Coking Reactivity of Laboratory-Scale Unit for Two Heavy Petroleum

Apr 2, 2013 - For a more comprehensive list of citations to this article, users are encouraged to ... Analysis: Computational-Intelligence-Based Model...
0 downloads 0 Views 1MB Size
Article pubs.acs.org/IECR

Coking Reactivity of Laboratory-Scale Unit for Two Heavy Petroleum and Their Supercritical Fluid Extraction Subfractions Linzhou Zhang, Shuyun Li, Ling Han, Xuewen Sun, Zhiming Xu, Quan Shi, Chunming Xu, and Suoqi Zhao* State Key Laboratory of Heavy Oil Processing, China University of Petroleum, Beijing P. R China, 102249 ABSTRACT: The building of a separation-reaction network for heavy feedstock requires an understanding of the reactivity for subfractions. This paper proposes a study on the coking reactivity for two heavy petroleum samples, vacuum residua (VR) from China Liaohe and Venezuela Orinoco crude oils, and their supercritical fluid extraction fractionation (SFEF) subfractions. The properties of feedstocks and their SFEF series were analyzed, including density, molecular weight, elemental content, Conradson carbon residue (CCR), SARA components (saturates, aromatics, resins and asphaltenes) and structure parameters by nuclear magnetic resonance (NMR). All the samples were subjected to a laboratory-scale batch reactor to investigate the coking reactivity. Coke yield increased as the SFEF subfraction became heavier; while the yields of naphtha, diesel, and gas oil all decreased. The results show that the classic linear correlation between coke yield and CCR are not appropriate for separated fractions of VR. Instead, a negative power function was observed. Furthermore, the coke yield shows a linear relationship with aromaticity for both SFEF series. To precisely estimate coking reactivity for feedstocks and subfractions, a new SARA component based prediction model was carried out, which reveals the contribution of each component.



INTRODUCTION

carbon ratio (H/C), kinetic viscosity, and density have been considered. Besides the mathematical correlation between the product yield and properties, the impact of overall structural parameters of vacuum residua has been studied. Characterization of feed and all delayed coking products were reported using structure data calculated from nuclear magnetic resonance (NMR) results.8,9 A consistent classification was reported to obtain the total distribution of carbon and proton atoms of products in the delayed coking process. Changes in carbon and hydrogen distributions (products versus feedstock) reveal the nature of the thermal cracking reactions. The main change is a substantial increase in the number of aromatic and paraffinic carbons at the expense of naphthenic carbons. Chen et al.10 also found that aromatic carbons increase during coking, the origin was considered from naphthenes. Liang et al.11 measured the yield of 17 feedstocks. A number of stepwise regression models were established for the coking yield of gases, naphtha, light gasoil, heavy gasoil, and coke. CCR and metals were found sensitive for coke yield. The contribution of molecular weight, CCR, H/C, hydrocarbon type composition, and metal content were different to gases, naphtha, light gasoil, and heavy gas oil. Saturates, aromatics, resins, and asphaltene (SARA) components, separated by solubility and polarity, are popular subjects for studies in many fields concerning petroleum.13−16 Their conversion behaviors during the coking reaction were also investigated. Favre et al.12 carried thermal conversion at 420 °C for Safaniya residue and found that saturates and resins are more reactive. Gray et al.13,14 proposed detailed studies on

Delayed coking is an important process for petroleum residue upgrading, especially poor quality residue with high metals and asphaltenes, which are not appropriate for catalytic upgrading. Characterizing the composition of feed and product would be valuable for understanding the dependence of coking reactivity on the structure or property. A detailed review has been published by Sawarkar et al. recently.1 Several modeling approaches have been conducted to predict the coking products yields. Gary2 established a number of correlations between coking yields to Conradson carbon residue (CCR) and API gravity for the straight-run distillation residues, which has been widely used in refineries. Maples et al.3 found that the coke yield, C5+ oil, and gas oil (diesel plus coking gas oil) yield correlated well with CCR. Chen et al.4 pointed out that the coke yield was of linear dependence on the CCR content, and the slope will vary from 0.9 to 1.66. These works have attempted to correlate the coking product (especially coking yield) with CCR based on the results from simulated mircoscale pyrolysis instruments. However, the dependence of gas yields with CCR was fair, and no correlation could be found for gasoline with CCR.3 On the other hand, many studies introduced multibulk properties in the models to cover more aspects of the feedstock. Schabron et al.5 used a correlation factor to predict coking yield, in which the contribution of molecular weight in toluene, the fractional heteroatom content, and weight percentage of the asphaltenes were included. A coking index6 and a dispersed particle solution model7 were also proposed too. Fan et al.8 took the coking reactivity of five Chinese vacuum residua and combined them with published experimental results to propose a correlation of coking gases, liquid, and coke yields with a characteristic parameter KZ. The influences of hydrogen over © 2013 American Chemical Society

