Fischer-Tropsch synthesis using iron based catalyst in

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 2 ( 2 0 1 7 ) 2 9 2 2 2 e2 9 2 3 5

Available online at www.sciencedirect.com

ScienceDirect journal homepage: www.elsevier.com/locate/he

Fischer-Tropsch synthesis using iron based catalyst in a microchannel reactor: Performance evaluation and kinetic modeling Yong Sun a,*, Zhe Jia a, Gang Yang b,c, Lian Zhang d, Zhi Sun e,** a

Edith Cowan University School of Engineering, 270 Joondalup Drive Joondalup WA 6027, Australia Anpeng High-tech Energy Corp, Beijing, China c National Engineering Laboratory of Acid/base Coupled Production Technology, Institute of Process Engineering, Chinese Academy of Sciences, Beijing, 100190, China d Monash University Department of Chemical Engineering, VIC 3800, Australia e National Engineering Laboratory for Hydrometallurgical Cleaner Production Technology, Institute of Process Engineering, Chinese Academy of Sciences, Beijing 100190, China b

article info

abstract

Article history:

Fischer-Tropsch (FT) synthesis was carried out in a microchannel reactor using an iron-

Received 10 July 2017

based catalyst. The performance of microchannel reactor was evaluated in the aspect of

Received in revised form

CO conversion versus time on stream, catalyst deactivation, pressure drop and gas hour

25 September 2017

space velocity. The result indicates an excellent mass and heat transfer in the micro-

Accepted 3 October 2017

channel reactor. The negative impact of external and internal film diffusional limitation

Available online 31 October 2017

could be avoided in this microchannel reactor at experimental conditions. The effect of reaction temperature, operational pressure, syngas ratio and space velocity upon CO

Keywords:

conversion and hydrocarbon selectivity were extensively investigated. The kinetic

Fischer-tropsch synthesis

modeling was conducted and the mechanisms i.e. carbide, enlic, alkyl, formate and CO

Iron-based catalyst

insertion were extensively explored. A mechanism derived from Eley-Rideal-type mecha-

Microchannel reactor

nism was found to be the most statistical and physical relevance at the experimental

Kinetic modeling

conditions during FT synthesis using iron-based catalyst in this microchannel reactor. © 2017 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

Introduction Natural gas, coal, biomass obtained from abundant resources i.e. agricultural, forestry industry etc, and biogas may be converted to fuels and bulk chemicals such as FT (FischerTropsch) synthesis products, methanol/or DME (dimethoxyethane), hydrogen and ammonia, via synthesis gas generated

by feedstock dependent combination of reforming, oxidation and gasification reaction [1e3]. To realize economics of this type of process, the scale needs to be over certain threshold. For example, the Pearl GTL (gas-to-liquid) operated by Pearl and Qatar petroleum with the productivity reaching 140,000 bpd (barrels per day) is regarded as a successful example of profitable economical scale of GTL processing [4]. However, there are many gas wells that are too small to meet the current

* Corresponding author. Edith Cowan University School of Engineering, 270 Joondalup Drive Joondalup WA 6027 Australia. ** Corresponding author. E-mail addresses: [email protected], [email protected] (Y. Sun), [email protected] (Z. Sun). https://doi.org/10.1016/j.ijhydene.2017.10.022 0360-3199/© 2017 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.

29223


economical productivity demands of industrial scale GTL. In addition, the limitation of transporting biomass to realize its cost-effectiveness calls for the processing scale being less than 500 bpd [5]. These scenario all drive both industries and researchers to pursue the alternative technologies that challenge the dependency of scale. Because of high degree of process intensification in field of strongly endothermic and exothermic reactions [6e9], the microstructured technology is regarded as one of reliable and promising solution for small and medium scale catalytic process, and many research groups and companies such as Velocys, and Micrometrics have been very active and led R&D researches in this area during the last decade [10e13]. As FT synthesis is the one of core technology for GTL and BTL (biomass to liquid), the room left for optimization of reaction system, preparation of high performance catalyst with better selectivity and longer stability still remains big. Because of advantages of utilizing low syngas (H2/CO) ratio, larger selectivity in petrol, diesel and olefin fractions, and cost-effectiveness in production when it is compared with cobalt based catalyst [14], the iron-based catalyst still shows its obvious attractive advantages in practical applications for high temperature FT synthesis [15e18]. Although the FT synthesis using iron-based catalyst in many different types of conventional reactor system i.e. fixed bed reactor (FBR), micro-fixed bed reactor, slurry phase CSTR (continuous stirred tank reactor) etc have been widely investigated [19e21], the reports of using iron-based catalyst in the microchannel reactor are still very limited. For sake of process intensification, easiness of implementing into BTL process with a low syngas ratio and providing valuable data for comparisons among different reactor systems, the FT synthesis using iron-based catalyst in a microchannel reactor was therefore performed. In this work, the performance (both mass transfer and process parameters) of FT synthesis in microchannel reactor was evaluated, the effects of process parameters i.e. reaction temperature, operational pressure, syngas ratio and space velocity upon CO conversion and hydrocarbon selectivity were extensively investigated. In addition, the modeling of CO consumption rate in this microchannel reactor using Langmuir-Hinshelwood and EleyRideal mechanistic approach was also deployed to gain some insightful understandings of reaction route during CO initiation.

