A Compact Kinetic Model for Biomass Pyrolysis at Gasification

Oct 2, 2017 - On the Validation of a One-Dimensional Biomass Pyrolysis Model Using .... Pharmaceutical industry productivity was robust in 2018, a yea...
1 downloads 0 Views 2MB Size
Subscriber access provided by UNIV OF ESSEX

Article

A compact kinetic model for biomass pyrolysis at gasification conditions. Himanshu Goyal, and Perrine Pepiot Energy Fuels, Just Accepted Manuscript • DOI: 10.1021/acs.energyfuels.7b01634 • Publication Date (Web): 02 Oct 2017 Downloaded from http://pubs.acs.org on October 2, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Energy & Fuels is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

A compact kinetic model for biomass pyrolysis at gasification conditions Himanshu Goyal† and Perrine Pepiot∗,‡ 1

Robert F. Smith School of Chemical and Biomolecular Engineering, Cornell University, Ithaca, New York 14853, USA, and Sibley School of Mechanical and Aerospace Engineering, Cornell University, Ithaca, New York 14853, USA E-mail: [email protected]

Abstract

2

3

Computational Fluid Dynamics (CFD) tools are increasingly gaining importance to

4

obtain detailed insight into biomass gasification. A major shortcoming of the current

5

CFD tools to study biomass gasification is the lack of computationally affordable chem-

6

ical kinetic models, which allows detailed predictions of the yield and composition of

7

various gas and tar species in complex reactor configurations. In this work, a detailed

8

mechanism is assembled from the literature and reduced to a compact model describing

9

the gas phase reactions of biomass gasification in the absence of oxygen. The reduction

10

procedure uses a graph-based method for unimportant kinetic pathways elimination

11

and quasi-steady-state species selection. The resulting reduced model contains 39 gas

12

species and 118 reactions, and is validated against the detailed model and two exper-

13

imental configurations: the pyrolysis of volatile species, such as levoglucosan (LVG), ∗

To whom correspondence should be addressed Robert F. Smith School of Chemical and Biomolecular Engineering, Cornell University, Ithaca, New York 14853, USA ‡ Sibley School of Mechanical and Aerospace Engineering, Cornell University, Ithaca, New York 14853, USA †

1

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

14

in a tubular reactor, and the fast pyrolysis of biomass particles in a droptube reac-

15

tor. The reduced model predicts the evolution of major gas products (e.g. CO, CO2,

16

CH4, H2) and various classes of tar (e.g. single-ring aromatics, oxygenated aromatics,

17

PAHs) produced during biomass gasification. The capability of the reduced model to

18

adequately capture the chemical process in a complex reactor geometry at an accept-

19

able computational cost is demonstrated by employing it in a simulation of a pseudo

20

two-dimensional laboratory-scale fluidized bed reactor (FBR).

21

Nomenclature

22

A

surface area of biomass particle based on an equivalent diameter (Section 4.3, m2 )

23

A∗

surface area of biomass particle derived from experimental data (Section 4.3, m2 )

24

dp

equivalent diameter of biomass particle (Section 4.3, m)

25

Eact

activation energy (J/mol)

26

H

enthalpy (J)

27

h

convective heat transfer coefficient (W/m2 K)

28

l

characteristic length of biomass particle (m)

29

nin

number of inflowing particles per time step in the PaSR

30

np

number of particles in the PaSR

31

npair

number of particles changing partners per time step in the PaSR

32

nR

number of reactions in kinetic model

33

nS

number of species in kinetic model

34

nstr

number of inflowing streams in the PaSR 2

ACS Paragon Plus Environment

Page 2 of 46

Page 3 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

35

Nu

Nusselt number

36

qconv

convective heat transfer rate (J/s)

37

∗ qconv

corrected convective heat transfer rate (J/s)

38

qrad

radiative heat transfer rate (J/s)

39

∗ qrad

corrected radiative heat transfer rate (J/s)

40

S

chemical source term in the PaSR

41

SRexp

solid residue in experiment (Section 4.3)

42

SRsim

solid residue in simulation (Section 4.3)

43

S1,S2

simulation cases (Section 4.3)

44

T

gas phase temperature (K)

45

Tp

biomass particle temperature (K)

46

Twall

wall temperature (Section 4.3, K)

47

t

time (s)

48

tend

integration time to calculate the errors for the targets (s)

49

tp

thickness of biomass particles (Section 4.3, m)

50

T

set of targets used in DRGEP

51

hT iD

average value of target T in the PaSR using the detailed model

52

hT iR

average value of target T in the PaSR using the reduced model

53

Tsp

trapped species that are released from biomass during devolatilization

54

Tsp∗

trapped species that remain inside biomass after devolatilization 3

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

55

Y

species mass fractions

56

Ychar

char yield (Section 4.3)

57

∆T

temperature difference between biomass particle surface and the surrounding gas (K)

58

59

∆t

PaSR simulation time step (s)

60

ǫT

a posteriori error on target T

61

Φ

composition of the mixture in the PaSR

62

Φi

composition of the ith particle in the PaSR

63

Φm

composition of the mixture after the mixing fractional step in the PaSR

64

Φstr

composition of the PaSR inflow stream

65

λf

gas thermal conductivity (W/m.K)

66

τmix

particle mixing time scale in the PaSR (s)

67

τpair

particle pairing time scale in the PaSR (s)

68

τres

residence time in the PaSR (s)

69

σ

Stefan-Boltzmann constant

70

ωc

char emissivity

71

ωp

biomass particle emissivity

72

ωw

wood emissivity

4

ACS Paragon Plus Environment

Page 4 of 46

Page 5 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

73

1

Introduction

74

Biomass (e.g. wood, energy crops, agricultural residue, municipal waste etc.) is recognized

75

as an essential renewable source of energy that can help in reducing the current dependence

76

on fossil fuels. Thermochemical conversion in fluidized bed reactors (FBR) is a promising

77

technology to convert low-value lignocellulosic biomass into high energy density gaseous or

78

liquid fuel. This process utilizes heat and/or physical catalysts to convert biomass to an

79

intermediate gas or liquid, followed by an additional conversion step to transform that gas

80

or liquid into a biofuel. It has the ability to robustly handle a wide range of feedstock and

81

to produce both liquid and gaseous fuels.

82

Thermochemical conversion of biomass can be divided into two major classes: pyrolysis

83

and gasification. Pyrolysis is performed at relatively low temperatures (773 K - 873 K)

84

maximizing the yield of liquid fuel, whereas gasification is performed at higher temperatures

85

(1073 K - 1273 K) maximizing the yield of gaseous fuel. In this work, we focus on the latter,

86

namely biomass gasification. One of the major challenges in making biomass gasification an

87

economically viable technology is the reduction or elimination of tars, which are complex

88

mixtures of condensable hydrocarbons. 1,2 Different tar species exhibit different properties,

89

for example, heterocyclic compounds (e.g. phenol) exhibit high water solubility, whereas

90

polycyclic aromatic compounds (PAH) can condense at relatively high temperatures. 2 Based

91

on its composition, tar can condense in downstream equipment causing fouling or plugging,

92

and can also produce hazardous tar-water mixtures, 2 and therefore needs to be removed. At

93

present, design and scale-up of FBR for biomass gasification are mostly empirically-based,

94

relying heavily on expensive and lengthy pilot-scale reactor studies. Yet, measurements

95

in these reactors are unlikely to be detailed enough to improve our understanding of tar

96

formation processes for various operating conditions and feedstocks, necessary to efficiently

97

optimize the conversion process. 3 Mathematical modeling and simulation tools provide a

98

much more flexible and affordable framework to investigate the controlling chemical and

99

physical processes, with the potential to play a determining role in the development and 5

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

100

deployment of the technology.

