Kinetic Modeling of Bitumen Hydroprocessing at In-Reservoir

ARTICLE pubs.acs.org/EF

Kinetic Modeling of Bitumen Hydroprocessing at In-Reservoir Conditions Employing Ultradispersed Catalysts Herbert Loria,* Gustavo Trujillo-Ferrer, Clementina Sosa-Stull, and Pedro Pereira-Almao Department of Chemical and Petroleum Engineering, Schulich School of Engineering, University of Calgary, Calgary, AB T2N 1N4 ABSTRACT: Heavy oil and bitumen’s low American Petroleum Institute (API) gravity and high viscosity make the economics of their industrialization difficult. Therefore, new recovery techniques must be developed to enhance these materials. Ultradispersed catalytic hydroprocessing of heavy oil and bitumen has been proposed as one of these novel techniques and has been tested in laboratories and pilot plants. In this work, a kinetic model for ultradispersed catalytic hydroprocessing of bitumen is proposed. Kinetic parameters were estimated from experimental results obtained in a tubular pilot plant reactor. The predicted products composition was in good agreement with experimental values with average absolute errors of less than 7%. Additionally, experimental liquid products viscosity and residue conversions followed an exponential correlation. This correlation, in combination with the results from the kinetic modeling, was employed to create a computational program that calculates products distribution from bitumen hydroprocessing and provides reaction conditions to achieve a specific liquid products viscosity.

1. INTRODUCTION At present, world resources of heavy oil and bitumen are estimated to be 5.6 trillion barrels, compared with the rapidly depleting remaining light conventional crude oil reserves of 1.02 trillion barrels.1 However, these vast unconventional resources of petroleum remain untapped, and most of these deposits are in North America.2 Consequently, production and processing of heavy oils and bitumen must be increased. These oils are more viscous than conventional oils due to the abundance of large heavy hydrocarbons which impose transportation problems and poor recovery potential. They also contain a significant fraction of heteroatoms, such as sulfur, nitrogen, and metals, which make processing more difficult and limit marketability. Therefore, in order to make the exploitation of heavy crude oil and bitumen reserves more economically and environmentally viable, more advanced technologies, including more efficient catalysts for heavy oil and bitumen upgrading, are required. Upgrading is generally defined as any fractionation or chemical treatment of bitumen or heavy oil that increases its value. Therefore, its minimum objective is to reduce the viscosity of the oil allowing transportation by pipeline without adding a solvent, and its maximum objective is to obtain a higher quality crude oil (“a synthetic crude oil”).3 Upgrading of heavy fractions can be accomplished in two different ways. The first, by carbon removal, is a thermal process, which consists of the formation of a solid carbonaceous product called coke to improve the hydrogen to carbon ratio of liquid products. However, the coke product has to be disposed of, becoming a major problem for upgrading facilities. The second way of upgrading is by hydrogen addition, a thermal catalytic process comprised of incorporation of hydrogen to the feed to elevate the hydrogen to carbon ratio and reduce the formation of coke.4 Typically, carbon removal has been the first choice for heavy oil upgrading; however, its high yields of coke and increasing environmental restrictions are making these technologies unsustainable. Therefore, hydroprocessing may become an option r 2011 American Chemical Society

to avoid coke formation in the upgrading of heavy fractions. Hydroprocessing reactions can be both destructive and nondestructive. Destructive hydroprocessing or hydrocracking is the conversion of compounds with higher molecular weight to lowerboiling point compounds.5 Carbon
carbon bonds break and are saturated with hydrogen. These processes require severe conditions (>400 °C) and high hydrogen partial pressure to minimize polymerization and coke formation. Nondestructive hydroprocessing or hydrotreating consists in hydrogenation reactions during which the quality of oil improves by removing certain contaminants from its molecular structure.5 Such contaminants are typically sulfur, nitrogen, and metals. In these reactions, the boiling point of the oil does not change considerably and the conditions are milder than those for destructive hydroprocessing reactions. Hydroprocessing catalysts are mostly made of transition metals supported on solid pellets made from ceramic materials. A major problem for supported catalysts is the deactivation which, depending on its mechanism, can be a physical and/or a chemical phenomenon that decreases the activity. The main mechanisms of deactivation are poisoning, which is the result of a strong chemisorption of impurities on the catalytic surface; coking, the formation of a carbonaceous residua on the catalyst that could block pores and catalytic sites; and sintering, which is the agglomeration of active sites leading to reduction of activity.6 In the case of heavy oils and bitumen, the hydroprocessing mechanism is similar to thermal cracking but having hydrogen transfer/hydrogenation superimposed, which will help to improve the quality of the upgraded product while decreasing the coke production.7 Ultradispersed catalysts have been studied for heavy oil and bitumen hydroprocessing as an alternative for typical supported catalysts. Some previous developments include Received: January 17, 2011 Revised: March 19, 2011 Published: March 23, 2011 1364