washed by de-ionized water until the pH value of slurry reached 7.0. After filtration, the K2CO3 (10 wt%) solution was mixed with the filter cake and the mixture was agitated in an ultrasonic bath for sake of better mixing. The obtained slurry was repeatedly filtered and washed and dried under 393 K for 36 h and then calcined at 723 K for 4 h with flow rate of air at 1 2.5 L.g1 cat.h . The prepared catalyst was then sieved to obtain the particle range of 37e75 mm. To test metal loading, the prepared catalyst was digested and the metal concentration was determined by ICP-OES (induced coupled plasma-optical emission spectrometer) using a microwave digestion method. The detailed property of the prepared catalyst is shown in Table 1.

Characterization of catalyst XRD (X-ray powder diffraction): Philips X-Pert diffractometer using Co Ka radiation at a wavelength of l ¼ 0.15406 nm with 2q changing from 10 to 90 at speed of 3 deg.min1. SEM morphology: JSM-7001F þ INCA X-MAX Field emission electron microscope. Specific surface area and porosity: Micromeritics ASAP 2020 automated gas sorption system using N2 adsorption at 77 K at a saturation pressure of 0.1 MPa. BET (BrunauerEmmett-Teller) specific surface area was assessed within the range of relative pressure from 0.05 to 0.3. The porosity was measured by mercury porosimeter (Auto Pore IV 9500, Micromeritics, USA). Elemental analysis: ICP-OES (OPTIMA 7100DV, Perkin Elmer, USA) with the aid of microwave, the detailed sample digestion procedures could be found from prior report [24].

Configuration of microchannel reactor and FT synthesis The FT synthesis was carried out in a microchannel reactor system. The schematic diagram of the system and the configuration of microchannel reactor are shown in Fig. 1. The dimension of stainless steel metal plate and a channel are 80 80 mm and 1 1 40 mm (depth width length), respectively. The Dh (hydraulic diameter) of the channel is 1 mm. The total volume of the channels is about 0.32 ml. The heater was inserted into both the top and bottom plates to keep

Experimental

Table 1 e Physical properties of catalyst and reactor system.

Catalyst preparation

Property

The catalyst in this work was prepared by the wet coprecipitation method [22,23]. About 10 g of Fe(NO3)3$9H2O and Cu(NO3)2$5H2O was firstly dissolved in 500 ml ultrapure water in a glass flask, after which a certain amount of silica solution was added and mixed well. Afterwards, 100 ml ammonia solution was added to this 500 ml mixture in flask with good agitation. During co-precipitation, the flask was placed in a 333 K thermostatic water-bath and the pH value of co-precipitated slurry was maintained at 9.0. After the coprecipitation, the slurry was filtered and the filter cake was

Bulk density/kg.m3 Porosity/Surface area/m2.g1 Volume of reactor/ml Mass of catalyst/g Average diameter/mm Tortuosity factor/CO diffusivity in wax/m2.s1

Value

ICP-OES

Composition (wt %)

1300 0.65 85 0.32 0.25 40 2 [47] 1.2 108

Fe Cu K Si O*

40.3 4.33 8.57 2.01 44.79

* represents by difference.

