Kinetic Studies of Upgrading Pine Pyrolytic Oil by Hydrotreatment

Fuel Processing Technology, 19 (1988) 31-50 31 Elsevier Science Publishers B.V., Amsterdam - - Printed in The Netherlands

Kinetic Studies of Upgrading Pine Pyrolytic Oil by Hydrotreatment

YU-HWA E. SHEU, RAYFORD G. ANTHONY

Kinetic, Catalysis and Reaction Engineering Laboratory, Department of Chemical Engineering,

and

ED J. SOLTES

Wood Chemistry Laboratory, Department of Forest Science, Texas A&M University, College Station, Texas 77843 (U.S.A.)

ABSTRACT

Pyrolytic oil, produced by the Tech-Air Corporation from Southern Pine sawdust and bark, was hydrotreated in a trickle bed reactor system. Changes in product composition as a function of reaction temperature, hydrogen pressure, space velocity and catalyst type were determined quantitatively by size exclusion chromatography - gas chromatography (SEC-GC) analyses and elemental analyses. The catalysts used in the reactions were Pt/Al203/Si02, and sulfided CoMo/;,- A1203, Ni-W/7-A1203 and Ni-Mo/7-A1203. The reaction temperatures ranged from 623 K to 673 K, and the reaction pressures varied from 5,272 kPa (750 psig ) to 10,433 kPa ( 1,500 psig ). Weight- hourly space velocity was changed from 0.5 to 3.0 h - ~.

Two models, one for overall oxygen removal and the other for the compositional changes in hydrotreated oil, were developed. Oxygen removal was not a function of space velocity and was modeled by an empirical function of temperature and pressure. A pseudo first order reaction network was used to relate the kinetics of a five "lump" model.

INTRODUCTION

Biomass, which includes forest and agricultural products and residues, ani- mal wastes, and municipal solid waste, is a renewable resource that could play a significant role in providing for future energy needs. A number of biomass thermal conversion processes are currently being developed, including combustion, gasification, and pyrolysis. Pyrolysis refers to incomplete thermal degradation which yields three major products: char (mostly carbon), oil (usable as a liquid fuel), and fuel gas (usable in any combustor). Conditions for biomass pyrolysis and products are somewhat dictated by reactor type and design. An excellent and detailed summary of biomass pyrolysis is given by Chatterjee [ 1 ].

Although reported attempts at pyrolytic oil utilization are usually related to

0378-3820/88/$03.50 © 1988 Elsevier Science Publishers B.V.

32

direct combustion, pyrolytic oil is a poor boiler fuel. It is viscous, not com- pletely volatile, corrosive, exhibits high oxygen content, and does not mix with conventional fuels.

In order to improve pyrolytic oil as a liquid fuel, the oil must be processed to reduce viscosity, improve volatility, remove acidity, and lower oxygen content. A catalytic hydrotreating reaction was proposed to improve the quality of the oil. Reduction of oxygen content results in more hydrocarbon-like molecules with the higher energy content preferred in fuels. Hydrotreating can effectively lower the oxygen content, resulting in better fuels while usually promoting some cracking to enhance volatility.

Only limited information is available on hydrotreatment of biomass-derived oils. All hydrotreating work on biomass pyrolytic oil to date has been done in batch reactors. Lin [2] reported that use of plat inum and other noble metal catalysts for hydrotreating pyrolysis oils yields a gasoline-diesel mixture. Soltes and L in [3, 4] reported that pyrolytic oils (tars) from a variety of agricultural residues and their hydroprocessed products were similar in composition re- gardless of biomass feedstock used.

Pyrolytic oil obtained from pine bark and wastes by the Tech-Air pyrolysis process was used in this research. Analyses of the Tech-Air pyrolytic oil, reported by Elder [5 ] and Elder and Soltes, [6-9 ] indicate that solvent extract- able phenolics comprise approximately 13 wt.% of the oil. Pyrolytic oil also contains volatile organic acids, as well as a variety of neutral components [9].

The major voltatile components (detectable by the gas chromatographic analysis) in the oil are the phenolics, e.g. phenol, cresol, some alkyl phenols, guaiacol, and eugenol [ 10 ]. A high percentage of the components of pine pyrolytic oil have high molecular weights. Less than 15 wt.% of the pine pyrolytic oil can be detected by gas chromatography. The oil is very hygroscopic and has about 10 wt.% moisture. Furthermore, the oil is heat-sensitive and subject to re-polymerization, so it cannot be fractionated by distillation.

An analytical method combining size exclusion chromatography (gel permeation chromatography) (SEC) and high resolution gas chromatography (GC) was developed by Sheu et al. [10] to analyze pyrolytic oil and its hydrotreated product. The separation method utilizes the concept that SEC separates pyrolytic oil into fractions which are composed primarily of the following types of chemical compounds; nonvolatiles, alkanes, phenols, and aromatics. Gas chromatography (GC) with the use of an internal standard (1-decene) gives percentages of volatiles. Changes in composition during the hydrotreating process were determined by comparing the SEC-GC analyses and the elemental analyses of the charge stocks and products.