dx.doi.org/10.1021/ef200094h | Energy Fuels 2011, 25, 1364–1372

Energy & Fuels the use of oil-soluble compounds that decompose under reaction conditions and produce very small particles; these particles are able to locally generate hydrogen spillover and increase the upgrading of residual feedstocks.8 Ultradispersed catalysts, which do not require support and are formulated in reduced particle sizes (550 °C), 50.66 wt %. All the experiments were conducted at the following conditions: reaction temperatures of 320
380 °C and residence times of 9
51 h, keeping constant the total pressure and the hydrogen-to-oil ratio at 2.76 MPa (400 psi) and 625 standard (std) cm3/cm3, respectively. In the case of the emulsion of Athabasca bitumen and ultradispersed catalysts fed to the pilot plant, the amount of metallic precursors produced 1200 ppmw of metal with respect to the bitumen. The products distribution was analyzed by high-temperature simulated distillation carried out in a gas chromatographer from Agilent Technologies modified by Separation Systems, Inc. to estimate the amount of residue at 550 °Cþ using the procedure ASTM D7169-2005 described by Carbognani et al. 21 The hydroprocessing products were defined as gases, naphta (IBP
216 °C), distillates (216
343 °C), VGO (343
550 °C), and residue (>550 °C). The experimental results showed that the quantity of liquid product (naphta, distillates, VGO, and residue) that was obtained ranged between 94.6 and 99.5 wt %, whereas the amount of produced gases varied between 0.5 and 5.4 wt %, depending on the experimental conditions. Viscosity of the liquid products at 40 °C was determined using a Brookfield viscometer model DV-IIþ Pro,

ARTICLE

Figure 2. Proposed kinetic model for the ultradispersed catalytic hydroprocessing of bitumen.19

viscosities of the liquid products from the experimental results at 40 °C varied between 98 and 1655 cP, depending also on the experimental conditions.

3. KINETIC MODELING 3.1. Reaction Mechanism. Hydroprocessing reactions generally present three phases: solid (coke), liquid (unconverted residue and liquid products), and gas (hydrogen and product gases). In the case of all hydroprocessing experiments performed with ultradispersed catalysts in this work, only traces of coke were observed. This amount of coke was considered insignificant and, thus, coke was not taken into account in the reaction model and the fluid
solid mass transfer limitation involving coke was not considered in the mass balance. Excess of hydrogen allows a rapid attainment of equilibrium between the gas and liquid phase and keeps a constant hydrogen concentration in the liquid phase;22 this was the case of the experiments presented in this work. The rate of hydrogen transfer across the liquid
gas interface was also found to be negligible in the hydrogenation of coal liquids;22 similar conclusions were reached by Qader et al.23
26 based on the magnitude of the activation energies for hydrocracking reactions of coal liquids. Ancheyta et al.27 found that, in continuous gas oil hydroprocessing reactors, axial dispersion was very low and external mass transfer gradients can also be neglected. For the case of the kinetic modeling of the experiments carried out for this paper, internal mass transfer gradients are not taken into account because ultradispersed catalysts are considered to be unsupported nonporous materials.9 On the basis of the previous statements, reactions for hydroprocessing of bitumen can be assumed to be kinetically controlled. 3.2. Kinetic Model. The kinetic model employed in this paper is adapted from the high severity hydrocracking of heavy oils proposed by Sanchez et al.19 Figure 2 shows the proposed kinetic model. The model includes 5 lumps, unconverted residue (R), vacuum gas oil (VGO), distillates (D), naphta (N), and gases (G), and 10 reactions. For each reaction, a kinetic expression (ri) was formulated as a function of the lump weight percent (wt %) composition (yi) and kinetic rate constants (kn). All reactions were assumed to be first order since hydroprocessing of hydrocarbons is supposed to be of first order with 1366

dx.doi.org/10.1021/ef200094h |Energy Fuels 2011, 25, 1364–1372

Energy & Fuels

ARTICLE

respect to the hydrocarbon molecule.23
26 Product compositions were determined with pilot plant mass balances and distillation curves. The reaction rates are given by the following expressions: Residue :

rR ¼
ðk1 þ k2 þ k3 þ k4 ÞyR

ð1Þ

balance equation (eq 8) for each lump by means of integration in time intervals corresponding to each evaluated residence time. For each time interval (τ0 e τ e τj), eq 8 can be integrated in both sides as: Z τj Z yi ðτj Þ dyi ¼ ri dτ ð10Þ yi ðτ0 Þ