Received: Revised: Accepted: Published: 5593

October 23, 2012 March 23, 2013 April 2, 2013 April 2, 2013 dx.doi.org/10.1021/ie302891b | Ind. Eng. Chem. Res. 2013, 52, 5593−5600

Industrial & Engineering Chemistry Research

Article

carried out in a laboratory-scale coking unit, and the products, including gases, liquids, and coke were collected. The coking product yields of all subfractions and their relationships with properties and structure parameters have been discussed.

the structure and thermal converstion of asphaltenes. Guo et al.15 performed coking reactivity of a residue and its four groups by thermogravimetry. The data showed that saturates also contribute to coke formation, with a yield of 5.68 wt % of total saturates, and the yields for aromatics, resins, and asphaltenes are 15.38, 32.44, and 53.71 wt %, respectively. They explained that saturates in vacuum residua have strong hydrogen abstraction abilities to deplete transferable hydrogen in hydroaromatic structures, resulting in radicals of the coke precursor uncapped and concentrated, thus accelerating coke formation.16 These works seem in contrary to view points of Long et al.17 and Roberts et al.,18 which stated that MCR are additive. Although the literature gives investigation and characterization on coking behavior in many aspects, most of them only focus on the overall residue. The deep separation subfractions were less discussed because of the limitations of the separation method and the practical industrial application. Elliot19 pointed out deep distillation can reduce total coke yield for deep distillation residue. Rahimi et al.20 studied the reactivity of Athabasca bitumen resid fractions and found the coke yield based on residue increases with boiling point. However, as the cut boiling point increased, the amount of the residue decreased. So, if the coke yield was calculated from the total feedstock (distillates and residue), it decreased with the rise of cutting temperature. These findings indicate that processing higher boiling point feedstocks can be achieved without generating larger amounts of coke. On the basis of their different solubilities in petroleum, supercritical fluid extraction and fractionation (SFEF) separates heavy petroleum into fractions.21,22 This technology was first developed as a laboratory-scale separation method and provided approximately a dozen of extracted narrow subfractions and a nonextractable end-cut for heavy oil characterization. The deep separation facilitates the characterization and reactivity of the subfractions of a heavy composition.23 Hydrodesulfurization (HDS), hydrodenitrogenation (HDN), and fluid catalytic cracking (FCC) of several SFEF subfractions have been reported in previous papers.24,25 Moreover, a classification index KH has been proposed to qualify the secondary upgrading ability of the separated subfractions.26 However, till now, the coking reactivity and correlation of SFEF subfractions have not been systematically investigated. The coking product prediction model for SFEF subfractions is also required by the progression in the industrial-scale supercritical based solvent separation process. Recently, a demonstration unit named SELEX-asp process was built by our group and PetroChina, which attempts to separate heavy residua by supercritical extraction and upgrading the fractions with different processes.27 The determination of cutting point and posted upgrading process requires a practical model for product calculation. The result based on whole residue may be not applicapable for SFEF subfractions. Moreover, the understanding of coking reactivity of these fractions will be helpful to find the key component that affects the coking or expected product yield and gives the possibility of removing or changing them by changing the SFEF pressure or temperatures. To study the coking reactivity of SFEF subfractions and build a relative coking product prediction model, we separate two vacuum residua, which were derived from a low sulfur content crude oil from China and a high sulfur content crude oil from Venezuela, respectively. The feedstocks were separated into a number of fractions by SFEF. The coking reactions were