The objectives of this research were to convert this heavy, high-oxygen-content oil into a lighter hydrocarbon mixture to obtain a better liquid fuel; dem- onstrate performance and operability of a trickle bed reactor; and obtain a

33

kinetic and reactor model. The P t catalysts used by Lin [ 2 ], as well as, sulfided Co-Mo, N i -Mo and NiW were to be used.

EXPERIMENTAL

Feedstock

Pyrolytic oil produced by the Tech-Air Corporation from pine sawdust and bark was used in this research. The physical properties of Tech-Air pyrolytic oil are listed in Table 1. The purified grade decahydronaphthalene (decalin) from Fisher Scientific was used in this study as a solvent.

Catalyst

Details regarding the characteristics of the catalysts used in this research are given in Table 2. The 5% Pt /Ale03 powder catalyst was mixed with Ludox AS-40 binder (70% SiOe, 30% H20 mixture sold by du Pon t ) in such a pro- portion that the final catalyst contained 30% SiO2. A paste was made when the binder was added. It was then taken into a syringe with a 1/16 inch (0.1588 cm) plunger. The catalyst was extruded, dried, and calcined in air at 756 K

TABLE1

Properties of pine pyrolytic oil

Property Result Method

Density 1.1415 g/cm 3 Water content 9.7% Heating value (wet basis) 27,800 kJ/kg Solubility (wt.%)

Water 39.7% Acetone 99.6% Methylene Chloride 93.5 % Toluene slight Hexane slight

pH 2.9

Filterable Solids (wt.%) Acidity (milli- equivalents acid per gram ofoil)

0.3%

0.64

ASTM D-85-70

ASTM D-240-64

5% oil dispersed in water Acetone insoluble Titration by NaOH

34

TABLE2

Specifications of the hydrotreating catalysts

Catalyst Manufacturer Size Composition Surface Pore Pore size area volume diameter (m2/g) (cm3/g) {.~)

Pt/A1203 Strem{78-166) powder Co-Mo Harshaw (HT-400) 1/16"E. b Ni-Mo Harshaw (HT-500) 1/12"E. Ni-W Harshaw (Ni-4301) 1/12"E.

5% Pt 100" 0.52" 100" 3% CoO, 15% MoO3 200 0.45 94 3.5% NiO, 15.5% Mo03 200 0.46 88 6% Ni, 19% W 230 0.37 104

"For the pellet catalyst. bE: Extrudates.

(483 °C) for four hours. In order to obtain an active catalyst,the P t /Al2Off Si02 pellet was reduced in situ prior to the experiments. The reduction was done by passing the H2 through the catalytic bed at 673 K, 8,720 kPa at a flow rate of 200 cm3/min (21.2°C, 1 bar) for one hour.

For Harshaw's catalysts, the sulfided form was tested. Presulfiding of the catalysts was done in situ prior to the experiments. A mixture of 90% H2 and 10% H2S by volume was passed through the catalyst bed at a flow rate of 40 cm3/min (measured at 21.2 ° K, 1 bar) at 673 K and atmospheric pressure until the outlet gas showed no further sign of H2S consumption.

Apparatus

A schematic diagram of the trickle-bed reactor system is shown in Fig. 1. The reactor consists of a 316 stainless steel tube, 3/4 inch (1.905 cm) O.D., 0.065 inch (0.1651 cm) wall thickness, and 32 inches (81.3 cm) long. The top 20 inches (50.8 cm) was packed with catalyst and the bottom 12 inches {30.48 cm) was packed with Pyrex ® 3 mm diameter glass beads (Cat. No. 7268-3 ).

An automatic temperature controller (Techne AC-40) was used to control the temperature of the salt bath. The heating medium was a Hitec ® heat trans- fer salt which is a mixture of 53% potassium nitrate, 40% sodium nitrite and 7% sodium nitrate {American Hydrotherm Co. ). Air was bubbled through the salt bath to ensure a more uniform temperature profile.

A thermowell [ 1/8 in. i.d. (0.3175 cm) ] was extended from the bottom to the top of the reactor. A 0.04 inch (0.1015 cm) OD chromel-alumel thermocouple (Omega Model 4001KC1 ) was inserted into the thermowall and moved up and down to measure the axial temperature profile of the reactor. The reactor temperature was nonisothermal, because only the bottom 22 inches (55.88 cm) of the reactor was in contact with the salt bath. Figure 2 shows two typical temperature profiles inside the reactor at two different salt-bath temperature settings.

By immersing 22 inches of the reactor in the salt-bath, the temperature inside the reactor increased gradually along the reactor length. This prevented

~R.=¢=~_ -C

S- I I

--c><3

I

Metering Pump

Metering Pump

Q

35

S-I

Fig. 1. Schematic diagram of the trickle-bed reactor system. BPR: back pressure regulator, C: water cooled exchanger, F: filter, G: gas sample collector, L: liquid sample collector, MF: mass flow meter, P: pressure gauge, R: regulator, TC: temperature controller, TI: temperature indicator, W: wet test meter, S-I and S-II: separators.

the volatiles in the pine pyrolytic oil from flashing suddenly into the gas phase. With the catalyst packed at the top section of the reactor, the pyrolytic oil was hydrotreated before the oxygen-containing compounds polymerized at high temperature. Table 3 shows typical operating conditions.