VGO : rVGO ¼ k1 yR
ðk5 þ k6 þ k7 ÞyVGO

ð2Þ

Distillates : rD ¼ k2 yR þ k5 yVGO
ðk8 þ k9 ÞyD

ð3Þ

With substitution of the reaction rate equations for each lump (eqs 1
5) in eq 8, the following set of equations can be found Z

Naphta :

rN ¼ k3 yR þ k6 yVGO þ k8 yD
k10 yN

ð4Þ

Residue : ¼
k1

ð5Þ

The dimensions for the variables present in eqs 1
5 were defined as ri = wt %/time, kn = 1/time, and yi = wt %, where the suffix i represents the different lumps, residue, VGO, distillates, naphta, and gases and the suffix n is the reaction number according to Figure 2. 3.3. Plug-Flow Reactor Model. The kinetic model was incorporated into a plug-flow reactor model. This model assumes an isothermal operation where reactions are assumed irreversible of the first-order. Axial dispersion and internal and external mass transfer gradients are neglected on the basis of the statements from section 3.1. With the use of these assumptions, the generic mass balance for each component i can be expressed by

τj τ0

Z VGO :

Z

yR dτ
k2

τj

τ0

Z yR dτ
k3

τj τ0

Z yR dτ
k4

τj

yR dτ

τ0

Z ¼ k1

yi ðτj Þ

dyVGO

yi ðτ0 Þ τj

τ0

Z

yR dτ
k5

Z

τj

yVGO dτ
k6

τ0

Z

τj

yVGO dτ
k7

τ0

τj

yVGO dτ

τ0

ð12Þ Z Distillates : Z

yi ðτj Þ yi ðτ0 Þ

τj τ0

ð6Þ

where Fi is the weight percent flow rate (F = {(wt%)(volume)}/ time) and V is the reactor volume. Since it is assumed that the experimental plug-flow reactor has a constant volumetric flow (Q = volume/time), then the weight percent flow rate (Fi) can be defined as Fi = Qyi. Now, the left side of eq 6 can be rewritten as

dyR

ð11Þ

¼ k2

dFi ¼ ri dV

yi ðτj Þ

yi ðτ0 Þ

Z

Gases : rG ¼ k4 yR þ k7 yVGO þ k9 yD þ k10 yN

τ0

dyD Z

yR dτ þ k5

τj

τ0

Z yVGO dτ
k8

τj

τ0

Z yD dτ
k9

τj

yD dτ

τ0

ð13Þ Z Naphta : Z ¼ k3

yi ðτj Þ

yi ðτ0 Þ τj

τ0

dyN Z

yR dτ þ k6

τj

τ0

Z yVGO dτ þ k8

τj

τ0

Z yD dτ
k10

τj

τ0

yN dτ

ð14Þ

dFi dyi dyi dyi ¼Q ¼ ¼ dV dV dðV =Q Þ dτ

ð7Þ

where τ is the residence time. The reactor mass balance for each component i thus becomes dyi ¼ ri dτ

Z Gases : Z ¼ k4

ð8Þ

yi ðτ0 Þ τj τ0

dyG Z

yR dτ þ k7

τj

τ0

Z yVGO dτ þ k9

τ0

Z yD dτ þ k10

Now, let Di and Si be defined as Z yi ðτj Þ Di ¼ dyi ¼ yi ðτj Þ
yi ðτ0 Þ ¼ yi, j
yi, 0 yi ðτ0 Þ

T

yi ð0Þ ¼ ½yR0 , yVGO0 , yD0 , yN0 , yG0 

Z

ð9Þ

where yi0 represents the initial weight percent distribution of each lump in the Athabasca bitumen. 3.4. Kinetic Rate Constants Estimation. Kinetic rate constants were estimated using a quasi-linearization method, the first step in this procedure was to reformulate the reactor mass

τj

τj

τ0

yN dτ

ð15Þ

Equation 8 is identical to the mass balance equation of a batch reactor, except that residence time replaces the batch reaction time. Initial conditions for eq 8 are given by