101

While the field has seen recent major advances, further improvements are still required,

102

especially in the description of chemical processes, before numerical tools can be utilized to

103

their full potential. Gomez-Barea et al., 4 in their review paper, recognize a strong need for

104

modeling efforts in biomass devolatilization and tar chemistry. But the chemistry of this con-

105

version process is extremely complicated to model due to the high variability of the feedstock,

106

the complex structure of biomass particles, as well as the interaction between chemistry and

107

the multi-phase flow dynamics typically found in gasification reactors. 5 These difficulties have

108

hindered the development of detailed kinetic models for biomass thermochemical conversion

109

chemistry, and have entailed the use of detailed mechanisms for combustion and pyrolysis of

110

various hydrocarbon species, developed in the combustion literature, to represent biomass

111

gasification. For instance, Debiagi et al. 5 and Norinaga et al. 6 have developed detailed ki-

112

netic models for thermochemical conversion of biomass starting from the kinetic models for

113

various hydrocarbons available in the combustion literature. However, the resulting detailed

114

mechanisms consist of a large number of elementary and non-elementary reactions (O(104 ))

115

and chemical species (O(103 )), making them computationally unaffordable to use in CFD

116

simulations. These detailed mechanisms are thus more suitable for zero-dimensional config-

117

urations neglecting the transport processes. In the absence of a computationally affordable

118

chemical mechanism, most of the existing modeling studies of biomass thermochemical con-

119

version either neglect the gas phase reactions 7–9 or use very simple kinetic models. 10–21 These

120

kinetic models describe biomass devolatilization and evolution of gas-phase primary products

121

using a few model compounds and global reactions, whose rates are fitted using available

122

experimental data, such as Thermo Gravimetric Analysis (TGA). While these global models

123

can be fitted a priori to provide trends in terms of the major controlling parameters, such

124

as reactor operating temperature, they are not appropriate whenever more quantitative or

125

detailed information is sought from CFD calculations. For instance, these models can not be

126

used to understand how tertiary tars are created in highly unsteady multiphase flows. An

6

ACS Paragon Plus Environment

Page 6 of 46

Page 7 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

127

intermediate level of chemical detail is then desirable that can provide refined predictions in

128

simulations, while remaining computationally affordable. The goal of this paper is to develop

129

such a model for biomass gasification. Note that we will focus on the initial volatile release

130

and subsequent gas phase evolution in the absence of oxygen, the resulting model requir-

131

ing to be complemented by a kinetic model to fully describe the long-term heterogeneous

132

reactions of gasification.

133

Several automated kinetic reduction techniques have been developed in the combustion

134

community. In general, these reduction techniques analyze a detailed mechanism for a given

135

set of conditions to predict the redundant species and reactions and remove them from the

136

chemical mechanism. Recently, Løvas ˙ et al. 22 used a combined reaction flow and sensitivity

137

analysis to develop a compact mechanism for gas phase reactions of biomass combustion.

138

This mechanism was developed in a homogenous reactor configuration for a fixed inlet gas

139

composition and variable temperature.

140

In this work, we focus on the secondary gas-phase reactions occurring at gasification

141

conditions, and compile a detailed mechanism from the literature describing those reactions.

142

We then use the DRGEP (Direct Relation Graph with Error Propagation) technique 23 to

143

extract a reduced model from the detailed one. The reduction procedure accounts for the

144

variability in the primary products expected to be found in gasification reactors by using

145

a statistical Partially Stirred Reactor (PaSR) configuration. Coupled with an appropriate

146

biomass devolatilization model (here, the work of Corbetta et al., 24 as described below), the

147

resulting reduced model describes the secondary gas phase reactions of biomass devolatiliza-

148

tion products in a pure nitrogen environment at temperatures relevant for gasification (1073

149

K-1273 K). Note that partial oxidation or steam reforming are not included in this study.

150

The remainder of this paper is organized as follows: Section 2 describes how the reference

151

chemical kinetic model for the gas phase chemistry and biomass devolatilization chemistry

152

are assembled from the literature. In Section 3, the automatic reduction procedure used to

153

generate a compact model that accurately reproduces the dynamics of the detailed model

7

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

154

is presented. Validation is detailed in Section 4. Finally, in Section 5, the applicability of

155

the reduced model in complex CFD configurations is demonstrated by simulating a pseudo

156

two-dimensional laboratory-scale FBR.

157

2

158

A description of the reference detailed chemical models for the solid biomass devolatilization

159

and the subsequent secondary gas-phase reactions of the primary devolatilization products

160

is first provided.

161

2.1

162

The high variability of feedstock and the structural complexity of biomass particles prevent

163

the development of detailed kinetic models for describing the transition of the solid biomass

164

into gas and char during devolatilization. In the absence of a more detailed description of

165

the biomass devolatilization, the lumped chemical model developed by the CRECK modeling

166

group 24 is used here to describe the first step of biomass gasification, i.e., devolatilization.

167

This model consists of 24 reactions involving 12 solid species, 7 trapped gases slowly releasing

168

from the solid matrix, and 20 gas phase products. The initial composition of biomass is rep-

169

resented by a combination of cellulose, hemicellulose, and 3 types of lignin. The rates of the

170

lumped reactions are fitted to match a series of thermogravimetric weight loss experiments.

171

It is worth noting that these reactions are irreversible, implying that the gas composition

172

surrounding the particle does not affect the chemistry going on at the particle level. This

173

model has been validated against a series of experiments for various operating conditions

174

and feedstock. 24–26

Reference detailed chemical kinetic model

Biomass devolatilization model

8

ACS Paragon Plus Environment

Page 8 of 46

Page 9 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

175

2.2

Primary product decomposition and tar formation

176

The biomass devolatilization model creates a variety of gas phase species, called primary

177

products, whose evolution in the gas phase at gasification temperature must be modeled.

178

These molecules usually are high molecular weight heterogeneous species, such as levoglu-

179

cosan or phenolic compounds. Some of these molecules, often found in combustion sys-

180

tems (e.g. phenol), have been extensively studied, and accurate mechanisms for their de-

181

composition to small hydrocarbons are available. Recently, we have developed and ex-

182

tensively validated a kinetic model for hydrocarbon pyrolysis and oxidation for combus-

183

tion applications, the latest version containing a large selection of alkanes up to dodecane

184

and aromatic species, such as phenol, toluene, benzene, xylene, two-ringed aromatics (e.g.

185

α-methylnaphthalene). 27–32 In the combustion process, ethylene is placed at the center of

186

molecular growth and we assume that the same holds true for the tars growth in biomass

187

pyrolysis and gasification. To assemble the detailed model, we combine several chemical

188

modules independent from one another, namely, a biomass devolatilization model to form

189

primary products from solid biomass, a primary product decomposition model, and a de-

190

tailed scheme for polycyclic aromatic hydrocarbon formation. A few of the primary gas phase

191

species produced by devolatilizing biomass are quite specific to the biomass constitutive com-

192

ponents, and have not received the same amount of characterization as typical combustion

193

molecules. To describe their decomposition, the lumped chemical reactions available from

194

Calonaci et al. 26 are used. To describe their decomposition, the lumped chemical reactions

195

available from Calonaci et al 26 are used. They include levoglucosan, 5-hydroxymethyl fur-

196

fural, hydroxyacetaldehyde, xylose, coumaryl, and their direct decomposition products (see

197

supporting information document for more details). In the pyrolysis conditions of relevance

198

here, the uni-molecular decomposition reactions of those compounds are controlling their

199

overall decomposition rates, as they are responsible for creating the initial radical pool. The

200

chemical mechanism on which those reactions are added being different from the one they

201

have been developed for, especially in terms of the small radicals chemistry, we found that 9

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

202

it was necessary to adjust slightly the uni-molecular decomposition rates for levoglucosan,

203

5-hydroxymethyl furfural, and hydroxyacetaldehyde to properly capture the rates experi-

204

mentally measured by Shin et al. 33 In those cases, the rates proposed by Shin et al. 33 were

205

in general adopted, staying as close to the branching ratios of Calonaci et al. 26 as possible

206

whenever competing reactions were involved. One exception is for hydroxyacetaldehyde, for

207

which Calonaci et al. and Shin et al. 33 decomposition pathways were combined and manually

208

adjusted to reflect the added pathways.

209

The detailed model consists of 396 molecular species and 3210 elementary reaction steps,

210

and is validated against the experiments by Shin et al. 33 in Fig. 1, and Norinaga et al. 6 in

211

Fig. 2. Shin et al. 33 studied the pyrolysis of levoglucosan (LVG), 5-hydroxymethyl furfural

212

(HMF), and hydroxyacetaldehyde (HAA) in a flow tube reactor for temperatures ranging

213

from 773 K to 1023 K. Norinaga et al. 6 studied the secondary pyrolysis of nascent volatiles

214

generated from the fast pyrolysis of cellulose in a tubular reactor in the temperature range of

215

973 K to 1073 K. Assuming that there is no significant axial mixing in the tubular reactors

216

of both experiments, these reactors are modeled as zero-dimensional isobar homogeneous

217

systems. As can be seen from Figs. 1 and 2, simulation results show very good agreement

218

with the experiments, especially considering the high uncertainty on initial conditions that

219

sometimes exist in the pyrolysis experiments. Additional validation cases considering the

220

evolution of phenolic and aromatics compounds can be found in the supporting information

221

document.