29224


Heater Thermal couple Top plate

Graphite sealing

Reaction plate

Bottom plate

9

TT

MCR

6 PRV

TT

7

P

TT

5

8

4 MFC MFC MFC

P

Hot

Hot

10 MFC

CO

5%H2/N2

Cold

12

11

2

1

H2 1

1

N2

Cold

3

MS

TCD

FID

GC-MS

FID

13

RGA

Fig. 1 e Simplified schematic diagram of reaction system, where 1 refers to gas cylinders, 2 reduction gas, 3 sulfur trap, 4 mass flow controllers, 5 pressure gauge, 6 pressure release valve, 7 ball valve for reduction line, 8 thermal couples, 9 microchannel reactor and configuration, 10 hot and cold trapping system, 13 after burner.

the microchannels maintaining at the setting temperature. With this configuration, no detectable temperature fluctuation was observed among all experimental conditions and maximum temperature difference between inlet and outlet is less than ±2 K. The high purity H2 and CO with different ratios were fed through the reactor at different flow rates. The tail gas was analyzed by an on-line RGA (refinery gas analyzer) and liquid collected from both hot and cold traps was analyzed by an off-line GC-MSD/FID (gas-chromatography-mass spectrometer detector/flame ionization detector, of which the split flow module was employed and the sample was sent into MSD and FID for qualitative and quantitative analysis). The ramping rates of the reactor were kept at 30 K min1 during reduction and at 1 K min1 during FT synthesis.

Catalyst performance and kinetic data collection Before FT synthesis, the catalyst was first reduced at 623 K 1 with 5 bar under a stream of 5% H2/N2 (v/v) at 2.56 L.g1 cat.h (the volumetric flow rate was calculated at STP-standard temperature and pressure) with a duration of 48 h. Then reactor was pressurized and heated up to the specific operational conditions. For microchannel performance evaluation and catalyst deactivation study, five different subgroups according to temperatures from 503 to 543 K, were set up. All subgroups started from baseline condition for 48 h, which is run 8 in Table 2 with detailed operational parameters as the following: pressure (2 MPa), syngas ratio (1.5) and GHSV 1 (6 L.g1 cat.h ). Afterwards, these subgroups were adjusted to


Table 2 e Reaction conditions and corresponding mass balances used for kinetic modeling. Runs

TOS/h

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17* 18* 19* 20* 21*

Parameters

24 36 51 92 132 176 220 259 310 343 380 412 490 525 559 592 80 92 104 116 128

SR

T

P

SV

1 1 1.5 1.5 1 2.5 2 2.5 2.7 2.7 1 1 1.5 2 1.5 1.5 1 1 1 1 1

503 513 533 533 533 533 533 523 533 533 533 533 533 533 533 533 503 513 523 533 543

3 3 3 2 3 2 2 2 2 2.2 2.1 2.5 3 3 2.6 3 3 3 3 3 3

6.0 6.0 8.04 6.0 6.0 6.0 6.0 5.34 6.0 6.72 7.38 8.04 5.34 6.72 7.38 8.04 6.0 6.0 6.0 6.0 6.0

Xco/%

Mass balance/%

16.81 28.91 16.90 27.95 35.90 63.01 65.77 41.20 39.43 37.81 36.58 35.07 36.77 30.21 26.50 25.89 17.14 29.28 36.16 65.62 69.18

93.5 95.3 95.0 91.6 95.0 98.3 90.9 94.3 91.2 97.6 90.6 90.1 97.0 91.5 92.1 91.3 90.2 93.4 95.5 94.9 90.4

rCO ¼

Selectiviy% ¼

4 X i¼2

ni SAi

100 FCO;in FCO;out FCO2 ;out

(1)

Evaluation of mass transfer in microchannel reactor To estimate the external mass transfer limitations, the Mears criterion [25] was employed as the following: M¼

rFT rs rn km C

(3)

. 3:24 PCO PCO2 PH2 P1 H2 O Kp 1 þ 0:003PCO PH2 P1 H2 O

(6)

dependent constants and can be correlated as the followings [23e26]: 1 1 A ¼ 8:85 106 exp 4494 493:15 T

(7)

1 1 B ¼ 1:93 102 exp 8236 493:15 T

(8)

Kp ¼

(mol.h1), FCO2 ;out CO2 molar flow rate from outlet (mol.h1), XCO CO conversion (%),FCO;in CO molar flow rate from inlet (mol.h1), and FCO;out CO molar flow rate from outlet (mol.h1). For selectivity study, four different sets (T 513e543 K, P 2.0e4.0, 1 MPa, SR 0.7e2.7, GHSV 5.3e8.04 L.g1 cat.h ) of experiments were conducted. In each condition, the FT started from baseline condition for 48 h. Then, it was adjusted to different conditions accordingly. To ensure enough liquid product collection, the duration of each individual condition was maintained for 12 h before trap swap. For kinetic study, the experimental data set in Table 2 was performed in order to discriminate different kinetic models and obtain the parameters regressions.