¼ ½50:66, 32:46, 13:48, 3:4, 0T

yi ðτj Þ

Si ¼

τj

τ0

yi dτ

ð16Þ

ð17Þ

where yi,0 and yi,j represent the weight percents of the lump i at the 0th and jth evaluated residence time, respectively. Since the weight percent differences and integrals from eqs 11
15 can be evaluated numerically, this procedure yields the following 1367

dx.doi.org/10.1021/ef200094h |Energy Fuels 2011, 25, 1364–1372

Energy & Fuels

ARTICLE

system of equations in matrix form:

3 2
SR DR 7 6 6 6 DVGO 7 6 SR 7 6 6 6 DD 7 ¼ 6 0 7 6 6 6 D 7 6 0 4 N 5 4 DRG 0 2


SR 0 SR 0 0


SR 0 0 SR 0


SR 0 0 0 SR

0
SVGO SVGO 0 0

The system given by eq 18 is overdetermined since [Si] is not a square matrix. In order to find the solution that is the closest with respect to experimental data, the least-squares method can be used.28 For the system [Di] = [Si] [kn] (eq 18), the least-squares solution is found when [Di]
[Si] [kn] = minimum. The least-squares method was implemented in a MATLAB computational program. In the present work, the kinetic rate constants were calculated from experimental data in the form of weight percent of the different lumps as a function of the residence time. Initial guesses for the kinetic rate constants were selected randomly. During the calculation procedure, each iteration was finished when |kn(lþ1)
kn(l)| (l = iteration number) became less than a set criterion and the model converged on the experimental data; otherwise, new guesses were made for the kinetic rate constants and the iteration continued. The model was said to converge when the following three criteria, described by K€oseoglu and Phillips,29 were achieved: (1) closeness of fit between experimental and model products weight percents (least-squares criterion). This criterion was achieved when a low performance index (PI) was obtained. The lower mod the PI, the closest the experimental (yexp i,j ) and model (yi,j ) products weight percents are. The performance index is defined as PI ¼

 2    exp mod  yi, j
yi, j   j¼1  m

∑i ∑

AAE% ¼

exp yi, j

N

 100

0
SVGO 0 SVGO 0

0
SVGO 0 0 SVGO

0 0
SD SD 0

0 0
SD 0 SD

4.1. Kinetic Rate Constants and Arrhenius Parameters. By employment of the previously described method, it was first found that kinetic rate constants corresponding to reactions producing gases (reactions 7, 9, and 10) were null (k7 = k9 = k10 = 0), except for reaction 4 (unconverted residue producing gases, k4 6¼ 0). From these results, it can be said that gas production exclusively comes from unconverted residue, at least in the range of temperatures for this study (320
380 °C). In addition, the fact that k10 = 0 means that naphta hydroprocessing is negligible at the reaction conditions employed in this work’s experiments. These results are in agreement with those obtained by Sanchez et al.;19 they found the same behavior for gas formation from severe hydrocracking of heavy oils at higher temperatures (380
420 °C). The proposed kinetic model from Figure 2 was modified following these results. A new kinetic model, in which reactions 7, 9, and 10 are not considered, was proposed in order to have a better accuracy in the calculation of the rest of the kinetic rate constants. The model was modified as shown in Figure 3. On the basis of these new considerations, the reaction rates of the new proposed model are

ð19Þ

ð20Þ

where N is the total of lumps weight percents obtained from the experiments at different residence times.

ð18Þ

4. RESULTS AND DISCUSSION

Residue : rR ¼
ðk1 þ k2 þ k3 þ k4 ÞyR

where m is the total number of evaluated residence times. (2) Since only irreversible reactions are taken into account, only positive kinetic rate constants must be obtained. (3) The kinetic rate constants must follow the Arrhenius law temperature dependence. The MATLAB program not only provided the convergence between experimental and model results but also yielded the calculated weight percent of each lump, the kinetic rate constant values, and the average absolute error percent (AAE%) between the experimental and model results. The AAE% is defined as      exp  yi, j
ymod  i , j m  

∑i j∑¼ 1

3 k1 6 7 6 k2 7 7 36 6 k3 7 6 7 0 7 6 k4 7 7 0 76 7 76 k 5 7 6 0 7 7 76 6 k6 7 7
SN 5 6 7 6 k7 7 6 7 SN 6 k8 7 6 7 6k 7 4 95 k10 2