After the reaction, the catalyst was regenerated with air to remove the coke deposited on the catalyst during the reaction. Then, the catalyst bed was ac- t ivated again for the next run.

Analysis

Gas sample

The gas sample was analyzed using a Carle Series S, 111 H analytical gas chromatograph, with columns and programs as reported by Philip et al. [ 11 ].

36

o v

W n-

I - ~C n" W O.

W I -

400

380

340

300

260

220

180

Glass Catalyst Bed = 1 _ Bead ~1

0 4 8 12 16 20 24 28 32

REACTOR LENGTH (inch)

Fig. 2. Temperature profile inside the reactor. Reaction conditions: Pt/A12OJSiO2 catalyst, 8,720 kPa, WHSV=2 h -1, salt bath temperatures: [] 623 K (350°C) and o 673 K (400°C).

TABLE 3

Typical reactor operating conditions

Gas feed (H2) Liquid feed Weight hourly space velocity (WHSV)" Salt bath temperature H2 pressure Catalyst bed

: 100 cc/min (at 60 °F, i bar) per gram of pine pyrolytic oil input. : at a ratio of 2 g decalin/gram of pine pyrolytic oil

: 0.5 to 3.0 h -1

: 623 to 673K : 5,272 to 10,443 kPa (750 to 1500 psig) : 60 g for each load

"WHSV = Gram of pine pyrolytic oil input per hour Gram of catalyst in the reactor

A Hewlett-Packard 3385A Automation System integrator and printer was used in conjunction with this GC.

Liquid sample

Size exclusion chromatography (gel permeation chromatography) (SEC) combined with GC was used to analyze pine pyrolytic oil and its hydrotreated product. The method separates the oil into fractions composed principally of nonvolatiles, alkanes, phenols, and aromatics. Changes in composition during the hydrotreating process were determined quantitatively by comparing the SEC-GC analyses and the elemental analyses of the charge stocks and prod-

37

ucts. The details of the SEC-GC analytical method have been described by Sheu et al. [ 10 ].

An elemental analysis was obtained for the total feed and products. Carbon and hydrogen balances were closed within 2%, whereas, the oxygen balances failed to close by 10 to 20%. This deviation was attributed to the inability to measure the oxygen content of the coke. The total material balances on a mass basis were closed within 2%. For additional information on experimental procedures, the reader is referred to Sheu [12].

RESULTS AND DISCUSSION

Oxygen removal

The amounts of oxygen in pine pyrolytic oil and the hydrotreated products were determined by elemental analysis. The effect of reaction temperature, pressure, and space velocity on oxygen removal are shown in Figs. 3 (d), 4 (d) and 5(d). The points in the figures are the experimental data and the solid lines were generated by using an oxygen removal model. A clear trend of the effect of reaction temperature and pressure on the oxygen removal in the hydrotreating process is shown in Figs. 3(d) and 4(d). Figure 5(d) shows that changes in space velocities did not affect oxygen removal. No data for the sul- tided NiW/AI203 catalyst are included. For all runs conducted with this catalyst significant coking occurred which rapidly plugged the reactor.

Even though the oxygen removal was not a function of space velocity, eqn. (1) is used because the experimental data were dependent on temperature and pressure. An Arrhenius type equation and an nth order pressure effect are assumed in eqn. (1).

dC°xy /~ Cm pn clZ ='" ~oxy (1)

The corresponding integrated equation is:

L

( R - ~ o ( 1 - T° ~ d z f o r m ¢ l (2) (C°~Ym- 1 ) - C ° m - 1 - vo~y -Pn koexp \ T ( z ) ] ] 0

and

- I n (1--Xoxy) =pnl for m=l .

Where,

38

4O

35

3O

25 I-- "I" ~ 2o

10

4 0

3 5

3O

25 I-. "I" 20

w

10

- (a) 40:

35

3O

~ 25

- - ~ 10

- - f i

I I 1 I I I 1 I

Z

0

I i I I

i 6O ~ 55 ~

45 ~

4 o

35

3o

2 5

2o

15

10 -

5 -

l i i 6 0 0 6 2 s 6 5 0 6 7 5

I 700

REACTION TEMPERATURE (°K)

Fig. 3. The effect of temperature on product distributions, • heavy nonvolatiles, [] light nonvolatiles, • phenolics,/x aromatics + alkanes, O coke + water + gases for (a) sulfided Ni-Mo/A1203, (b) sulfided Co-Mo/A1203, (c) Pt/Al2OJSiO2, and (d) on oxygen removal for A sulfided Ni- Mo/Al203, [] sulfided Co-Mo/A1203, and • Pt/A1203/Si02. Pressure=8,720 kPa, WHSV=2 h- 1. Solid lines are model predictions.