222

3

223

The reference kinetic model for the gas phase reactions described in the previous section

224

is too complex to be used even in simple CFD configurations. Therefore, the objective is

225

to identify the most important chemical reaction pathways for gasification. This section

226

describes how this objective is achieved by extracting a compact model with 39 species and

Reduced chemical model development

10

ACS Paragon Plus Environment

Page 10 of 46

Page 11 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

227

118 reactions from the reference model with 396 species and 3210 reactions. Note that the

228

reduction process does not affect the solid-to-gas devolatilization model 24 in any way.

229

3.1

230

As all reactions are not important at all conditions, the very first task is to identify the

231

conditions at which gasification will most likely take place in an actual gasifier, in particular,

232

the heating rate to which biomass will be subjected. This step will help refine the range

233

of conditions over which the reduced chemical kinetic model should be valid, and focus the

234

reduction procedure on the relevant kinetics. Table 1 shows the parameters used to repre-

235

sent the devolatilization of biomass particles in a FBR. The size of biomass particles varies

236

between 300 µm to 1 mm to represent the general size range found in many laboratory

237

gasification studies. Gas phase properties, such as density, specific heat capacity, conductiv-

238

ity, and viscosity, are computed assuming pure nitrogen at the temperatures and pressure

239

stated in Table 1. To estimate typical heating rates, the Nusselt correlation from Gunn 34 is

240

used along with the parameters from Table 1. For these parameters Biot numbers for the

241

biomass particles vary between 0.34 to 0.58, implying that thermal gradients will be present

242

in the biomass particles. However, the purpose of Table 1 is to establish relevant conditions

243

in order to apply our chemistry reduction algorithms. For this, an assumption of constant

244

internal temperature suffices. Simulations of biomass devolatilization using the Corbetta et

245

al. model, 24 and neglecting secondary gas phase reactions, show that most primary gases are

246

released from the biomass between 773 K and 873 K. With this assumption, Fig. 3 indicates

247

that particles of size between 300 µm and 1 mm experience a heating rate of O(103 ) K/s

248

during devolatilization. Therefore, a value of 1000 K/s is chosen as representative of the

249

heating rate for the reduction procedure.

Relevant gasification conditions

11

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

250

3.2

Simulation configuration

251

The relative importance of chemical reaction pathways as estimated by the DRGEP reduc-

252

tion methodology depends on the chemical compositions and sample kinetic trajectories on

253

which it is applied. As the reference model is too complex to be used in a realistic reactor

254

configuration, we choose a statistical treatment to sample as broadly as possible the chemical

255

states and trajectories occurring in a gasification reactor. For this purpose, the computation-

256

ally inexpensive and idealized Partially Stirred Reactor (PaSR) is used. Inside a PaSR, the

257

composition and properties of the fluid are represented by an ensemble of notional particles

258

each carrying its own species composition and temperature. The properties of each notional

259

particle evolve due to mixing, reaction, and inflow/outflow events such that the mean ther-

260

mochemical properties of the represented fluid are statistically spatially homogeneous, but

261

the fluid itself is imperfectly mixed at the molecular level. The use of a PaSR as sampling

262

tool forces the reduction to be quite conservative, thereby preventing important pathways

263

to be removed from the reduced kinetic model. It must also be noted that because a PaSR

264

is not a mathematical representation of a physical system, simulation results cannot not be

265

directly compared to, or interpreted in light of, experimental data.

266

The fluid is assumed to be an ideal gas-phase mixture that evolves in the PaSR at

267

a constant pressure, so that the full thermochemical state or composition of the mixture

268

Φ is completely characterized by the species mass fractions Y and the mixture enthalpy

269

H : Φ ≡ {Y, H}. The PaSR is continuously fed by a user-specified number nstr of inflow

270

streams of prescribed compositions Φstr ; it will be described later for the biomass system

271

under consideration. At any time t, the reactor contains a constant, even number np of

272

notional particles, the ith particle having composition Φ(i) (t). These compositions evolve in

273

time due to mixing, reaction, and inflow and outflow events. These processes are described

274

in more detail. Inflow and outflow events occur at discrete times and change the particle composition Φ in a discontinuous manner. In the inflow/outflow event, nin particles are selected at 12

ACS Paragon Plus Environment

Page 12 of 46

Page 13 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

random with equal probability, and their compositions are replaced by the inflow streams’ compositions. The integer number nin (= np × ∆t/τres ) is chosen according to the specified mean residence time τres and time step ∆t. Between these discrete times, the composition evolves by a mixing fractional step and a reaction fractional step. In the mixing fractional step, particles are paired and ordered so that particles i and i+1 are partners for odd i (1≤ i < np ), and the ordinary differential equations, Φ(i) (t) − Φ(i+1) (t) dΦ(i),m =− dt τmix



275

Φ(i+1) (t) − Φ(i) (t) dΦ(i+1),m =− dt τmix



(1)

276

are solved for each pair of particles over time interval ∆t. In this equation, τmix is the

277

specified pairwise mixing time scale. At each time step, npair particles are selected randomly

278

with equal probability and shuffled to change partners. The integer number npair (= np ×

279

∆t/τpair ) is chosen according to the specified pairing time τpair , typically taken equal to τmix .

280

The compositions after mixing evolve under isobaric, adiabatic conditions over a time ∆t

281

according to  dΦ(i),m (t) = S Φ(i),m (t) dt

(2)

282

where S is the chemical source term defined by the user-provided reaction mechanism. This

283

reaction fractional step finally yields the particle compositions at t + ∆t: Φ(i) (t + ∆t).

284

The PaSR simulation setup described here will be used to provide relevant compositions

285

of biomass devolatilization products expected to be found in real gasification reactors and

286

to compare the reduced model developed in this section with the reference model.

287

3.3

288

The automatic chemical mechanism reduction technique DRGEP 23 is used to extract a

289

reduced model from the reference, detailed, model. The reduction procedure follows the

Reduction using DRGEP

13

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

290

steps outlined in Pepiot et al. 23 for species and reaction elimination, and is performed using

291

the YARC reduction tool, 35 a Perl/C implementation of DRGEP and associated reduction

292

techniques.

293

• Reduction targets selection The first step in the reduced model development is

294

to identify a set of targets T , most often specific species, that the reduced model

295

must reproduce accurately. In biomass gasification, it is desirable to predict the yield

296

of gaseous products and tar species. Therefore, 5 major gas products: CO, CO2 ,

297

H2 , CH4 , and C2 H4 , water (H2 O), and 3 major tar species: benzene (C6 H6 , a single-

298

ring aromatic), naphthalene (C10 H8 , a polycyclic aromatic), and phenol (C6 H6 O, an

299

oxygenated aromatic) are selected as targets. Moreover, 13 primary devolatilization

300

products described by the reference devolatilization model, such as HAA, HMFU, LVG,

301

are also incorporated into the targets list.

302

• Sample Composition Database To evaluate the relative importance of species and

303

reactions for the given set of targets, DRGEP requires an ensemble of sample compo-

304

sitions representative of the simulations in which the reduced model will eventually be

305

used. For this purpose, we use a PaSR configuration, and assume that the particles in

306

the PaSR simulation will follow trajectories in composition space that are representa-

307

tive of those they would encounter in an actual reactor. The simulation parameters are

308

chosen based on previous experience and best practices 36 to ensure a broad range of

309

compositions relevant for our application, and the residence time is adjusted to match

310

the characteristic timescale of the overall pyrolysis chemistry process.

311

Two inflow streams are continuously fed to the PaSR to represent the release of the

312

primary products from the devolatilizing biomass into the hot nitrogen environment.

313

The first inflow stream consists of nitrogen gas at temperatures varying between 1073

314

K to 1273 K, while the second inflow stream consists of the primary products released

315

during biomass devolatilization. The second stream needs to account for the fact that

14

ACS Paragon Plus Environment

Page 14 of 46

Page 15 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

316

a gasification reactor contains biomass particles at different stages of devolatilization,

317

acting as variable sources of primary products. To include this variability, the biomass

318

devolatilization process is represented stochastically by sampling from the probability

319

distribution function (PDF) of the extent of biomass devolatilization. This PDF is

320

constructed by simulating biomass devolatilization a priori using the reference chemical

321

model.