EXPERIMENTAL SECTION Feedstock and Separation Procedure. Vacuum residues were derived from China Liaohe and Venezuela Orinoco crude oils. The preparation of narrow-cuts from feedstock by SFEF has been described elsewhere.21,22 The supercritical solvent was n-pentane for both two feedstocks. The selection of supercritical solvent was based on its solubility and feedstock property. In summary, feedstock (∼1000 g) was heated and pumped into the SFEF unit. The extraction and fractionation sections of the SFEF unit were heated to 200 and 220 °C, respectively. The pressure in the SFEF unit was initially set at 4 MPa and then increased to 12 MPa at a rate of 1 MPa/h. A series of narrow-cut samples with yield 5 wt % to feed (∼50 g) were collected sequentially. The SFEF end-cut was the remaining fraction that was not extractable with supercritical solvent. Three runs have been repeated to collect sufficient narrow cuts for each residue. The corresponding cuts were blended and then subjected to coking reaction and characterization. Coking Reaction. A 200 mL batch reactor with shaft riser was set up as seen in the flow sheet in Figure 1. The shaft can

Figure 1. Flow sheet of batch coking reactor.

take the reactor into and out of the furnace for quickly increasing or cooling down the temperature of the reactor. At first, a 100 g sample was sealed in the reactor and then purged by nitrogen. The furnace temperature was preset and statibilized before the reactor was inserted into the furnace to give a quick reaction temperature rise. Coking reaction was carried out at 500 °C in atmospheric pressure. The reaction time has been predetermined at 1 h, which can accommodate the complete coking reaction, and this time was applied for all the cuts of two feedstocks. After ending the reaction, the reactor was raised quickly from the furnace to cool down. Coke was collected from the batch reactor after the reaction and then fluxed by toluene for 2 h. 5594

dx.doi.org/10.1021/ie302891b | Ind. Eng. Chem. Res. 2013, 52, 5593−5600

Industrial & Engineering Chemistry Research

Article

The coke was then removed by filtration on a paper filter and the yield was calculated by weighting the filter after drying in a vacuum. The toluene was removed from the liquid by rotor evaporator, after which, the liquid was added to those evaporated from the reactor. The gas product was washed by NaOH solution to trap H2S. H2S content was caculated by titration. The volume of gases was determined by a water replacement method and the weight yield was calculated based on the composition using the ideal gas equation of state. The obtained liquids were assigned to naphtha (350 °C) based on the simulated distillation using the AC (Analytical Contral Co., Netherlands) high-temperature simulate distillation system. Property and Compositional Analysis. SARA analysis was carried out by using the American Society for Testing and Materials (ASTM) D2007-11 standard. The number average molecular weight of feedstock and SFEF subfractions were determined by using a Knauer K-7000 vapor pressure osmometer at 60 °C with toluene as solvent. The tested concentration ranged from 5 to 30 g/L. Elemental analysis was conducted by using Flash EA 1112 and Antek 7000 element analyzers. The metal elements contents were analyzed by VistaPRO simultaneous ICP−OES. The density is determined by pycnometer. CCR is analyzed by YT-30011 equipment produced by Shanghai Yutong Instrument. Proton nuclear magnetic resonance (NMR) measurements were performed with a Varian Unity Inova 500 MHz spectrometer, fitted with a 5 mm, 4-nucleus probe at 21.5 °C. Each sample (0.1 g) was dissolved in deuterated chloroform (1.0 mL) and placed in a 5 mL NMR tube. A solution of tetramethylsilane in chloroform (15 mL per 100 g) was used as a reference. Spectra were acquired at a sweep-width of 8000 Hz, a pulse width of 1.9 μs. Pulse acquisition time of 5 s and a relaxation delay of 10 s at an exact frequency of 500.0 Hz. Bulk property (elemental analysis, density, and CCR) for feedstock and subfraction 1, 4, 7, 10, 13 and end-cut of Orinoco VR have been previously published in the molecular level study on polar heteroatom species on these samples.28

Table 1. Property of Liaohe VR and Orinoco VR property CCR, wt % density (20 °C), g/cm3 molecular weight (VPO), g/mol SARA, wt %

content of elements, wt %

H/C atomic content of metal elements, wppm

saturates aromatics resins asphaltenes C H S N Ni V Ca Fe

Liaohe VR

Orinoco VR

18.11 1.0079 1587 16.43 34.91 41.12 5.61 86.73 10.58 0.50 1.51 1.5 155.0 3.5 104.0 325.0

26.19 1.0524 1152 7.30 32.36 37.23 14.58 82.69 9.68 4.8 1.09 1.4 175.5 751.7 12.8 16.7

Figure 2. Yield−pressure curves for Orinoco and Liaohe vacuum residua in supercritical fluid extraction.