Coxy = concen t r a t ion of oxygen in pyroly t ic oil = gram of oxygen in hydro t r ea t ed oil

g ram of oxygen in raw pyrolyt ic oil C ° x y = initial oxygen concen t r a t i on (equal to one)

L = tota l catalyt ic bed length (cm) m = react ion order n = pressure effect order P = react ion pressure (kPa) To = reference t empera tu re (K) Xoxy = oxygen convers ion z = axial length Ea = ac t iva t ion energy

k = rate c o n s t a n t

39

40 40

30 ~ 30 o 20 20

10 - ~ 10

Q

30 " ~ ~ , - ~ (c)

~ 5o

1" 2O ~ ~- ~: ,o i l J

z

i ~ 20

4 . 5

PRESSURE, MPa

_•• (b)

•

I I I

[ J I 6.0 7,5 9.0 10.5

PRESSURE, MPa

Fig. 4. The effect of pressure on product distributions, • heavy nonvolatiles, [] light nonvolatiles, • phenolics, Zk aromatics+alkanes, O coke+water+gases for (a) sulfided Ni-Mo/Al203, (b) sulfided Co-Mo/A1203, (c) Pt/Al~0JSi02, and (d) on oxygen removal for /x sulfided Ni-Mo/ A1203, [] sulfided Co-Mo/Al20~, and • Pt/A12OJSi02. Temperature=673 K. WHSV=2 h -1. Solid lines are model predictions.

The integral I = fL ko exp (Ea/R To ( 1 - To~ T ( z ) ) ) dz was evaluated by using the six point Gauss quadrature numerical technique [ 13 ].

The nonlinear regression program from the "Statistical Analysis System" (SAS) package [14] was applied to determine the parameters, m, n, ko, and Ea which are listed in Table 4.

No data using Pt /AI2OJSiO2 catalyst for hydrodeoxygenation (HDO) has been reported in the literature. The P t catalyst exhibits the best activity for oxygen removal for pine pyrolytic oil among the four catalysts tested.

Although the experimental data shows that changes in space velocity did not affect oxygen removal, the "activation energies" developed from this oxygen removal model for sulfided Co-Mo and Ni-Mo catalysts agree with those reported by others [15, 16 ]. The "activation energy" for HDO using sulfided Co- Mo catalyst agrees with the activation energy (E ,=84 .4 kJ /mol ) reported by

40

~e

I-- "1- (.~ IJJ

40

35

30

25

20

15

10

5

o

_ ~ o o

I 1 I l

40

35

30

~e 25 I - - ~ 2o

10

5

I (b)

40

35

30

25

l.IJ .20

15

10

(c)

I I I 1

SO > 55 0 2~ 50 LD 4 5

40 Z LU 35 (9 >-

o ~ o o

30 25 20 I

0 1

SPACE VELOCITY (WHSV)

(d)

£t

L I l 2 3 4

Fig. 5. The effect of space velocity on product distributions, • heavy nonvolatiles, [] light nonvolatiles, • phenolics, A aromatics+alkanes, © coke+water+gases for (a) sulfided Ni-Mo/ A1203, (b) sulfided Co-Mo/A1203, (c) Pt/A12OJSi02, and (d) on oxygen removal for A sulfided Ni-Mo/A1203, [] sulfided Co-Mo/Al203, and • Pt/A12OJSi02. Pressure = 8,720 kPa. Tempera- ture = 673 K. Solid lines are model predictions.

T A B L E 4

Parameters for oxygen removal model (eqn. 1 )

Catalyst m n ko E~ (cm- 1.kPa-") a (kJ/mol)

Pt/A12OJSi02 1.0 1.0 1.50 _+ 0.02 × 10- 7 45.5 _+ 3.2 Sulfided

Co-Mo 1.0 0.3 1.07 _+ 0.07 X 10-3 71.4 _+ 14.6 Sulfided

Ni-W 1.0 0.5 7.00+0.22×10 -5 61.7+7.1

"ko: evaluated at 648 K ( 3 7 5 ° C ) .

41

Ternan and Brown [15]. No hydrogen pressure effect on HDO of biomass liquids using sulfided Co-Mo catalysts has been reported.

The "activation energy" for HDO using sulfided Ni-Mo catalyst is 61 + 7.1 kJ/mol. This value agrees with the activation energy for HDO of dibenzofuran using sulfided Ni-Mo catalyst [16].

Dibenzofuran , Biphenyl, E, = 68.4 kJ/mol;

Dibenzofuran , Cyclohexylbenzene, Ea = 76.2 kJ/mol.

Effect of temperature on reaction products

The effect of the reaction temperature on hydrotreated pyrolytic oil was evaluated at 623, 648, and 673 K. All reactions were run at 8,720 kPa and weight hourly space velocity equal to 2 h-1. One difference in the reaction products at different reaction temperatures was that the liquid products at 623 K did not have a water layer separated from the hydrotreated oil. Thus, the liquid product at 623 K contained the hydrotreated pyrolytic oil, solvents, and water.