322

The parameters used for these simulations are summarized in Table 2. The PaSR

323

simulations are performed for three nitrogen temperatures: 1073K, 1173K, and 1273K.

324

A database of 18,000 distinct chemical compositions is created by randomly sampling

325

the compositions encountered in the PaSR simulations.

326

• Automatic reduction and error estimation The automatic reduction procedure

327

proceeds through two distinct steps. In the first step, DRGEP analyzes the composition

328

database and quantifies the coupling between species and reactions in the chemical

329

mechanism for the chosen target species in the form of importance coefficients also

330

known as DRGEP coefficients. Species and reactions with the lowest value of DRGEP

331

coefficients are removed from the mechanism in an iterative manner providing a list

332

of kinetic models of decreasing complexity. More details about the implementation of

333

DRGEP technique can be found for example in. 37

334

In the second step, the PaSR test configuration (parameters provided in Table 2) is

335

simulated using each of the reduced models generated in the first stage, and a posteriori

336

errors on the targets are computed, defined for any target T as

εT =

R tend 0

| hT iD (t) − hT iR (t)|dt R tend | hT iD (t)|dt 0

(3)

337

In this equation, hT i (t) designates the average of quantity T at time t over all particles

338

contained in the PaSR, and tend is taken here as 15 PaSR residence times. A posteriori

339

errors as a function of the number of species, nS for a few targets are shown in Fig. 4. 15

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

340

The reduction process provides several mechanisms with decreasing number of species

341

and reactions; we choose the smallest possible mechanism for which the error is less

342

than 10% for most of the species, and at most a factor of 2 for a few groups of species.

343

The mechanism with 60 chemical species and 486 reactions is found to be the smallest

344

acceptable model generated automatically by the DRGEP procedure and is shown by

345

a dotted line in Fig. 4.

346

3.4

Additional reduction

347

In the DRGEP technique, a species or a reaction is removed from the chemical mechanism

348

only when it is identified as unimportant for every single composition in the database, which

349

imposes a stringent criterion on species and reaction removal. A species or a reaction can

350

be important for the local dynamics of a target, but may not impact its global statistical

351

behavior. Several techniques have been developed in recent years to identify those addi-

352

tional species and reactions, for example, the DRGEP with Sensitivity Analysis method. 38

353

In this work, we use an ad-hoc semi-automatic technique that quantifies the impact of species

354

and reactions on the global statistical behavior of the targets. This technique utilizes the

355

global production/consumption rates of each species for every reaction obtained from the

356

simulations of the PaSR test configuration using the intermediate mechanism. The coupling

357

between species and reactions on the targets is quantified in a manner similar to the DRGEP

358

technique, but using the global production and consumption rates of species instead. Po-

359

tential species and reactions that may have a minimal impact on the prediction of global

360

statistics of the targets are removed from the mechanism, and the resulting model is sim-

361

ulated in the PaSR test configuration to calculate a posteriori errors on the targets using

362

Eqn. 3. A 44 species and 118 reactions mechanism is found to be the smallest acceptable

363

model after this step, and is shown by diamond symbols in Fig. 4.

16

ACS Paragon Plus Environment

Page 16 of 46

Page 17 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

364

3.5

Quasi-steady state approximations

365

Once the above-mentioned strategies have removed as many species and reactions as pos-

366

sible, quasi-steady state (QSS) approximations are introduced that replace the differential

367

equation for a given species by an algebraic expression much faster to solve. All suitable

368

QSS species are computed using algebraic expressions consisting of non-QSS species. To

369

keep the calculation of QSS species simple and fast, it is made sure that all of the algebraic

370

expression are linear. 39 With this constraint, 5 QSS species are identified: C2 H5 , CH2 OH,

371

CH2 CHO, C7 H7 , and CH2 CO. The final reduced kinetic model has 39 non-QSS species (including N2), 118 reactions

372

373

(including both forward and backward reactions), and 5 QSS species.

374

4

375

In this section, the accuracy of the reduced model (39 species and 118 reactions) developed in

376

Section 3 is assessed for a number of configurations. Three different test cases are performed:

377

1) The PaSR test configuration is simulated using both the reduced and the reference model,

378

and major product and tar species are compared; 2) The reduced model is used in a zero-

379

dimensional reactor configuration to simulate the pyrolysis experiments of Shin et al.; 33 3)

380

The reduced model is integrated with the CFD solver NGA 40 to simulate biomass gasification

381

in the laboratory-scale Drop Tube Reactor (DTR) of Chen et al. 41 These validation cases

382

are discussed in the following subsections. Additional validation considering the pyrolysis of

383

phenol can be found in the supporting information document.

384

4.1

385

The PaSR configuration used to create composition database in Section 3 is used again here

386

to compare the reduced model to the reference model for the temperatures ranging from

387

1073 K to 1273 K. The parameters used for these simulations are shown in Table 2. After a

Validation of the reduced model

Comparison in a partially stirred reactor

17

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

388

statistically steady state is reached, the mass fractions of gaseous products and tar species are

389

averaged over 10 residence times to get mean steady state mass fractions. Since the reaction

390

pathways are significantly altered due to the high reduction ratio, the mass fraction of a

391

few species are not compared individually. Instead, these species are divided into different

392

groups based on their molecular weights, and the sum of their mass fractions is compared.

393

Figure 5 compares the mean steady state mass fractions of relevant individual species

394

and groups of species obtained from the reduced and reference model. Predictions of the

395

reduced and reference model for major product gases: CO, CO2 , CH4 , H2 , and C2 (species

396

with 2 carbon atoms), water (H2 O), and three classes of tars: single-ring aromatics (A1 ),

397

poly-aromatic hydrocarbons (PAH), and oxygenated aromatics (AO) are in good agreement.

398

In addition, the reduced model is also able to predict light (LPV), medium (MPV), and

399

heavy (HPV) weight primary devolatilization products, and small radicals pool (RAD).

400

Those acronyms and their definitions are summarized in Table. 3. Note that the assumption

401

of constant heating rate used in the reduction procedure is shown to have negligible impact

402

on the results, as is described in the supporting information document.

403

The PaSR simulations are carried out on a Beowulf cluster with Nehalem X series pro-

404

cessors. The time per iteration per processor for using the reduced model is O(10−3 s) and

405

for the reference model it is O(10−1 s), corresponding to a reduction in the CPU time by

406

∼99% by the reduced model. It must be noted that in the PaSR simulations the majority

407

of the time is spent on the integration of the chemical source terms, therefore, the saving in

408

the CPU time is dominated by this term. Reduction in computational expense is expected

409

to be even higher in CFD simulations as additional scalar transport equations need to be

410

solved for each species at every grid point in the computational domain.

411

4.2

412

The reduced model is used to simulate the tubular reactor experiments of Shin et al. 33 de-

413

scribed in Section 2.2. Simulation results are compared with the experimental measurements

Pyrolysis in a tubular reactor

18

ACS Paragon Plus Environment

Page 18 of 46

Page 19 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

414

in Fig. 6, and show overall a good agreement. When compared to the simulations performed

415

with the detailed, reference mechanism (Fig. 1), we see that the prediction of hydroxymethyl

416

furfural decomposition is virtually unchanged by the reduction process, but more significant

417

changes are observed for levoglucosan and hydroxyacetaldehyde, for which the decomposi-

418

tion rate has been reduced. While agreement with the experimental data is still satisfactory,

419

those results indicate a larger sensitivity of those molecules to the underlying small radical

420

chemistry.

421

4.3

422

Numerical simulations of the one-dimensional DTR of Chen et al. 41 are conducted and

423

compared to the experimental results. A schematic of the experimental DTR can be seen

424

in Fig. 7. Chen et al. 41 studied gasification of millimetric sized biomass particles (beech

425

wood) in the DTR at 1073 K and 1223 K. In the experiments, particles are flake-like and are

426

characterized by their equivalent spherical diameters. Biomass particles and nitrogen stream

427

are continuously injected from the top of the reactor, while the exhaust gas is sampled at

428

the bottom. A portion of this exhaust gas is then examined by several gas analyzers. The

429

distance between the locations of biomass injection and gas collection is varied to get four

430

residence lengths: 0.3 m, 0.5 m, 0.7 m, and 0.9 m. The total amount of gas, tar, and

431

char produced is measured at these residence lengths. Moreover, yields of major gas phase

432

components are also provided.