the subfractions. Both density and NMW of the subfraction increased while the H/C went down as the extraction pressure rose. Those properties for the extracted SFEF subfractions varied slightly, but a significant gap in properties was found between the end-cut and extracted subfractions. Two feedstocks and their subfractions with various properties provided a series of samples for the study on coking reactivity. Coking Reactivity and Its Dependence on CCR Content. Two VRs and the SFEF subfractions were subjected to a laboratory-scale batch coking reactor to perform coking reactions. Table 2 lists the products yields of Liaohe VR and Orinoco VR. Both feedstocks generated around 30 wt % coke and nearly 50 wt % total liquids. Less coking yield and more total liquid yield of Liaohe VR implies that it has better coking reactivity than Orinoco VR. Products yields of SFEF subfractions were shown in Figure 4. Coke yield increases and total liquid yield decreases as the subfraction gets heavier. The yields of naphtha, diesel, and gas oil all showed downward trends. Both end-cuts showed poor coking reactivities with coke yields higher than 50 wt %. CCR has for a long time been considered as the parameter to estimate the coke yield of delayed coking.2 Figure 5 shows the CCR contents as a function of all the SFEF subfractions. It is clear that the CCR content increased as the SFEF subfraction became heavier, which was in agreement with the increase trend of the coke yield. It could be calculated that the ratio is 1.77 for Liaohe VR and 1.34 for Orinoco VR. Coke yields and



RESULTS AND DISCUSSION Properties of Feedstock and SFEF Subfratcion. Properties of Liaohe VR and Orinoco VR were listed in Table 1. It can be seen that both feedstocks have high CCR and density over 1 g/cm3. Both materials have similar nickel content (155.0 wppm and 175.5 wppm for Liaohe VR and Orinoco VR, respectively), but their vanadium contents were quite different (751.7 wppm for Orinoco VR and only 3.5 wppm for Liaohe VR), suggesting their difference in origin. Liao VR has asphaltenes sum up to 5.61 wt %, while that of Orinoco VR was much higher, with a value of 14.58 wt %. In addition, Orinoco VR has much higher sulfur content (4.8 wt %) than that of Liaohe VR (0.5 wt %). Figure 2 shows yields of extractable subfractions as a function of pressure for both feedstocks. The extractable subfractions were collected with a yield of 5 wt %. For Liaohe VR, 14 extractable subfractions were obtained with a total yield of 69.6 wt %, while for Orinoco VR, extractable subfractions were 13, and total yield was 63.82 wt %. End-cuts, which were nonextractable fractions during SFEF, were obtained from the bottom of the extract kettle and formed solid materials in room temperature. After separation, an obvious trend in the properties could be observed. Figure 3 illustrates the densities, H/C, and NMW of 5595

dx.doi.org/10.1021/ie302891b | Ind. Eng. Chem. Res. 2013, 52, 5593−5600

Industrial & Engineering Chemistry Research

Article

Figure 3. Property of SFEF subfractions for Liaohe VR and Orinoco VR: (left) density; (middle) hydrogen to carbon atomic ratio; (right) number average molecular weight.

values of coke/CCR vary in a wide range (from around 1 to 5). Coke/CCR values for the first SFEF subfractions (SFEF 1) from Liaohe VR and Orinoco VR were 4.97 and 3.77, respectively. The values were far away from the previous value, indicating that the coke yield is not linearly dependent to the CCR content. CCR analysis was performed in a microreactor (crucible), which loaded only ∼5 g feedstock. In other words, making a comparison between the CCR and the coke yield from the laboratory scale coking unit was comparing the coking yield from two different reactors. The light fraction in the feed and cracking products could easily have escaped from the liquid phase in CCR analysis due to the large surface area. On the contrary, the laboratory scale coking unit loaded more feedstock (∼100 g), and the light fraction and cracking products could only vaporize through the upper part of the reactor. Previous works on thermal pyrolysis of VR films in different thicknesses have revealed the effect of mass transfer during the coking of heavy petroleum.29−31 It could be assumed that mass transfer in the liquid phase during CCR analysis forms less impact on the coke yield than the laboratory scale

Table 2. Coking Products Yields for Two Feedstocks product

Liao VR

Orinoco VR

gases, wt % total liquid, wt % naphtha, wt % diesel, wt % gas oil, wt % coke, wt %