Figure 6 (a) - (c ) shows the gas chromatogram for the SEC phenolic fraction of the pine pyrolytic oil and for two different reaction products using P t / A 1 2 0 J SiO2 at 623 and 673 K. Significant changes in the phenolic fraction is observed. The chromatogram of the phenolic fraction at 623 K (Fig. 6 (b)) has less volatile phenols than those in pine pyrolytic oil (Fig. 6 (a)), which indicates that the small molecular weight phenols, guaiacol, methyl guaiacol, etc., have been converted into aromatics during the hydrotreating reaction. The volatile phenols in pyrolytic oil decreased from 12.3% (before reaction) to 6.9%, and the aromatic and alkane fractions increased from 1.5 to 9.87 and 0 to 1.9%, re- spectively. The phenolic peaks identified in Fig. 6 (c) are different from those identified in Fig. 6 (a). This suggests that phenolics are intermediate products which are produced and consumed during the hydrotreating reaction. Com- paring the two reaction products, the volatile phenols increased from 6.9 @ 623 K to 10.6 wt.% @ 673 K. The lumped aromatics fraction also increased from 9.8 to 16.2 wt.%. The following reaction scheme is suggested:

k3 k4 Light nonvolatiles , Phenolics , Aromatics

k2

Part of the phenols in pine pyrolytic oil are hydrogenated to produce aromatics; phenols, are produced also from the nonvolatiles during the reaction; and some aromatics are produced from the light nonvolatile fraction. The alkanes are produced by hydrogenating the esters in the pine pyrolytic oil.

42 ¢

I (a) 5

I L i I ~ I

(b)

t I i r I

I 9 (c)

67 1 1

t I r i i i

15 25 35 45 55 65

RETENTION TIME (rain.)

Fig. 6. Typical gas chromatograms of the phenolic fractions (Size Exclusion Fraction # 3). GC conditions: SP-2100 column, carrier gas He: 40 ml/min, temperature program; 50 to 280°C/min, initial hold 3 min. (a) Pine pyrolysis oil-numbers correspond to [ 1 ] m,p-cresol, [2 ] guaiacol, [3 ] dimethyl phenol, [4] 4-methyl guaiacol, [5] 4-ethyl guaiacol, [6] eugenol, [7] isoeugenol. (b) Phenolic fraction from hydrotreated product using Pt/A12OJSi02 at 623 K, 8,720 kPa, and 2 h - ' . (c) Phenolic fraction from hydrotreated product using Pt/Al2OJSi02 at 673 K, 8,720 kPa, and 2 h - ' . Numbered peaks are: [1] 1-decene, [2] o-cresol, [3] m,p-cresol, [4] and [5] C2-phenol, [6] and [7] C3-phenol, [8] 2-phenylphenol, [9] C+-phenol, [10] naphthol, [11] Cs-phenol, [12] C6- phenol. 1-decene is the internal standard.

Kinetic and reactor model

Because many species are contained in pine pyrolytic oil and its hydrotreated products, the lumping together of constituents with similar functional groups is a useful approach. Lumped kinetic models provide a means of quan- tifying the effects of process variables on product yields and also of summariz- ing large amounts of experimental data.

The following major "lumps" are used in the kinetic model.

Heavy nonvolatiles = nonvolatiles in SEC fraction 1 + nonvolatiles in SEC fraction 2.

Light nonvolatiles

Phenols

Aromatics

Alkanes Coke + H20 ÷ Outlet Gases

The liquid yield, YL, is

= nonvolatiles in SEC fraction 3 + nonvolatiles in SEC fraction 4.

= volatiles in SEC fraction 3 (compounds detectable by GC).

= volatiles in SEC fraction 4 +volatiles in SEC fraction 5 (excluding solvents).

-- volatiles in SEC fraction 2

1 -- liquid yield.

defined as follows:

E (wt. of fractions by SEC) - (solvents in hydrotreated oil) YL-

43

pine pyrolytic oil input

The ~ (wt. of fractions by SEC ) represents the weight of liquid product which includes the hydrotreated oil and solvents.

Table 5 shows an example of the lumped fractions of the pine pyrolytic oil and a hydrotreated product. Because the alkane concentrations were small, alkanes and aromatics were combined for the kinetic model. A pseudo first order kinetic model is used to relate the reaction rates of lumped species. The proposed model consists of the following kinetic scheme

k l k3 k4 HNV ) LNV , Phenolics ~ [Aromatics+Alkanes]

TABLE 5

Typical lumped fractions for kinetic modeling ~ (Pt/A12OJSi02, 673 K, 8,720 kPa, and 2 hr -1 WHSV)

Lumped Pine Hydrotreated fraction pyrolytic oil pyrolytic oil

(wt.%) (Exp. No. 220) (wt.%)

1 Heavy Nonvolatiles: 0.4932 0.2457 2 Light Nonvolatiles: 0.3690 0.2941 3 Volatile Phenols: 0.1232 0.1063 4 Aromatics: 0.0146 0.1625 5 Alkanes: 0.0 0.0327 6 [Gases+H20+Coke]: 0.0 0.1587

1.0000 1.0000

aSheu [12].