Fast pyrolysis of biomass in a drop tube reactor

433

The DTR presents a multiphase and multiphysics system, therefore, simulations of this

434

reactor require a reactive multiphase flow solver. For this purpose, the reduced model and

435

the biomass devolatilization model are integrated with the reactive multiphase CFD solver

436

NGA, 40 with a Euler-Lagrange strategy 42 to model gas-solid flows. NGA has been exten-

437

sively validated and used for various DNS and LES multiphase reactive flow systems. 43–50

438

Simulations were conducted for biomass particles with the equivalent spherical diameter (dp )

439

of 520 µm and two gas temperatures: T = 1073 K (simulation S1) and 1223 K (simulation 19

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

440

S2). Parameters used for S1 and S2 are provided in Table 4. In the DTR simulations, two

441

modeling issues are encountered related to: 1) the reference devolatilization kinetics model

442

and 2) the shape of the particles. These are discussed in the following two subsections before

443

the results of the DTR simulations are presented.

444

Biomass devolatilization model. In the DTR simulations, the biomass devolatiliza-

445

tion chemistry is modeled by the reference devolatilization model discussed in Section 2.1.

446

Preliminary DTR simulations could not accurately predict the experimental yields of CH4 ,

447

C2 H4 , and solid residue. This difference is attributed to the fact that a significant portion of

448

these two gases remains trapped in the solid matrix in the chemisorbed state. For simula-

449

tion S1, Fig. 8(a) shows the evolution of the trapped species (Tsp ≡ CO, CO2 , and CH3 OH)

450

that are completely released from the biomass particle during the devolatilization, whereas

451

Fig. 8(b) shows that a few trapped species (Tsp∗ ≡ COH2 , CH4 , C2 H4 , and H2 ) remain inside

452

the biomass particle even after the complete devolatilization. Figure 8(c) shows that even

453

long after the completion of biomass devolatilization, the amount of solid residue is much

454

higher than that of char; this difference is also attributed to the trapped species Tsp∗ . This

455

is corroborated by the fact that the predicted value of char, Ychar =0.10 is close to the solid

456

residue measured in the experiments, SRexp =0.08 ± 20%.

457

Similar discrepancies have been very recently investigated by Anca-Couce et al., 51 who

458

performed biomass pyrolysis experiments and used the reference devolatilization model to

459

predict the experimental yields of various species. They introduced several modifications in

460

the devolatilization mechanism to significantly improve the agreement between the modeled

461

results and their experimental database, mainly for the yields of light hydrocarbons and the

462

yield and composition of char. The focus of the present paper being the secondary gas-phase

463

reactions, we introduce here a simple ad hoc modification of the Corbetta et al. model 24

464

considering only the current experiment at hand, as described below, and refer the reader

465

to the study of Anca-Couce et al. 51 for a more comprehensive treatment of this issue. To

466

improve the predictions of CH4 , C2 H4 , and solid residue, we adjust the parameters of the

20

ACS Paragon Plus Environment

Page 20 of 46

Page 21 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

467

reactions, present in the reference devolatilization model, governing the release of Tsp∗ . In

468

the reference devolatilization model, activation energies (Eact ) for the release of Tsp∗ are much

469

higher compared to Eact for the release of Tsp . To make the release of Tsp∗ faster, we replace

470

the Eact for the release of Tsp∗ by the Eact for the release of trapped CO. Simulation S1 is

471

repeated with the modified values of the Eact ; the resulting evolution profiles of Tsp and Tsp∗

472

are shown in Fig. 8(d) and Fig. 8(e), respectively, and the evolution of char and solid residue

473

is shown in Fig. 8(f). These figures show that as the devolatilization proceeds all the trapped

474

species get released from the biomass particles and the amount of solid residue, SRsim =0.1 is

475

close to the experimental value, SRexp =0.08 ± 20%. Therefore, the reference devolatilization

476

model with the modified value of Eact for Tsp∗ is used in the DTR simulations.

477

Shape of the particles Biomass particles used in the experiments have flake-like shape.

478

The shape of the particle affects the drag force from the surrounding gas and the heat transfer

479

rate experienced by the particle. In the simulations particles are treated as spheres, therefore,

480

a correction must be made to include the effect of the proper particle shape while calculating

481

drag force and heat transfer rate. Chen 52 experimentally measured the slip velocity of the

482

biomass particles and estimates a correction factor of 1.5 that can be multiplied with the drag

483

correlation for a spherical particle to estimate the drag on a flake-like particle. In the DTR

484

simulations, this correction factor is used in the drag calculation for the biomass particles.

485

Although drag is corrected for the flake-shaped particles, any correction for heat transfer

486

rate is not provided in the experimental study. Therefore, we calculate correction factors to

487

estimate the convective and radiative heat transfer rates of the flake-shaped particles based

488

on the calculation for spherical particles.

489

Convective heat transfer rate for a particle can be expressed as

qconv = hA∆T

490

(4)

where A is the surface area of the particle, ∆T is the temperature difference between the

21

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 22 of 46

491

particle surface and the surrounding gas, and h is the convective heat transfer coefficient.

492

h can be calculated from the Nusselt number, N u, as h =

493

conductivity of the gas surrounding the particle and l is the characteristic length, which is

494

equal to the diameter (dp ) for a sphere and the thickness (tp ) for a flake-shaped particle. The

495

Nusselt number is calculated using Gunn’s correlation. 34 Average area of the flake-shaped

496

particles is calculated based on the experimental measurements 52 of particle dimensions. The

497

ratio of the average area of the flake-shaped particles (A∗ ) to that of an equivalent spherical

498

particle (A) is calculated to be 1.33. Using the thickness, tp , as the characteristic length in

499

the expression for h, the convective heat transfer rate for the flake-shaped particles becomes

∗ qconv

=



dp h tp



N uλf , l

where λf is thermal

 A∗ A ∆T A

(5)

500

For a biomass particle of equivalent diameter dp = 520 µm, the experimentally measured

501

average particle thickness (tp ) is 250 µm. Substituting these values in Eq. 5, we get ∗ qconv

=



 520 h (1.33A) ∆T ∼ 2.8hA∆T = 2.8qconv 250

(6)

502

Equation 6 implies that the convective heat transfer rate for the flake-shaped particles (cor-

503

responding to an equivalent spherical diameter of 520 µm) is about 2.8 times faster than

504

that for the equivalent spherical particles. The radiative heat transfer rate from the reactor walls to the biomass particle is modeled

505

506

by 4 qrad = Aωp σ(Twall − Tp4 )

(7)

507

where, ωp is the particles emissivity, σ is the Stefan-Boltzmann constant (= 5.6704×10−8

508

Wm2 K−4 ), Twall is the reactor wall temperature, and Tp is the particle surface temperature.

509

ωp is calculated as a linear combination of wood (ωw =0.7) and char emissivity (ωc =0.92). 53

22

ACS Paragon Plus Environment

Page 23 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

510

Energy & Fuels

For a flake-shaped particle, A is replaced by A∗ , and we get

∗ qrad

=



A∗ A



4 Aωp σ(Twall − Tp4 ) = 1.33qrad

(8)

511

Equation 8 implies that the radiative heat transfer rate for the flake-shaped particles (cor-

512

responding to an equivalent spherical diameter of 520 µm) will be about 1.33 times faster

513

than that of the equivalent spherical particles.

514

To evaluate the effect of these corrections for convective and radiative heat transfer rates

515

∗ ∗ on biomass devolatilization, S1 is performed with the corrected rates (qconv and qrad ) and the

516

uncorrected rates (qconv and qrad ). Fig. 9 shows that using the corrected heat transfer rates

517

significantly improves the prediction of the shrinkage rate of the particles. Therefore, in the

518

DTR simulations, convective and radiative heat transfer rates for the spherical particles are

519

multiplied with 2.8 and 1.33, respectively, to make correction for the shape of the particles.

520

Comparison with experimental data. After incorporating the modifications in the

521

reference devolatilization model and the heat transfer rates, simulations S1 and S2 are

522

run until steady state is reached. Figure 10 compares the simulation predictions of the

523

mass fraction of major gas products and particle diameters at various reactor lengths to

524

the experimental values. Agreement between the simulation predictions and experimental

525

measurements is very good considering the possibility of the high degree of variability in

526

various parameters and physical properties. These simulations are performed on a single

527

core of a MacBook laptop and required O(1 hour) to reach steady state, which shows the

528

affordability of the current reduced model to simulate laboratory-scale reactors.