VGO :

rVGO ¼ k1 yR
ðk5 þ k6 ÞyVGO

ð21Þ ð22Þ

Distillates : rD ¼ k2 yR þ k5 yVGO
k8 yD

ð23Þ

Naphta : rN ¼ k3 yR þ k6 yVGO þ k8 yD

ð24Þ

Gases : rG ¼ k4 yR

ð25Þ

Kinetic rate constants obtained from the reaction system depicted in Figure 3 and eqs 21
25 are shown in Table 1. Activation energies and frequency factors for each reaction were calculated employing the linearized Arrhenius equation: ln kn ¼ ln An


EAn RT

ð26Þ

where An and EAn are the frequency factor and activation energy, respectively, for each reaction n, R is the gas constant, and T is the reaction temperature in absolute units. The Arrhenius equation parameters, linear regression coefficients (rL2) for eq 26 and AAE % obtained at each temperature are also presented in Table 1. Experimental compositions and those obtained by solving eqs 21
25, employing the kn values given in Table 1, were 1368

dx.doi.org/10.1021/ef200094h |Energy Fuels 2011, 25, 1364–1372

Energy & Fuels

ARTICLE

Table 1. Kinetic Parameters for the Ultradispersed Catalytic Hydroprocessing of Bitumen kinetic rate constants, h
1 T, °C

k1

k2

k3

k4

k5

k6

k8

320

0.00214

0.00131

0.00030

0.00006

0.00670

0.00491

0.00105

0.02

350

0.00845

0.00610

0.00113

0.00073

0.00282

0.00141

0.00046

4.58

360

0.01324

0.00951

0.00251

0.00085

0.00181

0.00073

0.00032

6.61

380

0.02650

0.03093

0.01130

0.00902

0.00045

0.00014

0.00007

0.69

EA, kJ/mol

136

167

192

261

145

190

146

ln A (A, h
1)

21.47

27.24

30.63

43.01

21.59

29.60

20.15

rL2

0.9983

0.9949

0.9660

0.9683

0.9998

0.9952

0.9979

AAE%

Figure 5. Arrhenius plot for the different kinetic rate constants. Figure 3. Modified kinetic model for the ultradispersed catalytic hydroprocessing of bitumen.

Figure 4. Comparison between experimental and model compositions from the different lumps defined in the kinetic model.

compared with respect to a 45° straight line shown in Figure 4. The equations were solved using the fourth order Runge
Kutta method, and the solution of the equations was also part of the developed MATLAB program. In Figure 4, it is observed that the different lumps composition is quite well predicted since the AAE% is less than 7%, as it was

shown in Table 1. This indicates that the proposed kinetic model is adequate for the hydroprocessing of bitumen at moderate reaction conditions. Figure 5 shows the Arrhenius plot for all kinetic rate constants; as presented in Table 1, linear regression coefficients were close to the unit. Figure 5 illustrates the absence of deviations inside the limits of the temperature range that was studied in the experiments; these deviations typically suggest the appearance of external mass transfer limitations.30 Thus, it can be assumed that the proposed kinetic model is fully representative of the range of the studied temperatures and reactions are kinetically controlled and external mass transfer gradients can be neglected, justifying the assumptions made in section 3.1. The values of the activation energies for the proposed kinetic model, shown in Table 1, are within the range of those reported by Sanchez et al.19 (114
225 kJ/mol). However, Sanchez et al.19 employed for their calculations a heavy oil (22°API) lighter than Athabasca bitumen (11°API). Therefore, the effect of ultradispersed catalysts can be translated to the reduction of the energy uptake for bitumen upgrading, since activation energies for Athabasca bitumen hydroprocessing reactions are within the range of those required for hydrocracking reactions of lighter heavy oils. 4.2. Catalytic Activity. A preliminary kinetic evaluation of Athabasca bitumen hydroprocessing in a batch reactor, employing the same catalytic emulsion and similar operating conditions as those employed in this work, was previously performed by Galarraga et al.20 They presented two different kinetic models to 1369

dx.doi.org/10.1021/ef200094h |Energy Fuels 2011, 25, 1364–1372

Energy & Fuels

ARTICLE

Table 2. Catalytic Activity for the Different Lumps from the Kinetic Model at Different Operating Conditions catalytic activity,ai = (rPFR,i/rBatch,i) T, °C 320