(5)

ð1 þ BPCO Þ2

1 Where rCO2 is the CO2 consumption rate (mol.kg1 cats ) and it is obtained from WGS (water gas shift) reaction. And the WGS is assumed to occur independently to FT reaction and it proceeds with a formate intermediate mechanism [28,29]. The 1 rCO is the CO consumption rate (mol.kg1 cats ),PCO , PH2 O and PH2 are the partial pressure of the CO, H2O and H2 (MPa). In this work, the baseline condition was employed to calculate Mears criterion, and their corresponding partial pressure of different gases are CO (0.27 MPa), H2 (1.58 MPa), H2O (0.16 MPa) and CO2 (0.12 MPa), respectively. TheA, B and Kp are temperature

(2)

where SAi refers to the molar flow rate of hydrocarbon

(4)

APCO P3=4 H2

rCO2 ¼

different temperatures for deactivation study accordingly. The CO conversion and hydrocarbon selectivity were obtained from RGA and calculated on the basis of carbon mass balance as the followings: FCO;in FCO;out 100 XCO % ¼ FCO;in

1 s Where rFT is FT reaction rate (mol.kg1 cats ), r is the density of catalyst (kg.m3), ris catalyst radius (m), nis reaction order (herein, for easiness of calculation, we assuming it is a first order reaction, -), km is mass transfer coefficient (m.s1), and Cis concentration of reagent syngas gas (mol.m3). In order to 1 correlate rFT (mol.kg1 cats ), the generalized kinetic model for iron based catalyst is employed for estimation, since they were tested and evaluated by prior scholars with very broad experimental conditions in different reactor systems [26,27]. The detailed expressions were the followings:

rFT ¼ rCO þ rCO2

TOS represents time on stream (h), SR syngas ratio (), SV gas hour 1 space velocity (L.g1 cat.h ), * refers to experimental conditions by changing reaction temperature while other process parameters were kept the same to run 8 (baseline).

29225

5078:0045 5:897 þ 13:96 104 T 27:59 108 T2 T

(9)

For correlations of mass transfer coefficient km (m.s1) in a microchannel reactor, the spherical shapes of catalyst were assumed and the Fr€ossling correlation was employed as the followings [26]: Sh ¼ 2 þ 0:6Re1=2 Sc1=3 ¼

km dp De

(10)

Re ¼

dp U y

(11)

Sc ¼

De y

(12)

Where Sh is Sherwood dimensionless number (), Re Reynolds dimensionless number (), Sc Schmidt dimensionless number (), Uthe superficial gas velocity (m.s1), nA molecular volume of CO (m3.g1.mol1), andy kinematic viscosity (m2.s1), andDAB is diffusivity of CO in wax (m2.s1). In this work, the Akgerman's correlation and the CO diffusivity in wax was used to approximate DAB at baseline condition

29226


(1.2 108 m2 s1) [30]. The detailed properties of catalyst and reactor conditions are summarized in Table 1. The gas-solid mass transfer coefficient is estimated to be 0.011 (m s1) at experimental conditions in the microchannel reactor. The corresponding Mears criteria value is estimated to be around 0.001, which is smaller than 0.15, indicating the ignorable external diffusional limitation at experimental conditions in this microchannel reactor. To estimate the intra-particle diffusional limitation, the dimensionless Thiele modulus was employed as the followings [30,31]: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi rs ðrCO Þ De CCO

(13)

ε De ¼ DAB t

(14)

Where Lg is geometry characteristic length (m), rs density of 1 catalyst (kg.m3), rCO CO consumption rate (mol.kg1 cats ), 2 1 De effective diffusivity (m .s ), εporosity of catalyst particle (), ttortuosity factor (), and CCO concentration of reagent gas (mol.m3). The calculated Thiele modulus on baseline condition is 0.11, which is less than 1, indicating the negative impact of intra-particle film diffusional limitation could be avoided by employing particle size range of 37e75 mm in microstructured reactor.