9.4 58.5 13.9 25.5 19.2 32.0

13.3 52.3 15.8 23.1 12.7 35.2

CCR for both end-cuts were similar, and the ratios were 1.04 and 1.06 for Liaohe end-cut and Orinoco end-cut, respectively. Those values show a deviation from the previous determined value (8.5−1.6) for delayed coking,4 but not quite different in value. The coke yield over CCR ratios showed significant difference for light SFEF subfractions. Figure 6 (top) showed the plot of coke yield as a function of CCR content. The coke yields raised as CCR increased but not in clear linear relationship for both SFEF series. The coke/CCR values were calculated and summarized in Figure 6 (bottom). It can been seen that the

Figure 4. Coking products yields of SFEF subfractions for Liaohe VR and Orinoco VR. 5596

dx.doi.org/10.1021/ie302891b | Ind. Eng. Chem. Res. 2013, 52, 5593−5600

Industrial & Engineering Chemistry Research

Article

they quickly evaporated in the crucible. But in the laboratory scale unit, those products were trapped in the liquid phase and the evaporating rate was controlled by the mass transfer. Thus, a secondary reaction (such as polycondensation) was promoted, which led to higher coke yield. As the extracted pressure increased, the light products from the derived SFEF cut were reduced and the difference between two reactors decreased. That is why the value of coke/CCR went down when the SFEF fraction got heavier. Here, we regressed the coke/CCR vs CCR data using a power function with negative exponent for both VR SFEF cuts as following: coke%/CCR% = a + b × (CCR%)c

(1)

where a, b, and c are 0.7914, 8.4356, and −0.8248, respectively. The regression correlation coefficient was 0.9847. The line shown in Figure 6 (red dash line) also revealed the good fitting. These results indicated that linear correlation with CCR is not sufficient for coke yield predication. CCR may be an average reactivity indicator for a bulk VR, but not for its fractions. The coking products correlation will be discussed in the following paragraphs. Influence of Structural Parameters on Coking Reactivity. The coking reactivity was related to their structure parameters. Figure 7 shows the coke yield and total liquid yield

Figure 5. CCR content of SFEF subfractions for Liaohe VR and Orinoco VR.

Figure 7. Coking product yield as function of H/C atomic ratio for two VR and corresponding SFEF series.

as a function of H/C. For all the samples, the coke yields and total liquid yields show a different relationship to H/C. As H/C increased, the total liquid yield rose while the coke yield went down. For Liaohe and Orinoco SFEF extracted subfractions, linear relationships were observed. But at the same H/C atomic ratio, there was a higher coke yield but lower total liquid yield for Liaohe VR than for Orinoco VR. H/C was inversely proportional to the aromaticity of the compounds. Moreover, as indicated in the literature,4 aromaticity was a more general parameter for coke yield. To investigate the dependence of the coke yields upon the aromaticity of material, we calculated the structural parameters using the modified Brown−Ladner method based on 1H NMR data.32 Figure 8 shows the yields of aromatic, naphthenic, and paraffinic carbons for both VR and SFEF series. A gradual increasing trend of the aromatic carbon fraction and a downward trend of the paraffinic carbon fraction were

Figure 6. Relationship between CCR content and coking yield for two VR and corresponding SFEF series. (Top) Plot of coking yield versus CCR content. (Bottom) plot of coking yield/CCR content versus CCR content.

unit. For light SFEF fractions, when the thermal cracking reactions took place, most of the products were volitilizable and 5597