529

5

530

The reduced gas-phase chemistry model, coupled with the biomass devolatilization model

531

of Corbetta et al., 24 is used to simulate a pseudo-2D configuration (rectangular geometry)

532

of an experimental FBR 54 using NGA. 40 Parameters used in this simulation are reported in

Application to a fluidized bed reactor

23

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

533

Table 5. Initially, the sand bed is fluidized without biomass particles by injecting the nitrogen

534

gas from the bottom of the reactor. Once a fluidized sand bed is achieved, biomass particles

535

are injected into the reactor at a constant mass flow rate. Simulation is run long enough

536

to reach a statistically steady state. Figure 11 shows the instantaneous values of the mass

537

fraction of various classes of tar normalized by their maxima at statistically steady state.

538

The location of the mass fraction maximum of oxygenated aromatics (AO) is very different

539

from that of single-ring (A1 ) and multiple-ring (PAH) aromatics. It indicates that different

540

tar species can have different length and time scales associated with their formation and

541

consumption. The mass fraction of the major gas and tar species at different reactor lengths

542

are shown in Fig. 12. As expected, CO is the major gas product followed by CO2 , CH4 , C2 ,

543

and H2 . Among tars, single-ring aromatics are the major species followed by oxygenated

544

aromatics, and Polycyclic Aromatic Hydrocarbons (PAH). Another important observation

545

made from Fig. 12 is that the mass fraction of all the light gases, and A1 and PAH increase

546

along the reactor height, while it decreases for AO.

547

The simulation was performed on 96 cores on the cluster mentioned in Section 4.1, and

548

required 3000 CPU hours and 9000 CPU hours per flow-through time (0.75 s) for pure sand

549

fluidization case and the reacting case with biomass injection, respectively. This simulation

550

shows the ability of the reduced model to be used with a CFD solver to simulate laboratory-

551

scale FBR in an affordable manner. The present reduced model combined with a CFD

552

solver provides the capability to track the evolution of major gas and tar species for different

553

operating conditions.

554

6

555

An adequate description of the chemical kinetics in the CFD tools is imperative for the

556

detailed simulations of biomass gasification, however, the large size of detailed mechanisms

557

make their use prohibitive in the CFD simulations. In this work, we assemble a detailed

Conclusion

24

ACS Paragon Plus Environment

Page 24 of 46

Page 25 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

558

chemical model (396 species, 3210 reactions) for the secondary gas-phase reactions of biomass

559

gasification and reduce it to a compact model (39 species, 5 quasi-steady state species,

560

and a total of 118 reactions) using automated strategies. The reduced model shows very

561

good reproducibility of the statistical yields of various species of interest at a fraction of

562

computational cost compared to the detailed model. The savings in computational time are

563

expected to be higher in CFD simulations where a set of extra Partially Differential Equations

564

(PDEs) need to be solved for the scalar transport equations. The reduced model, integrated

565

with the CFD solver NGA, is used to simulate a laboratory-scale Drop Tube Reactor (DTR)

566

experiment showing good agreement with the experiments. By simulating a pseudo two-

567

dimensional FBR with the reduced model, it is shown that an adequate description of the

568

gas phase reactions can be used with CFD tools in a computationally affordable manner.

569

The reduced model developed here is small enough to be integrated with a CFD solver to

570

study the secondary gas phase reactions of biomass gasification in laboratory-scale reactors.

571

Acknowledgement

572

This material is based in part upon work supported by the National Science Foundation

573

under Grant Number EEC-1342362.

574

Supporting Information Available

575

The following file is available free of charge.

576

• Supporting information file: Additional validation cases for the reference kinetic model

577

and the reduced model; Partially Stirred Reactor case with a variable heating rate; and

578

the reduced kinetic model in CHEMKIN format along with the thermodynamic and

579

transport data for all the gas phase species.

580

This material is available free of charge via the Internet at http://pubs.acs.org/.

25

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

581

References

582

(1) Palma, C. F. Applied energy 2013, 111, 129–141.

583

(2) Van Paasen, S.; Kiel, J.; Veringa, H. Tar formation in a fluidised bed gasifier ; 2004.

584

(3) Van der Hoef, M. A.; van Sint Annaland, M.; Deen, N. G.; Kuipers, J. A. M. Annu.

585

586

587

588

589

Rev. Fluid Mech. 2008, 40, 47–70. (4) Gómez-Barea, A.; Leckner, B. Progress in Energy and Combustion Science 2010, 36, 444–509. (5) Debiagi, P. E. A.; Gentile, G.; Pelucchi, M.; Frassoldati, A.; Cuoci, A.; Faravelli, T.; Ranzi, E. Biomass and Bioenergy 2016, 93, 60–71.

590

(6) Norinaga, K.; Shoji, T.; Kudo, S.; Hayashi, J.-i. Fuel 2013, 103, 141–150.

591

(7) Brunchmuller, J.; van Wachem, B.; Gu, S.; Luo, K.; Brown, R. AIChE Journal 2012,

592

593

594

595

58, 3030–3042. (8) Bruchmüller, J.; Luo, K. H.; Van Wachem, B. G. M. Proceedings of the Combustion Institute 2013, 34, 2373–2381. (9) Ku, X.; Li, T.; Løvås, T. Energy & Fuels 2015, 29, 5127–5135.

596

(10) Miller, R.; Bellan, J. Combustion Science and Technology 1997, 126, 97–137.

597

(11) Fletcher, D.; Haynes, B.; Christo, F.; Joseph, S. Applied mathematical modelling 2000,

598

24, 165–182.

599

(12) Xiong, Q.; Aramideh, S.; Kong, S.-C. Energy & Fuels 2013, 27, 5948–5956.

600

(13) Xiong, Q.; Kong, S.-C.; Passalacqua, A. Chemical Engineering Science 2013, 99, 305–

601

602

313. (14) Xue, Q.; Heindel, T. J.; Fox, R. O. Chemical Engineering Science 2011, 66, 2440–2452. 26

ACS Paragon Plus Environment

Page 26 of 46

Page 27 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

603

(15) Xue, Q.; Dalluge, D.; Heindel, T.; Fox, R.; Brown, R. Fuel 2012, 97, 757–769.

604

(16) Mellin, P.; Kantarelis, E.; Yang, W. Fuel 2014, 117, 704–715.

605

(17) Oevermann, M.; Gerber, S.; Behrendt, F. Particuology 2009, 7, 307–316.

606

(18) Oevermann, M.; Gerber, S.; Behrendt, F. fuel 2008, 5, 0.

607

(19) Papadikis, K.; Gu, S.; Bridgwater, A. Chemical Engineering Science 2009, 64, 1036–

608

609

610

1045. (20) Papadikis, K.; Gu, S.; Bridgwater, A. V. Chemical Engineering Journal 2009, 149, 417–427.

611

(21) Ku, X.; Li, T.; Løvås, T. Chemical Engineering Science 2015, 122, 270–283.

612

(22) Løvås, T.; Houshfar, E.; Bugge, M.; Skreiberg, Ø. Energy & fuels 2013, 27, 6979–6991.

613

(23) Pepiot-Desjardins, P.; Pitsch, H. Combustion and Flame 2008, 154, 67–81.

614

(24) Corbetta, M.; Frassoldati, H., A.and Bennadji; Smith, M. J., K.and Serapiglia; Gau-

615

616

617

618

619

620

621

622

623

thier, G.; Melkior, T.; Ranzi, E.; Fisher, E. M. Energy & Fuels 2014, 28, 3884–3898. (25) Ranzi, E.; Cuoci, A.; Faravelli, T.; Frassoldati, A.; Migliavacca, G.; Pierucci, S.; Sommariva, S. Energy & Fuels 2008, 22, 4292–4300. (26) Calonaci, M.; Grana, R.; Barker Hemings, E.; Bozzano, G.; Dente, M.; Ranzi, E. Energy & Fuels 2010, 24, 5727–5734. (27) Blanquart, G.; Pepiot-Desjardins, P.; Pitsch, H. Combustion and Flame 2009, 156, 588–607. (28) Narayanaswamy, K.; Blanquart, G.; Pitsch, H. Combustion and Flame 2010, 157, 1879–1898.