44

k5 [Aromatics + alkanes] , [COKE+WATER+GASES]

where HNV and LNV refers to "heavy" and "light" nonvolatiles. The rate expressions for the appearance of the lumped species are:

rl = - k ip, (3)

r2 = klfll - k2p2 - k3p2 (4)

r3 = k3p2 - k4P3 (5)

r4 -~ k2p2 + k4P3 - k5P4 (6)

r~ = k5fl 4 (7)

and the subscripts "1" = heavy nonvolatiles, "2" =light nonvolatiles, "3" = volatile phenols, "4" = [aromatics + alkanes ], and "5" = [coke + H20 + outlet gases ]. Equations (8) through ( 12 ) are the material balances with axial dispersion for each lumped reaction species occuring in the process.

d2 Y1 - d Y 1 -

dZ- ~ - PeM ~ + PeM Da

d2Y2 - dY2 - --dZ 2 - PeM ~ + PeM Da '3

exp (01(1--~-)) Y I = O

exp (02(1 -~- ) ) Y2

+PeM Da'~ exp (y3(1 - -~-) ) Y2

=PeM Da~ exp (01 (1- -~- ) ) Y1

d2Y3 - dY3 - To dZ----~--PeM-d~- + PeM Da~ exp (04(1--~-)) ]13

=PeM Da~ exp (0a(1--~-)) ]12

d2Y4 dY4 (05(1--~-)) Y4 dZ----- ~ - PeM ~ + PeM Da ~ exp

= P e M D a ~ e x p (04(1----~)) ]I3

+PeM Da`3 exp (02(1--~-)) Y2

(8)

(9)

(10)

(11)

45

d 2 Y5 d Y5 _~_ dZ 2 P e M ~ = P e M D a ~ e x p (~5(1-- )) Y4 (12)

where

Damkohler number (Da °) = Lkio/Ua, Peclet number ( P e M ) .= ULL/D. (13)

~i = E J R T o (14)

kio =Ai exp ( -7 i )

Ai = frequency factor (min -1) D = dispersion of the reacting species in liquid phase (cm2/min) Eai = activation energy (kJ /mol) UL = superficial velocity of the pyrolytic oil (cm/min)

density of reacting fraction i Yi =

density of pine pyrolytic oil Z = dimensionless catalytic bed length.

The boundary conditions for eqns. (8) through (12) are:

-- PeM ( Yi - yo) at Z = 1. dZ

- 0 dZ

at Z = 0 . ( i=1 to 5)

A method [17-19] of combining quasilinearization with orthogonal collo- cation was applied to estimate the parameters PeM, Da o, and ?i in eqns. (8) to (12) simultaneously. The method was originally developed for modeling the liquefaction of coal in a bubble column reactor.

The parameter estimation was performed on a single run basis. Thus, for each reaction run, a set of lumped fraction data from the SEC-GC analysis plus the temperature profile inside the reactor were given. Equations (8) to ( 12 ) were solved simultaneously and the parameters PeM, Da °, and ~i for each run were obtained. The activation energy (Ei) and frequency factor (Ai) of the lumped kinetic model were derived from the estimated parameters Da ° and Yi by eqns. (13) and (14). The average activation energies for different catalyst are given in Table 6 along with the average frequency factors for the various steps in the reaction network. Also listed in Table 6 is the Peclet number (PeM). The average frequency factors listed in Table 6 were evaluated at 8,720 kPa. The pressure effect on reaction rates will be discussed later.

46

TABLE 6

Kinetic and reactor parameters (Ai are at 8,720 kPa)

Pt /A12OJSi02 Sulfided Sulfided Ni-Mo/A1203

Co-Mo/A1203

E, (kJ /mol ) 74.0 _+ 2.9 a 74.5 __ 3.2 82.2 _+ 2.9 E2 91.8 _+ 4.3 96.4 _ 3.4 105.8 _+ 3.0 E3 80.6 _+ 2.6 81.8 _+ 3.6 90.4 _+ 3.0 E4 62.3 _+ 2.6 69.0 _+ 2.9 68.4 _+ 3.4 E5 69.6 __ 2.8 55.8_+ 3.6 74.9 __ 2.7

A1 (min - ' ) 3,860 _ 500 3,500 + 280 8,800 _+ 1000 A2 75,400 _+ 12,000 218,000 _+ 8,400 654,000 __ 59,000 A:~ 8,300 + 1,100 7,700_+ 460 30,600 _+ 3,600 A4 950 _+ 160 3,100 + 560 1,920 __ 50 A5 4,000 __ 650 450 _+ 47 16,400 _+ 160

PeM 10.2 _ 1.1 9.86 _+ 0.05 9.63 __ 0.38

aMean _-2- standard deviation.

The experimental and predicted values of the lumped species in the kinetic model at various reaction conditions are discussed in the following subsections.