dx.doi.org/10.1021/ie302891b | Ind. Eng. Chem. Res. 2013, 52, 5593−5600

Industrial & Engineering Chemistry Research

Article

seen that the ratio was around 1.0 to 1.1 for bulk feedstocks. But the ratios were increased for heavier SFEF cuts. The front SFEF subfractions had lower ratios than 1.0. But for heavier subfractions and end-cuts the value exceeded 1.0. For heavier front cuts, the coke originated from other carbons beyond aromaticity. As the data shows in literature,10 50% to 60% aromatic attached naphthenes converted to aromatics, which in turn made the total aromatic carbon balance 130% to 135% in the case of no CGO recycle. It can also be seen more additional coke carbon was produced for Liaohe VR than for Orinoco VR. So during delayed coking the aromatic carbon of compounds will increased due to the contribution of naphthanes in saturates and aromatic attached naphthanes. Although the slopes of coke yield as a function of aromatic carbon fraction for two SFEF series were similar, the intercepts were quite different. It indicates that the linear correlation is only apparent and other parameters, such as the heteroatom compounds, may affect coking yields. The Orinoco VR had lower nitrogen content but much higher sulfur content than Liaohe VR. Most of nitrogen atoms were known to be located on the aromatic rings for natural petroleum compounds.33 Similarly, more than half of the sulfur containing groups were a thiophene unit.23 Those heteroatoms also contributed to the aromaticity of the heavy petroleum, but the contribution varied in value due to the corresponding different π electron density. SARA Component Based Coking Product Yield Prediction Model. Some researcher showed that FCC reactivity is well correlated with SARA components.24 We investigated the SARA components of feed and subfractions and proposed a relative predictive model. Figure 10 shows the Figure 8. Fractions of aromatic (fa), naphthenic (fn) and paraffinic (fp) carbons for Liaohe VR and Orinoco VR SFEF series calculated from 1 H NMR and H/C.

observed. On the other hand, the naphthenic carbon fraction did not vary significantly for both SFEF series. The aromatic carbon fraction for the Orinoco SFEF series was higher by 0.05 to 0.06, the naphthenic carbon fraction was higher by 0.04 to 0.08 than the corresponding subfractions of Liao VR, while that for the paraffinic carbon fraction was lower by 0.08 to 014. The coke yields of both SFEF series as function of aromatic carbon fraction were shown in Figure 9. It can be seen the coke yield increased linearly as aromatic carbon fraction increased. The data from Liaohe and Orinoco SFEF series were linear regressed individually (the dash lines). A comparison of the ratio of coke yield to 100 times the aromatic factor, it can be

Figure 9. Coke yield as function of aromatic factor for all samples. A clear linear relationship was observed.

Figure 10. SARA components of Liaohe and Orinoco SFEF series. 5598

dx.doi.org/10.1021/ie302891b | Ind. Eng. Chem. Res. 2013, 52, 5593−5600

Industrial & Engineering Chemistry Research

Article

contributed least to coke yield, followed by aromatics. It was in agreement with literature15,16 that aromatics contribute coke more than saturates if there are individual reactions. The total liquids were mainly from saturates, followed by aromatics. The contribution of the resins and asphaltenes to total liquid yields were much lower. A similar trend could also be observed from the naphtha, diesel, and gas oil which was assigned from the collected liquid. It seems that the gas yields were most relative to the content of the aromatics and asphaltenes. The reason for this phenomena is not clear due to the complex reactions concerning gas formation. The SFEF separated the feedstock based on solubility, and thus generated a series of subfractions that varies in polarity and aromaticity. The proposed model could be used to estimate the coke reactivity of subfractions, which will help to decide the cutting point for practical industrial plant and optimize the separation-reaction networks.

SARA components of two SFEF series. Saturates dropped and resin rose, while aromatics increased to a maximum then decreased, and asphaltene contents were generally below 1 wt %. As the SFEF subfraction goes heavier, the polarity increases. It can be calculated that over 95% asphaltene was enriched in the end-cut. The end-cuts from both VR were very low in saturates contents. We proposed a first order correlation of the product yields upon the SARA component. The form is as follows: y% = a × sat% + b × ar% + c × re% + d × asp%

(2)

Where y %, sat %, ar %, re % and asp % are weight percent of product, saturates, aromatics, resins, and asphaltenes, and a, b, c, and d are their proportional coefficients, respectively. The coefficients represented the contribution to product yields in the first order. All the data were used in regression using leastsquares method. The results were listed in the Table 3. The plots of experimental and predicted data were comprised in Figure 11, which shows good prediction ability of the model.



CONCLUSION Two vacuum residua, Liaohe VR and Orinoco VR were separated into a number of subfractions with each yield of ∼5 wt % and nonextractable end-cuts by supercritical fluid extraction and fractionation (SFEF). The density, molecular weight, and CCR show an increasing trend, while hydrogen over the carbon ratio decreases as the SFEF subfraction becomes heavier. Coke yield increases as subfractions become heavier and total liquid yield decreases, in which naphtha, diesel, and gas oil all decrease. The end-cuts show poor coking reactivity with high coke yield for both series. The classic linear relationship between coke yield and CCR is no longer valid for the SFEF series. The coke yield ratio over CCR shows a decreasing trend, obeying a power function with a negative exponent for both VR SFEF cuts. It indicates that CCR is not sufficient for coke yield prediction for VR subfractions. The structural parameters obtained from 1H NMR show that the aromatic carbons strongly affect the coking reactivity. But two feedstocks show a difference in linear regression intercepts,