27

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

624

625

(29) Narayanaswamy, K.; Pepiot, P.; Pitsch, H. Jet fuels and Fischer-tropsch fuels: surrogate definition and chemical kinetic modeling. US National Combustion Meeting. 2013.

626

(30) Narayanaswamy, K.; Pepiot, P.; Pitsch, H. Combustion and Flame 2014, 161, 866–884.

627

(31) Narayanaswamy, K.; Pitsch, H.; Pepiot, P. Combustion and Flame 2015, 162, 1193–

628

1213.

629

(32) Narayanaswamy, K.; Pitsch, H.; Pepiot, P. Combust. Flame 2016, 165, 288–309.

630

(33) Shin, E.-J.; Nimlos, M. R.; Evans, R. J. Fuel 2001, 80, 1697–1709.

631

(34) Gunn, D. International Journal of Heat and Mass Transfer 1978, 21, 467–476.

632

(35) Pepiot, P. Automatic strategies to model transportation fuel surrogates. Ph.D. thesis,

633

Department of Mechanical Engineering, Stanford University, USA, 2008.

634

(36) Liang, Y.; Pope, S. B.; Pepiot, P. Combust. Flame 2015, 162, 3236–3253.

635

(37) Mehta, M.; Fox, R. O.; Pepiot, P. Industrial & Engineering Chemistry Research 2015,

636

54, 5407–5415.

637

(38) Niemeyer, K. E.; Sung, C.-J.; Raju, M. P. Combustion and flame 2010, 157, 1760–1770.

638

(39) Pepiot-Desjardins, P.; Pitsch, H. Combustion Theory and Modelling 2008, 12, 1089–

639

640

641

642

643

1108. (40) Desjardins, O.; Blanquart, G.; Balarac, G.; Pitsch, H. Journal of Computational Physics 2008, 227, 7125–7159. (41) Chen, L.; Dupont, C.; Salvador, S.; Grateau, M.; Boissonnet, G.; Schweich, D. Fuel 2013, 106, 61–66.

644

(42) Capecelatro, J.; Desjardins, O. Journal of Computational Physics 2013, 238, 1–31.

645

(43) Desjardins, O.; Pitsch, H. Number paper ICLASS06-084, Kyoto, Japan 2006, 28

ACS Paragon Plus Environment

Page 28 of 46

Page 29 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

646

647

Energy & Fuels

(44) Wang, L.; Pitsch, H. Prediction of pollutant emissions from industrial furnaces using large eddy simulation. 5th US Combustion Meeting, San Diego, CA. 2007.

648

(45) Knudsen, E.; Pitsch, H. Combustion and Flame 2008, 154, 740–760.

649

(46) Desjardins, O.; Moureau, V.; Pitsch, H. Journal of Computational Physics 2008, 227,

650

8395–8416.

651

(47) Desjardins, O.; Pitsch, H. Journal of computational physics 2009, 228, 1658–1677.

652

(48) Knudsen, E.; Pitsch, H. Combustion and flame 2009, 156, 678–696.

653

(49) Pepiot, P.; Jarvis, M.; Nimlos, M.; Blanquart, G. Chemical kinetic modeling of tar

654

655

656

657

658

659

660

formation during biomass gasification; 2010. (50) Van Poppel, B. P.; Desjardins, O.; Daily, J. W. Journal of Computational Physics 2010, 229, 7977–7996. (51) Anca-Couce, A.; Sommersacher, P.; Scharler, R. Journal of Analytical and Applied Pyrolysis 2017, (52) Chen, L. Fast pyrolysis of millimetric wood particles between 800o C and 1000o C. Ph.D. thesis, Lyon I, 2009.

661

(53) Park, W. C.; Atreya, A.; Baum, H. R. Combustion and Flame 2010, 157, 481–494.

662

(54) Jia, L.; Le-Brech, Y.; Shrestha, B.; Frowein, M. B.-v.; Ehlert, S.; Mauviel, G.; Zimmer-

663

mann, R.; Dufour, A. Energy & Fuels 2015, 29, 7364–7374.

29

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Table 1: Parameters used to estimate relevant biomass particle heating rates for the chemistry reduction procedure.

Parameter Value Gas temperature 1073 K - 1273 K Particle temperature during devolatilization 773 K - 873 K Pressure 1 atm Solid heat capacity 2300 J/kg.K Biomass bulk density 650 kg/m3 Biomass particle diameters 300 µm - 1 mm

30

ACS Paragon Plus Environment

Page 30 of 46

Page 31 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

Table 2: PaSR simulation parameters Parameter Value Number of notional gas particles in PaSR 96 Gas residence time 3s Mixing time 0.3 s Biomass particle heating rate 1000 K/s Temperature of raw biomass 300 K Temperature of pure nitrogen stream 1073 K to 1273 K Normalized mass flow rates of nitrogen stream 0.9 Normalized mass flow rates of biomass stream 0.1

31

ACS Paragon Plus Environment

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Table 3: List of species and group of species, and their acronyms. Acronyms A1 AO C2 HAA HMF HPV MPV LPV LVG PAH RAD

Name Single ring aromatics (e.g. Benzene) Oxygenated aromatics (e.g. Phenol) Gases containing two carbon atoms (e.g. Ethylene) Hydroxyacetaldehyde Hydroxymethyl Furfural Heavy weight Primary vapors (7+ carbon atoms, e.g. p-Coumaryl) Medium weight Primary vapors (4-6 carbon atoms, e.g. Levoglucosan) Light weight Primary vapors (2-3 carbon atoms, e.g. Glyoxal) Levoglucosan Polyaromatic hydrocarbons (e.g. Naphthalene) Small radicals pool (e.g. H, OH)

32

ACS Paragon Plus Environment

Page 32 of 46

Page 33 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

Table 4: Parameters for the drop tube reactor simulation. Parameter Value Domain length (Lx × Ly × Lz ) 0.9 m × 0.02 m × 0.02 m Number of cells (nx × ny × nz ) 900 × 1 × 1 Inlet nitrogen velocity 0.279 m/s Temperature of inlet nitrogen stream 1073 K (S1) and 1223 K (S2) Injection rate of biomass particles 7.545×10−7 Kg/s Biomass density 710 Kg/m3 Biomass particle size 520 µm Biomass composition (wt%) Cellulose Hemicellulose C-rich lignin H-rich lignin O-rich lignin Ash Moisture

33

ACS Paragon Plus Environment

43.91 23.85 3.24 14.99 6.71 0.4 6.9

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Table 5: Parameters for the FBR simulation. Parameter Domain length (Lx × Ly × Lz ) Number of cells (nx × ny × nz ) Inlet nitrogen velocity Inlet nitrogen temperature Number of sand particles Size of sand particles Density of sand particles Injection rate of biomass particles Size of biomass particles Density of biomass particles

Value 0.15 m × 0.02 m × 0.0015 m 300 × 40 × 3 0.2 m/s (6umf ) 1073 K 105 200 µm 2650 Kg/m3 5×10−6 Kg/s 200 µm 907 Kg/m3

Biomass composition (wt%) CELL 47.24 HCELL 31.49 LIGC 2.78 LIGH 6.48 LIGO 4.63 Ash 0.37 Moisture 7.0

34

ACS Paragon Plus Environment

Page 34 of 46

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

Mole fraction

Page 35 of 46

1

1

1

0.75

0.75

0.75

0.5

0.5

0.5

0.25

0.25

0.25

LVG

HAA

HMFU

0

0 0

0.2

0.4

0.6

0.8

0 0

0.2

0.4

0.6

0.8

0

0.2

0.4

0.6

0.8

Time [s]

Fig. 1: Pyrolytic decomposition of levoglucosan (LVG), hydroxyacetaldehyde (HAA), and 5-hydroxymethyl furfural (HMF): Comparison between simulation results using the reference chemical model (lines) and experiments (Shin et al., 33 symbols). Different symbols indicate different temperatures (square: 898 K, circle: 923 K, triangle: 948 K, and diamond: 973 K).