Performance evaluation of microchannel reactor The CO conversion versus time on stream (TOS) of five subgroups at the initial stage (baseline condition) is shown in Fig. 2a. The CO conversion on TOS is smooth and no

100

Subgroup_1 Subgroup_2 Subgroup_3 Subgroup_4 Subgroup_5

60 40

XCO ðtI Þ XCO ðtÞ 100% XCO ðtI Þ

(15)

where DA is the degree of deactivation of the catalyst, XCO ðtI Þ CO is conversion attI (h) on TOS when the reaction is stabilized, and XCO ðtÞis conversion at t (h) on TOS when the condition is adjusted back to the baseline condition. The profile of all five subgroups catalyst deactivation is shown in Fig. 2b. With good heat transfer in this microchannel reactor, the DA of all subgroups presents a similar range of less than ±5% catalyst deactivation over the period of 240 TOS. The pressure drop variation with the GHSV is shown in Fig. 2c. For microchannel reactor configuration in this work, no significant pressure drop was observed indicating the packing of fine particle in microchannel does not create significant pressure resistance across the microchannels. The CO conversion at different reaction temperature as a function of GHSV is shown in Fig. 2d. The responding patterns of CO conversion with different GHSV under different reactor temperatures show a similar pattern, indicating a good process performance in this microchannel reactor.

60

a

80

Xco/%

DA ¼

Xco/%

F ¼ Lg

appreciable fluctuation is observed, indicating an excellent heat transfer in the microchannel reactor system. After about 50 h, the CO conversion gradually stabilizes at around 42%. All subgroups present a similar CO conversion under the same initial conditions, this indicates that the reliable and reproducible data collected from microchannel reactor. In this work, in order to investigate the catalyst deactivation, the run 8 in Table 2 is chosen as the baseline conditions to assess the deactivation of catalyst and the cumulative degree of deactivation is defined as the following:

b

40

Subgroup_1 Subgroup_2 Subgroup_3 Subgroup_4 Subgroup_5

20

20 0

0

20

40

60

80

0 80

100

TOS/h

60

c

160

200

240

TOS/h

d

50

Xco/%

delta P/ bar.m

-1

1.2

120

0.8 2.0 MPa 2.5 MPa 3.0 MPa 3.5 MPa 4.0 MPa

0.4

0.0

5

6

7 -1 -1

GHSV/L.gcat h

8

533K 523K 513K 503K

40 30 20

5

6

7

-1 -1

8

GHSV/L.gcat h

Fig. 2 e Performance of microchannel reactor a) evolution of CO conversion at different TOS in different subgroups at initial stage, b) deactivation in different subgroups, c) pressure drop across reactor under different GHSV at different pressure while other conditions were kept at baseline condition, d) dependence of conversion and GHSV on reaction temperature while other conditions were kept at baseline condition.


model parameters were calculated by minimizing the sum of squares of residuals defined by calling the objective function as the following:

Reaction mechanism Reactor model In the microchannel reactor, the flow is laminar. The parabolic-type velocity profile in the microchannel will significantly increase complexity of reactor model. In this work, we use the averaged velocity profile to approximate the actual velocity profile during the development of the reactor model. Previous reports have indicated that using isothermal plug flow approximation in microchannel is valid when the diffusional limitation is not significant [13], which is also confirmed from the calculated Thiele number (0.11) in this work. However, the outcome of parity plot for CO consumption rate generally fall within the range of ±30%, with some exceptional higher up to ±40%, this indicates its limitation for reactor model approximation using this assumption. The outlet concentrations, which was related to the rate of reaction, were measured by an online RGA. The following differential equation was used to approximate reactor model: Wcat ¼ Fin;CO

rCO

Z

XCO;out

XCO;in

dXCO rCO

XCO Fin;CO ¼ Wcat

(16)

(17)

The corresponding boundary conditions (BC) were set as: W ¼ 0; Fi ¼ Fi;ðinletÞ W ¼ Wcatalyst ; Fi ¼ FiðexitÞ

mi PT Nc P mi

Fobj

! Nexp rexp rcal CO;i CO;i 1 X ¼ 100 exp Nexp i¼1 rCO;i

(20)

Where Nexp is the numbers of experimental conditions (), cal rexp co and rco are experimental and calculated molar rate for CO 1 consumption (mol.g1 cat.h ). The Levenberg-Marquardt (LM) was used for local optimization, while the genetic algorithm (GA) was utilized for global optimization. All programming was performed using MATLAB platform with built-in GA toolbox for model parameter regression. The MARR was calculated as the following: ! Nexp exp 1 X ri rcal i 100 MARR% ¼ exp Nexp j¼1 ri

(21)

For model discrimination, the physical relevance of the Arrhenius and Van't Hoff adsorption laws were employed as the followings: Ea;i ki ðTÞ ¼ ki exp