Table 3. Parameters and Average Absolute Deviation (AAD) of the SARA Component Based Coking Product Prediction Model gas total liquid naphtha diesel gas oil coke

a

b

c

d

AAD %

0.03 0.91 0.19 0.35 0.37 0.06

0.19 0.69 0.18 0.29 0.22 0.15

0.07 0.38 0.11 0.16 0.09 0.56

0.16 0.31 0.12 0.15 0.04 0.57

7.8 2.4 9.5 3.5 7.6 9.8

Resins and asphaltenes were generally believed to be responsible for the coke formation. Their coefficients were 0.56 and 0.57 in the model, respectively, which were much higher than those of saturates and aromatics. Saturates

Figure 11. Comparison between experimental data and data from SARA component-based prediction model. 5599

dx.doi.org/10.1021/ie302891b | Ind. Eng. Chem. Res. 2013, 52, 5593−5600

Industrial & Engineering Chemistry Research

Article

(18) Roberts, I. The chemical significance of carbon residue data; American Chemical Society, Division of Petroleum Chemistry Preprints: Washington, DC, 1989; Vol. 34. (19) Elliott, J. Maximize Distillate Liquid Products. Hydrocarbon Process., U.S. 1992, 71, 75−82. (20) Rahimi, P.; Gentzis, T.; Taylor, E.; Carson, D.; Nowlan, V.; Cotte, E. The impact of cut point on the processability of Athabasca bitumen. Fuel 2001, 80, 1147. (21) Shi, T.-P.; Hu, Y.-X.; Xu, Z.-M.; Su, T.; Wang, R.-A. Characterizing petroleum vacuum residue by supercritical fluid extraction and fractionation. Ind. Eng. Chem. Res. 1997, 36, 3988. (22) Yang, G.; Wang, R. A. The supercritical fluid extractive fractionation and the characterization of heavy oils and petroleum residua. J. Petrol. Sci. Eng. 1999, 22, 47. (23) Zhao, S.; Sparks, B. D.; Kotlyar, L. S.; Chung, K. H. Reactivity of sulphur species in bitumen pitch and residua during fluid coking and hydrocracking. Petrol. Sci. Technol. 2002, 20, 1071. (24) Xu, C.; Gao, J.; Zhao, S.; Lin, S. Correlation between feedstock SARA components and FCC product yields. Fuel 2005, 84, 669. (25) Yang, C.; Du, F.; Zheng, H.; Chung, K. H. Hydroconversion characteristics and kinetics of residue narrow fractions. Fuel 2005, 84, 675. (26) Shi, T.-P.; Xu, Z.-M.; Cheng, M.; Hu, Y.-X.; Wang, R.-A. Characterization index for vacuum residua and their subfractions. Energy Fuel 1999, 13, 871. (27) Zhao, S.; Xu, C.; Sun, X.; Chung, K. H.; Xiang, Y. China refinery tests asphaltenes extraction process. Oil Gas. J. 2010, 108, 7. (28) Zhang, L.; Xu, Z.; Shi, Q.; Sun, X.; Zhang, N.; Zhang, Y.; Chung, K. H.; Xu, C.; Zhao, S. Molecular characterization of polar heteroatom species in Venezuela Orinoco petroleum vacuum residue and its supercritical fluid extraction subfractions. Energy Fuel 2012, 26, 5795. (29) Gray, M. R.; Le, T.; McCaffrey, W. C.; Berruti, F.; Soundararajan, S.; Chan, E.; Huq, I.; Thorne, C. Coupling of mass transfer and reaction in coking of thin films of an Athabasca vacuum residue. Ind. Eng. Chem. Res. 2001, 40, 3317. (30) Gray, M. R.; McCaffrey, W. C.; Huq, I.; Le, T. Kinetics of cracking and devolatilization during coking of athabasca residues. Ind. Eng. Chem. Res. 2004, 43, 5438. (31) Radmanesh, R.; Chan, E.; Gray, M. R. Modeling of mass transfer and thermal cracking during the coking of Athabasca residues. Chem. Eng. Sci. 2008, 63, 1683. (32) Zhao, S.; Kotlyar, L.; Woods, J.; Sparks, B.; Chung, K. Molecular nature of Athabasca bitumen. Petrol. Sci. Technol. 2000, 18, 587. (33) McKenna, A. M.; Purcell, J. M.; Rodgers, R. P.; Marshall, A. G. Heavy petroleum composition. 1. Exhaustive compositional analysis of Athabasca bitumen HVGO distillates by Fourier transform ion cyclotron resonance mass spectrometry: A definitive test of the Boduszynski model. Energy Fuel 2010, 24, 2929.