350

380

t = τ, h

R

VGO

D

N

G

1

1.04

0.98

1.02

1.03

0.98

5

1.04

0.98

1.01

1.03

0.98

10 50

1.03 0.99

0.98 0.98

1.01 0.99

1.02 1.02

0.97 0.97

100

0.97

0.98

0.99

1.01

0.97

1

0.98

0.99

1.04

0.97

0.96

5

0.99

0.99

1.04

0.99

0.97

10

0.99

1.01

1.02

0.99

0.97

50

1.03

1.04

1.00

0.99

0.99

100

1.03

1.05

0.99

1.04

0.99

1

1.00

0.98

0.99

0.97

0.98

5

1.00

0.98

0.99

0.98

0.99

10

0.98

0.99

1.01

0.98

0.99

50

0.98

0.99

1.01

0.99

1.00

100

0.97

1.01

1.03

0.99

1.03

fit their experimental data. In one of them, they obtained their kinetic rate constants from the same reaction mechanism that was employed in this work (the 5 lumps kinetic model from Sanchez et al.19); however, their mass balance was based on a batch reactor model. In order to compare the performance of the ultradispersed catalysts from the previous study and the present one, the numeric catalytic activity at the same operating conditions was obtained. The catalytic activity30 (ai) is defined as the ratio of the reaction rates obtained from the plug-flow reactor model (rPFR,i) with respect to those calculated from the batch reactor model (rBatch,i): rPFR , i ai ¼ rBatch, i

ð27Þ

The numeric comparison between results from both models is possible because the plug-flow reactor model yielded a mass balance similar to one from a batch reactor model; as it was commented on in section 3.3. Since both models are mathematically similar, reaction rates calculated with the same value of residence (τ) and reaction time (t) can be directly compared. Table 2 shows the numeric catalytic activities for the different lumps presented in the reaction model at the same operating conditions. Table 2 illustrates that catalytic activities values for all presented cases are virtually 1, and this demonstrates that the performance of ultradispersed catalysts in pilot-plant tests is similar to the one observed in previous laboratory batch experiments. Thus, batch reactor experiments at the laboratory level can be used for preliminary evaluation of ultradispersed catalysts to be used at larger scales. 4.3. Liquid Products Viscosity Estimation. An interesting aspect to be studied is the correspondence between liquid products viscosity and residue conversion; since, according to Gray,3 the minimum objective in heavy oil upgrading is to reduce the viscosity of the oil to allow its transportation through

Figure 6. Exponential regression for the viscosity of liquid products at 40 °C with respect to the residue conversion for the ultradispersed catalytic hydroprocessing of Athabasca bitumen at moderate operating conditions.

Figure 7. Viscosity at 40 °C of the ultradispersed catalytic upgraded liquid products from Athabasca bitumen at moderate reaction temperatures and residence times; values calculated from the bitumen upgrading prediction computational program.

pipelines avoiding solvent addition. The viscosity of the different liquid products measured at 40 °C from the experiments carried out in this study was plotted versus the residue conversion in Figure 6. The residue conversion percentage was defined as ConvR = (1
yR/yR0)  100. Figure 6 depicts, as expected, an exponential decrement on the viscosity with respect to residue conversion. It is observed that for conversions higher than 9%, the experimental data fit with an exponential trend with a regression index (rexp2) of 0.9654; therefore, viscosity measurements can be employed to determine the extent of bitumen upgrading. 4.4. Bitumen Upgrading Prediction Computational Program. The MATLAB program developed in this work was improved by the addition of the viscosity-residue conversion correlation shown in Figure 6. With this addition, the MATLAB program is now a tool that not only allows the estimation of kinetic rate constants and computation of products composition for bitumen ultradispersed catalytic hydroprocessing but can also provide the liquid products viscosity at 40 °C for given reaction temperatures and residence times. 1370

dx.doi.org/10.1021/ef200094h |Energy Fuels 2011, 25, 1364–1372

Energy & Fuels The MATLAB program can be employed to build a graph that can provide the reaction conditions that are necessary to achieve a specific liquid product viscosity. Figure 7 shows an example of this kind of graph that can be helpful for planning future experiments or new upgrading schemes. Figure 7 demonstrates that liquid products from ultradispersed catalytic hydroprocessing of Athabasca bitumen with low values of viscosity at 40 °C can be obtained. The viscosity at 40 °C in the whole process is reduced from the bitumen’s original value of 7743 cP to viscosities in the range of 100
500 cP when moderate reaction temperatures and residence times are employed.