35

ACS Paragon Plus Environment

Energy & Fuels

1

0.12

0.1 CO2

CO

CH4

0.75

0.075

0.09

0.5

0.05

0.06

0.25

0.025

0.03

0

0 0

1

2

3

4

5

0 0

6

0.015

Mass fraction

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 36 of 46

1

2

3

4

5

6

0.2

0

1

2

3

4

5

6

2

3

4

5

6

2

3

4

5

6

0.1

H2

H2 O

C 2 H4

0.15

0.075

0.1

0.05

0.05

0.025

0.01

0.005

0

0

0 0

1

2

3

4

5

0

6

0.08

1

2

3

4

5

0

6

0.008

0.02

CH3 CHO

Acetone

Benzene

0.06

0.006

0.015

0.04

0.004

0.01

0.02

0.002

0.005

0 0

1

2

3

4

5

6

1

0

0 0

1

2

3

4

5

6

0

1

Time [s]

Fig. 2: Pyrolytic decomposition of major cellulose devolatilization products: Comparison between simulation results using the reference chemical model (lines) and experiments (Norinaga et al., 6 symbols). Different symbols indicate different temperatures (square: 973 K, circle: 1023 K, and triangle: 1073 K).

36

ACS Paragon Plus Environment

Page 37 of 46

10

4

o

Tp = 500 C o

Tp = 600 C

Heating rate [K/s]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

Particle diameter 300 µm 10

Start of volatilization o Tp = 500 C

3

End of volatilization o Tp = 600 C

Particle diameter 1 mm

2

10800

900

850

o

950

Reactor temperature [ C]

Fig. 3: Typical heating rates experienced by the biomass during gasification at the conditions described in Table 1.

37

ACS Paragon Plus Environment

Energy & Fuels

100 Using DRGEP

Error in H2 O

Error in CO

10−2

(a)

10−4 10−6 10−8

102

(b)

10−2 10−4 10−6 10−8

0

25

50

75

100 125 150 175 200

0

25

50

75

nS 10

10

0

25

50

10−8

100 125 150 175 200

nS 10

(f)

10−2

103 10−4

102 10−6

101

10−8 100 125 150 175 200

75

4

0

nR

Error in AO

10−6

75

10−6

(e)

10−4

50

10−4

10−8

100 125 150 175 200

(d) 10−2

25

10

−2

nS

0

0

(c)

100

Error in MPV

100

Error in A1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 38 of 46

0

25

50

75

nS

100 125 150 175 200

nS

0

100

200

300

400

nS

Fig. 4: Error in the PaSR predictions as a function of the number of species nS retained in the skeletal model during the reduction process: (a) CO, (b) H2 O, (c) MPV, (d) A1 , (e) AO, and (f) number of reactions nR retained in the model. Filled symbols: PaSR nitrogen stream temperature is 1073K; open symbols: PaSR nitrogen stream temperature is 1273K. Circles: automatic reduction; diamonds: semiautomatic reduction with quasi-steady state assumption.

38

ACS Paragon Plus Environment

Page 39 of 46

0.012

0.06

6

×10−3

H2 O

0.01

0.05

CH4

5 0.008

CO

0.04

4 0.006

0.03 0.02

0.002

CO2 0.01 1050

4

1100

1150

1200

1250

1300

2

0 1050

×10−3

C2

3

MPV

0.004

Mass fraction

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

1100

1150

1200

1250

1300

×10−4

H2

1 1050

1100

1150

1200

1250

1300

−9

2.5

×10

6 2

3

RAD

HPV 1.5

4 A1

2

1

PAH

2

1

0.5

AO

LPV 0 1050

1100

1150

1200

1250

1300

0 1050

1100

1150

1200

1250

1300

0 1050

1100

1150

1200

1250

1300

Temperature [K]

Fig. 5: Statistically steady state mass fractions of various light gases, tar species, devolatilization products, and small radical pool: Comparison between the reduced model (symbols) and the reference chemical model (lines) in a Partially Stirred Reactor (PaSR) configuration for the temperature ranging from 1073 K to 1273 K. Expanded species names are provided in section 4.1.

39

ACS Paragon Plus Environment

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Mole fraction

Energy & Fuels

Page 40 of 46

1

1

1

0.75

0.75

0.75

0.5

0.5

0.5

0.25

0.25

0.25

LVG

HAA

HMFU

0

0 0

0.2

0.4

0.6

0.8

0 0

0.2

0.4

0.6

0.8

0

0.2

0.4

0.6

0.8

Time [s]

Fig. 6: Pyrolytic decomposition of Levoglucosan (LVG), Hydroxyacetaldehyde (HAA), and Hydroxymethyl Furfural (HMFU): Comparison between simulation results using the reduced kinetic model (lines) and experiments (symbols, Shin et al. 33 ). Different symbols indicate different temperatures (square: 898 K, circle: 923 K, triangle: 948 K, and diamond: 973 K).

40

ACS Paragon Plus Environment

Page 41 of 46

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 ACS Paragon Plus Environment

Energy & Fuels

0.25

0.075

0.025 unmodified

(a)

0.02

COH2

CO2

unmodified

(c)

unmodified

(b)

0.2

char+trapped-gases

0.05

0.15

0.015 CH3 OH 0.01

Normalized yield

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 42 of 46

0.1

CH4

0.025 0.005 CO 0

H2

0 0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0.04

CO2

0.2

0.4

0.6

0.8

1

COH2

modified

(f)

0.2

char+trapped-gases

0.15

0.03

0.015

0

modified

(e)

modified

(d)

0.02

0 0.25

0.05

0.025

char

0.05

C 2 H4

CH3 OH 0.01

0.1

0.02

0.005

CH4

0.01

C2 H4

CO 0

H2

0 0

0.2

0.4

char

0.05

0.6

0.8

1

0

0.2

0.4

0.6

0.8

1

0

0

0.2

0.4

0.6

0.8

1

time [s]

Fig. 8: Evolution of trapped species, char, and solid residue during biomass devolatilization in DTR simulation S1 using the unmodified reference devolatilization model described in section 2.1 (first row) and the modified model described in section 4.3 (second row). (b) and (c) shows that for the original devolatilization model some trapped species are not released from the biomass, while those trapped species are released after the modification as shown in (e) and (f). Yields are normalized with the initial mass of biomass particle.

42

ACS Paragon Plus Environment

Page 43 of 46

600

Particle diameter [µm]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Energy & Fuels

T = 1073K 500

uncorrected

400

corrected

300 200

0

0.2

0.4

0.6

0.8

1

Reactor length [m]

Fig. 9: Biomass diameter (for simulation S1) at various reactor lengths: Comparison between experimental measurements (symbols), and simulation predictions with corrected heat transfer rate (solid line) and uncorrected heat transfer rate (dashed line).

43

ACS Paragon Plus Environment

0.8

0.1 CO

Mass fraction

0.6 0.4 CO2

0.2

0.08

T = 1073K C2 H4

0.06

CH4

0.04

C6 H6

0.02

H2 0 0.2

0.4

0.6

0.8

0 0.2

1

Reactor length [m]

0.4

0.6

0.8

1

0.8

Mass fraction

CO

0.6 0.4 CO2

0.2

0.08

T = 1223K CH4

0.06

C2 H4

0.04

C6 H6

0.02

H2 0 0.2

0.4

0.6

0.8

Reactor length [m]

T = 1073K 500 400 300 200

0

1

0 0.2

0.4

0.6

0.8

0.2

0.4

0.6

0.8

1

Reactor length [m]

0.1 T = 1223K

600

Reactor length [m]

1

Reactor length [m]

Particle diameter [µm]

Mass fraction

T = 1073K

Mass fraction

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 44 of 46

Particle diameter [µm]

Energy & Fuels

600 T = 1223K 500 400 300 200

0

0.2

0.4

0.6

0.8

1

Reactor length [m]

Fig. 10: Steady state mass fraction (dry basis) of various gas species and particle diameter at different reactor lengths: Comparison between simulation results (lines) and experimental measurements (symbols) for particle diameter, dp =520 µm, and two gas temperatures: 1073 K and 1223 K.

44

ACS Paragon Plus Environment

Page 45 of 46

Energy & Fuels

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 ACS Paragon Plus Environment

Energy & Fuels

0.5 0.2L

0.4L

0.7L

L

0.4 0.3

Mass fraction (dry basis)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 46 of 46

0.2 0.1 0 H2

CO

CO2

CH4

C2

0.02 0.2L

0.015

0.4L

0.7L

L

×10

0.01

×10

0.005 0 A1

AO

PAH

Species

Fig. 12: Steady state mass fraction of major gas and tar species at different reactor lengths in the pseudo two-dimensional FBR simulation. Mass fractions of AO and PAH are multiplied by 10 to compare their behavior with A1 .

46

ACS Paragon Plus Environment