which may be caused by the different heteroatom distribution of the feed and subfractions. A SARA component-based first order coking product prediction model was proposed. The model reveals that coke yields are mostly from resins and asphaltenes and liquid products are more dependent on saturates and aromatics.



AUTHOR INFORMATION

Corresponding Author

*Tel.: +86-10-8973-9015. E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported by the National Basic Research Program of China (2010CB226901), the Union Fund of National Natural Science Foundation of China (NSFC), China National Petroleum Corporation (CNPC) (U1162204) and NSFC fund (21176254).



REFERENCES

(1) Sawarkar, A. N.; Pandit, A. B.; Samant, S. D.; Joshi, J. B. Petroleum residue upgrading via delayed coking: A review. Can. J. Chem. Eng. 2007, 85, 1. (2) Gary, J. H.; Handwerk, G. E. Petroleum Refining: Technology and Economics; M. Dekker: New York, 1975. (3) Maples, R. E. Petroleum Refinery Process Economics; PennWell Corporation: Tulsa OK, 2000. (4) Chen, J. Changes of carbon and hydrogen composition in petroleum refining and their efficient utilization. Acta Pet. Sin. 1982, Vol. 2. (5) Schabron, J., F.; Speight, J., G. Évaluation du rendement en produit de cokéfaction différée de petrole lourd á l'aide de la teneur an asphalténes et du résidu de coke. Oil Gas. Sci. Technol. 1997, 52, 73. (6) Schabron, J. F.; Pauli, A. T.; Rovani, J. F., Jr; Miknis, F. P. Predicting coke formation tendencies. Fuel 2001, 80, 1435. (7) Schabron, J. F.; Pauli, A. T.; Rovani, J. F., Jr. Residua coke formation predictability maps. Fuel 2002, 81, 2227. (8) Rodriguez, J.; Tierney, J. W.; Wender, I. Evaluation of a delayed coking process by 1H and 13C NMR spectroscopy: 1. Material balances. Fuel 1994, 73, 1863. (9) Rodriguez, J.; Tierney, J. W.; Wender, I. Evaluation of a delayed coking process by 1H and 13C NMR spectroscopy: 2. Detailed interpretation of liquid NMR spectra. Fuel 1994, 73, 1870. (10) Chen, J.; Cao, H. Changes of chemical structures of heavy oils during processing and their conversion. Pet. Refin. Eng. 1994, 6, 9. (11) Liang, C.; Shen, B.; Liu, J.; Chen, X. Prediction for product distribution of delayed-coking process by stepwise regression model. J. E. China. Univ. Sci. Technol. 2009, 2, 7. (12) Favre, A.; Boulet, R.; Behar, F. Etude par simulation en laboratoire de l'operation de viscoréduction. Oil Gas. Sci. Technol. 1985, 40, 609. (13) Gray, M. R. Consistency of asphaltene chemical structures with pyrolysis and coking behavior. Energy Fuel 2003, 17, 1566. (14) Rahmani, S.; McCaffrey, W. C.; Dettman, H. D.; Gray, M. R. Coking kinetics of asphaltenes as a function of chemical structure. Energy Fuel 2003, 17, 1048. (15) Guo, A.; Zhang, H.; Yu, D.; Wang, Z. Coking performance of residue and its four groups by thermogravimetric analysis. Pet. Process. Chem. 2002, 7, 5. (16) Guo, A.; Wang, Z.; Que, G. Promoting effect of saturate hydrocarbons in initial coke formation from petroleum residua under thermal cracking. J. Fuel. Chem. Technol. 2001, 5, 5. (17) Long, R. B.; Speight, J. G. Studies in petroleum composition. Development of a compositional map for various feedstocks. Oil Gas. Sci. Technol. 1989, 44, 13. 5600

dx.doi.org/10.1021/ie302891b | Ind. Eng. Chem. Res. 2013, 52, 5593−5600