Effect of temperature

Figures 3 ( a )- ( c ) shows the comparison of experimental and predicted concentration profiles for the various species in the lumped kinetic model. The points in the figures are the experimental data. The solid lines were calculated from the lumped kinetic model using the parameters listed in Table 6. Figures 3(b) and 3(c) for sulfided Co-Mo and Pt/A12OffSi02 show an 8% to 10% decrease in the "heavy nonvolatiles" and "light nonvolatiles" by an increase of reaction temperature from 623 to 673 K. The lumped fractions, "Phenols", "Aromatics + alkanes", and "coke + water + outlet gases", increase with higher reaction temperature.

Even though the reaction rates increased with the reaction temperature, the temperature effect on the performance of sulfided Ni-Mo catalyst was not significant. Figure 3 (a) shows that the lumped fractions only change about 1 to 4% bY increasing the reaction temperature from 623 to 673 K.

Effect of reaction pressure

The effect of reaction pressure was evaluated at 5,272; 6,996; 8,720 and 10,443 kPa at 673 K and weight hourly space velocity equal to 2 h-1. A relationship

47

between the frequency factor in the lumped kinetic model and the reaction pressure of the following form was assumed

At (:~ phi

The calculated values of n~ and are listed in Table 7. All of the frequency factors increased with reaction pressure except for the Pt/A1203/SiO2 catalyst. The effect of pressure reflected by the parameters, n~, for the different catalysts is different for each catalyst, because of the difference in catalytic activity. Since this is a two phase system increasing pressure will increase the concentration of hydrogen in the liquid, which apparently increases the rates of the reactions. Hence, the rates in the lumped kinetic model increase with higher reaction pressure except for the following reaction when using Pt/A12OffSiO2:

k5 "Aromatics + alkanes" , "Coke + H20 + outlet gases"

which decreases with higher reaction pressure. The experimental and predicted profiles for the various reacting species at

673 K are shown in Figs. 4 (a) - (c ). The agreement between the experimental results and model predictions appears to be reasonably good. Overall, the increase in reaction pressure produced fewer "Heavy nonvolatiles" and "Light nonvolatiles", and more "Aromatics + alkanes" and "phenolics". Higher "Coke + H20 + outlet gases" are produced with sulfided Harshaw Co-Mo and Ni-Mo catalysts, but less "Coke + H~O + outlet gases" are produced with P t / A12OffSi02.

Effect of space velocity

The effect of weight hourly space velocity on the reaction products was evaluated at 0.5, 1.0, 2.0, and 3.0 h - 1. The experimental and predicted concentration profiles for the various reacting species are shown in Figs. 5 (a)- (c) . By comparing Figs. 5 (a ) - (c), one observes:

T A B L E 7

Effect of reaction pressure on the frequency factor (Ai = A ° P n~)

Pressure Pt /A12OJSiO2 Sulfided Sulfided effect Co-Mo/Al203 Ni-Mo/AI203

nl 1.35 0.83 0.16 n2 0.40 0.71 0.17 n3 1.58 0.93 0.31 n4 1.08 0.73 0.25 n~ - 0.64 1.34 0.30

48

( 1 ) The weight percentage of the phenolic fraction does not vary with different space velocities. Hence, the reaction rate of "Light nonvalatiles Phenols" is equal to the reaction rate of "Phenols Aromatics ÷ alkanes".

(2) The production of aromatics and alkanes favors low space velocity for the Pt/A1203/Si02 catalyst.

(3) The sulfided Co-Mo catalyst forms more "Coke ÷ H20 ÷ outlet gases" during the hydrotreating reaction than Pt/A120ffSi02 or sulfided Ni-Mo catalyst.

(4) For sulfided Ni-Mo catalyst, the most significant effect of space velocity is the change in the concentrations of "Light nonvolatiles" and "Coke ÷ H20 ÷ outlet gases" fractions.

CONCLUSIONS

Although the experimental data for oxygen removal shows that changes in space velocity do not have a significant effect on oxygen removal, the "activation energies" developed from the oxygen removal model for sulfided Co-Mo and sulfided Ni-Mo catalysts agree with those reported in the literature. No kinetic data using Pt/A12OJSi02 as catalyst for oxygen removal have been reported in the literature.

Data obtained from the SEC-GC analysis were modeled with pseudo-first order kintetics describing the reaction rates of lumped species. The model was composed of five fractions: (1) heavy nonvolatiles, (2) light nonvolatiles, ( 3 ) phenols, ( 4 ) [ aromatics + alkanes ], and (5) [ coke ÷ H2 ÷ outlet gases ]. An Ar- rhenius model was used to describe the temperature dependence of the rate constants. A clear trend of the effect of reaction temperature, weight hourly space velocity, and reaction pressure on hydrotreating pyrolytic oil using P t / Al2OffSi02, sulfided Co-Mo, and sulfided Ni-Mo catalyst were observed. The agreement between the experimental data and model predictions are good.

Small trickle -bed reactors such as those used in laboratories and pilot plants have some axial dispersion. The Peclet number estimated from the lumped kinetic model is about 10.