5. CONCLUSIONS The kinetic modeling of Athabasca bitumen hydroprocessing experiments, which were carried out in a pilot plant plug-flow reactor, was proposed in this work in order to predict the different product compositions and viscosity improvement. The five-lump kinetic model employed in this work, based on the work of Sanchez et al.,19 was capable of predicting the production of unconverted residue, VGO, distillates, naphta, and gases with AAE% < 7% with respect to the experimental data. Ultradispersed catalyst performances were similar to the one found in previous experiments carried out in laboratory batch reactors. The kinetic model in conjunction with a viscosityresidue conversion correlation permitted one to investigate, via numerical simulation, the bitumen ultradispersed catalytic upgrading extension. The modeling predictions from this work reveal that, depending on the reaction temperature and residence time, there could be a great reduction on the liquid products viscosity. These results are very promising, showing that catalytic upgrading employing ultradispersed catalysts may play a key role in the exploitation of the huge resources of bitumen and heavy oil in North America and elsewhere. Future developments in this research will include the kinetic modeling of ultradispersed catalytic hydroprocessing experiments employing different fractions of bitumen, catalyst concentrations, and formulations. ’ AUTHOR INFORMATION Corresponding Author

*Address: Department of Chemical and Petroleum Engineering, Schulich School of Engineering, University of Calgary, 2500 University Drive NW, Calgary, AB T2N 1N4. Phone: þ1 403 210 95 90. Fax: þ1 403 210 39 73. E-mail: [email protected].

’ ACKNOWLEDGMENT This work was supported in part by the National Council for Science and Technology of Mexico, The Alberta Ingenuity Centre for In Situ Energy funded by the Alberta Ingenuity Fund, the industrial sponsors Shell International, ConocoPhillips, Nexen Inc, Total Canada, and Repsol-YPF, and The Schulich School of Engineering at the University of Calgary, Canada. The authors thank C. E. Galarraga for providing the batch reactor experiments data, C. E. Scott for the fruitful comments that improved the original manuscript, and R. Gomez for his involvement in the pilot plant experiments. ’ REFERENCES (1) Hein, F. J. Heavy Oil and Oil (Tar) Sands in North America: An Overview & Summary of Contributions. Nat. Resour. Res. 2006, 15 (2), 67–84.