The SEC-GC analysis of hydrotreated pyrolytic oil shows that part of the phenols in pine pyrolytic oil have been hydrogenated and converted into aromatics, and phenols have been produced also from the nonvolatiles during the reaction. Some aromatics were converted from the light nonvolatile fraction directly. The alkanes were produced by removing the oxygen atoms from the esters in the pine pyrolytic oil during the hydrotreating reaction.

Pt/A12OJSiO2 catalyst has the best hydrotreating ability among the four catalysts tested. During the upgrading process, some oxygen functional groups in the pyrolytic oil have to be removed (hydrotreated) first, otherwise the oxygen-containing compounds in pyrolytic oil polymerize at high temperature.

49

The hydrotreated product at the reaction conditions applied in this research still contain a certain amount of oxygen and nonvolatiles. In order to remove more oxygen and convert the nonvolatiles into volatiles, a higher reaction temperature and pressure or a new more efficient catalyst should be applied.

ACKNOWLEDGMENT

The financial support from the U.S. Department of Agriculture under the Special Research Grants Program, Agreement No. 59-2481-0-2-089-0 is greatly appreciated. Special acknowledgement is also made to the Center for Energy and Mineral Resources, the Texas Agriculture Experimental Station, and the Texas Engineering Experimental Station, all part of Texas A&M University System.

REFERENCES

1 Chatterjee, A.K., 1981. State-of-the-Art Review of Pyrolysis of Wood and Agriculture Bio- mass. Final Report, Contract No. 53-319-R-0-206, AC Project P0380, USDA Forest Service, Washington, D.C.

2 Lin, S.C., 1981. Hydrocarbons via Catalytic Hydrogen Treatment of a Wood Pyrolytic Oil. Ph.D. Dissertation, Texas A&M University.

3 Soltes, E.J. and Lin, S.C., 1983. Vehicular fuels and oxychemicals from biomass thermochem- ical tars. Biotechnol. Bioeng. Syrup., 5: 53.

4 Soltes, E.J. and Lin, S.C., 1984. Hydroprocessing of biomass tars for liquid engine fuels. Progr. Biomass Conver., 5: 1.

5 Elder, T.J., 1979. The Characterization and Potential Utilization of the Phenolic Compounds Found in a Pyrolytic Oil, Ph.D. Dissertation, Texas A&M University.

6 Elder, T.J. and Soltes, E.J., 1979. Further Investigations into the Composition and Utility of a Commercial Wood Pyrolysis Oil. Paper presented at the American Chemical Society National Meeting, Honolulu, HI.

7 Elder, T.J. and Soltes, E.J., 1979. Adhesive Potentials of Some Phenolic Constituents of Pyrolytic Oil, Paper presented at the American Chemical Society National Meeting, Wash- ington, D.C.

8 Elder, T.J. and Soltes, E.J., 1980. Pyrolysis of lignocellulosic material: Characterization of the phenolic constituents of a pine pyrolytic oil. Wood & Fiber, 12: 217.

9 Soltes, E.J. and Elder, T.J., 1981. Pyrolysis. In: I.S. Goldstein (Ed.), Organic Chemicals from Biomass. CRC Press Inc., Boca Raton, FL.

10 Sheu, Y.E., Philip, C.V., Anthony, R.G. and Soltes, E.J., 1984. Separation of functionalities in pyrolytic tar by gel permeation chromatography - Gas chromatography. J. Chrom. Sci., 22: 497.

11 Philip, C.V., Bullin, J.A. and Anthony, R.G., 1979. Analysis of lignite-derived gases by au- tomated gas chromatography. J. Chrom. Sci., 17: 523.

12 Sheu, Y-H.E., August 1985. Kinetic Studies of Upgrading Pine Pyrolytic Oil by Hydrotreat- ment. Ph.D. Dissertation, Texas A&M University.

13 Stroud, A.H., 1974. Numerical Qaudrature and Solution of Ordinary Differential Equations. Springer-Verlag, New York.

14 SAS Institute, Inc., SAS User's Guide - 1986, Cary, North Carolina.

50

15 Ternan, M. and Brown, J.R., 1982. Hydrotreating a distillate liquid derived from subbitu- minous coal using a sulphided CoO-MoO3-A1203 catalyst. Fuel, 61:110.

16 Krishnamurthy, S., Panvelker, S. and Shah, T.Y., 1981. Hydrodeoxygenation of dibenzofuran and related compounds. AIChE J. 27 (6): 994.

17 Tarng, Y.-J. and Anthony, R.G., March 31, 1984. Estimation of Parameters in Differential Equations, quarterly report. U.S. Department of Energy, DOE/FC/10601-4.

18 Tarng, Y.-J, August, 1985. A Reactor and Kinetic Model for Liquefaction of Lignite in an Upflow-Tubular Reactor. Ph.D. Dissertation, Texas A&M University.

19 Tarng, Y.-J. and Anthony, R.G., 1988. A reactor and kinetic model for liquefaction of lignite in an upflow tubular reactor, Part I; Model development and Part II. Applications. Fuel Pro- cessing Technol., 18 (3): 237-271.

Documents

Kinetic Studies of Upgrading Pine Pyrolytic Oil by Hydrotreatment