ARTICLE

(2) Herron, E. H.; King, S. D. Heavy Oil as the Key to U.S. Energy Security, http://www.petroleumequities.com/cgi-bin/site.cgi?p=energysecurity.html&t=5 (accessed on January 2009). (3) Gray, M. R. Tutorial on Upgrading of Oildsands Bitumen, http:// www.ualberta.ca/∼gray/Library/Tutorials/Upgrading/sld001.htm (accessed on March 2009). (4) Speight, J. G. New Approaches to Hydroprocessing. Catal. Today 2004, 98 (1
2), 55–60. (5) Speight, J. G. The Chemistry and Technology of Petroleum; CRC Press/Taylor & Francis: Boca Raton, FL, 2007. (6) Scherzer, J.; Gruia, A. Hydrocracking Science and Technology; Macel Dekker: New York, 1996. (7) Mohanty, S.; Kunzru, D.; Saraf, D. N. Hydrocracking: A Review. Fuel 1990, 69 (12), 1467–1473. (8) Fixari, B.; Peureux, S.; Elmouchnino, J.; Le Perchec, P.; Vrinat, M.; Morel, F. New Developments in Deep Hydroconversion of Heavy Oil Residues with Dispersed Catalysts. 1. Effect of Metals and Experimental Conditions. Energy Fuels 1994, 8 (3), 588–592. (9) Pereira-Almao, P.; Hill, J.; Wang, J.; Vasquez, A. In Ultra Dispersed Catalyst for Processing Heavy Hydrocarbon Fractions, AICHE, 2005 Spring National Meeting, Atlanta, GA, AIChE: Atlanta, GA, 2005. (10) Vasquez, A. Synthesis, Characterization and Model Reactivity of Ultra Dispersed Catalysts for Hydroprocessing. MSc Thesis, University of Calgary, Calgary, AB, Canada, 2007. (11) Pereira-Almao, P. In Fine Tuning Conventional Hydrocarbon Characterization to Highlight Catalytic Upgrading Pathways, Variability of the Oil Sands Resource Workshop, Lake Louise, AB, May 1
4, 2007, Alberta Ingenuity Centre for In Situ Energy: Lake Louise, AB, 2007. (12) Loría, H.; Pereira-Almao, P.; Scott, C. E. A Model To Predict the Concentration of Dispersed Solid Particles in an Aqueous Medium Confined inside Horizontal Cylindrical Channels. Ind. Eng. Chem. Res. 2009, 48 (8), 4088–4093. (13) Loría, H.; Pereira-Almao, P.; Scott, C. E. A Model To Predict the Concentration of Submicrometer Solid Particles in Viscous Media Confined inside Horizontal Cylindrical Channels. Ind. Eng. Chem. Res. 2009, 48 (8), 4094–4100. (14) Loria, H.; Pereira-Almao, P.; Scott, C. E. Model To Predict the Concentration of Ultradispersed Particles Immersed in Viscous Media Flowing through Horizontal Cylindrical Channels. Ind. Eng. Chem. Res. 2010, 49 (4), 1920–1930. (15) Pereira-Almao, P. R.; Ali-Marcano, V.; Lopez-Linares, F.; Vasquez, A. Ultradispersed Catalysts Compositions and Methods of Preparation. International Patent WO 2007/059621 A1, May 31, 2007. (16) Trujillo-Ferrer, G. L. Thermal and Catalytic Steam Reactivity Evaluation of Athabasca Vacuum Gasoil. MSc Thesis, University of Calgary, Calgary, AB, Canada, 2008. (17) Galarraga, C. E.; Pereira-Almao, P. Hydrocracking of Athabasca Bitumen Using Submicronic Multimetallic Catalysts at Near In-Reservoir Conditions. Energy Fuels 2010, 24 (4), 2383–2389. (18) Ancheyta, J.; Sanchez, S.; Rodríguez, M. A. Kinetic Modeling of Hydrocracking of Heavy Oil Fractions: A Review. Catal. Today 2005, 109 (1
4), 76–92. (19) Sanchez, S.; Rodriguez, M. A.; Ancheyta, J. Kinetic Model for Moderate Hydrocracking of Heavy Oils. Ind. Eng. Chem. Res. 2005, 44 (25), 9409–9413. (20) Galarraga, C. E.; Scott, C. E.; Loria, H.; Pereira-Almao, P. Kinetic Models for Hydrocracking of Athabasca Bitumen using Submicronic NiWMo Catalysts at Low Severity Conditions. Manuscript under preparation. (21) Carbognani, L.; Lubkowitz, J.; Gonzalez, M. F.; Pereira-Almao, P. High Temperature Simulated Distillation of Athabasca Vacuum Residue Fractions. Bimodal Distributions and Evidence for Secondary “On-Column” Cracking of Heavy Hydrocarbons. Energy Fuels 2007, 21 (5), 2831–2839. (22) Guin, J. A.; Tarrer, A. R.; Lee, J. M.; Lo, L.; Curtis, C. W. Further Studies of the Catalytic Activity of Coal Minerals in Coal Liquefaction. 1. Verification of Catalytic Activity of Mineral Matter by Model Compound Studies. Ind. Eng. Chem. Process Des. Dev. 1979, 18 (3), 371–376. 1371

dx.doi.org/10.1021/ef200094h |Energy Fuels 2011, 25, 1364–1372

Energy & Fuels

ARTICLE

(23) Qader, S. A.; Hill, G. R. Catalytic Hydrocracking. Mechanism of Hydrocracking of Low Temperature Coal Tar. Ind. Eng. Chem. Process Des. Dev. 1969, 8 (4), 456–461. (24) Qader, S. A.; Hill, G. R. Hydrocracking of Gas Oil. Ind. Eng. Chem. Process Des. Dev. 1969, 8 (1), 98–105. (25) Qader, S. A.; Hill, G. R. Hydrocracking of Petroleum and Coal Oils. Ind. Eng. Chem. Process Des. Dev. 1969, 8 (4), 462–469. (26) Qader, S. A.; Wiser, W. H.; Hill, G. R. Rate Data of Tar Hydrocracking. Fuel 1972, 51 (1), 54–56. (27) Ancheyta, J.; Marroquin, G.; Angeles, M. J.; Macias, M. J.; Pitault, I.; Forissier, M.; Morales, R. D. Some Experimental Observations of Mass Transfer Limitations in a Trickle-Bed Hydrotreating Pilot Reactor. Energy Fuels 2002, 16 (5), 1059–1067. (28) Anton, H. Elementary Linear Algebra; 9th ed.; John Wiley & Sons, Inc.: Hoboken, NJ, 2005. € Phillips, C. R. Kinetic Models for the Non(29) K€oseoglu, R. O.; catalytic Hydrocracking of Athabasca Bitumen. Fuel 1988, 67 (7), 906–915. (30) Smith, J. M. Chemical Engineering Kinetics; McGraw-Hill: Boston, MA, 1981.

1372

dx.doi.org/10.1021/ef200094h |Energy Fuels 2011, 25, 1364–1372