rights / license: research collection in copyright - …30596/... · diss. eth no. 17741 solar...

Research Collection

Doctoral Thesis

Solar Gasification of Carbonaceous MaterialsReactor Design, Modeling, and Experimentation

Author(s): Z'Graggen, Andreas

Publication Date: 2008

Permanent Link: https://doi.org/10.3929/ethz-a-005633392

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For moreinformation please consult the Terms of use.

ETH Library

https://doi.org/10.3929/ethz-a-005633392

http://rightsstatements.org/page/InC-NC/1.0/

https://www.research-collection.ethz.ch

https://www.research-collection.ethz.ch/terms-of-use

Diss. ETH No. 17741

SOLAR GASIFICATION OFCARBONACEOUS MATERIALS

—REACTOR DESIGN, MODELING

AND EXPERIMENTATION

A dissertation submitted toETH ZURICH

for the degree ofDoctor of Sciences

presented byANDREAS Z’GRAGGENDipl. Masch.-Ing. ETH

born July 25, 1978citizen of Schattdorf (UR)

accepted on the recommendation ofProf. Dr. Aldo Steinfeld, examiner

Prof. Dr. Michael F. Modest, co-examiner

2008

Abstract

Solar steam-gasification of carbonaceous materials is proposed as an intermediate

step on the path toward a sustainable energy economy. The developed reactor

technology, together with a set of experimental results, is presented in the first part

of this thesis. Numerical models, described in the second part, were used to support

the engineering design in the first place and to gain further insight in the heat and

mass transfer processes affecting the reactor’s performance.

The developed reactor features a continuous vortex flow of steam laden with

feedstock particles confined to a cavity-receiver and directly exposed to concen-

trated solar radiation. This setup provides efficient radiative heat transfer to the

reaction site to drive the high-temperature highly endothermic process. A 5 kW pro-

totype reactor was tested in a high-flux solar furnace with three distinct feedstocks:

(1) Dry coke powder with steam fed separately yielded up to 87% coke conversion

in a single pass of 1 s residence time at temperatures in the range 1500–1800 K.

The solar-to-chemical energy conversion efficiency was 9%. (2) Coke-water slurry

injected continuously resulted in coke conversion of up to 87% and energy conver-

sion efficiency of up to 5%. Finally, (3) liquefied vacuum residue with steam fed

coaxially yielded maximal coke conversion of 50%, and energy conversion efficiencies

of 2%. The inferior efficiencies of the latter two feedstocks were either due to excess

water used to produce the slurry (2), or to particle deposition due to non-optimal

vacuum residue injection (3). Typical syngas composition produced was 60% H2,

26% CO, 12% CO2, and 2% CH4 for delayed coke and 71% H2, 16% CO, 9% CO2

and 4% CH4 for vacuum residue. The results indicate the technical feasibility of

simultaneous pyrolysis and steam-gasification of delayed coke particles in the range

2–200 µm and of liquefied vacuum residue using concentrated solar energy.

A lumped-parameters steady-state process model that couples radiative heat

transfer with the reaction kinetics is used to support the engineering design. In

a subsequent more detailed simulation, the reactor is modeled by means of a two-

ii Abstract

phase formulation that couples radiative, convective, and conductive heat transfer

to the chemical kinetics for polydisperse suspensions of reacting particles. The gov-

erning mass and energy conservation equations are solved by applying advanced

Monte-Carlo and finite-volume techniques. Validation is accomplished by compar-

ing the numerically calculated temperatures, product compositions, and chemical

conversions with the experimentally measured values obtained from testing a 5 kW

prototype reactor.

The validated reactor model is then used to optimize the reactor’s geometrical

configuration and operational parameters (feedstock’s initial particle size, feeding

rates, and solar power input) for maximum reaction extent and solar-to-chemical

energy conversion efficiency of a 5 kW prototype reactor and its scale-up to 300 kW.

The advantageous volume-to-surface ratio of the 300 kW scaled-up reactor and its

enhanced insulation lead to a solar-to-chemical energy conversion efficiency of 24%.

Moreover, the increased cavity dimensions permit the use of a coarser feedstock to

be reacted efficiently. Finally, the tube-shaped cavity was found to outperform the

commonly employed barrel-shaped cavity, mainly because most of the particles are

directly exposed to the incoming high-flux solar radiation.

Riassunto

Il processo di gassificazione solare di materiali ricchi in carbonio, usando vapore

acqueo, e proposto come passo intermedio sulla strada verso una economia basata

sull’uso sostenibile di energia. Nella prima parte di questa tesi viene descritto il

reattore prototipo e la rispettiva tecnologia sviluppata, seguiti da risultati speri-

mentali ottenuti per diversi edotti. La seconda parte della tesi e incentrata sullo

sviluppo di modelli numerici, i quali da un lato sono stati utilizzati come base per la

progettazione ingegneristica, dall’altro hanno permesso una migliore comprensione

dei processi di scambio di calore e massa che influenzano le prestazioni del reattore.

Nel reattore presentato le particelle di materia prima sono sospese in un flusso

continuo di vapore acqueo a forma di vortice, confinate in una cavita e direttamente

esposte a radiazione solare concentrata. Questa configurazione permette il trasfer-

imento efficiente di radiazione calorica agli edotti coinvolti in una reazione chimica

altamente endotermica che procede ad alte temperature. Un prototipo del reattore,

operante a 5 kW, e stato testato in un forno solare a flusso elevato con tre edotti

distinti. (1) Polvere asciutta di coke con vapore acqueo immesso separatamente e

risultata in conversioni di coke fino a 87% con un tempo di reazione medio di 1 sec-

ondo e temperature tra i 1500 e 1800 K. L’efficienza della conversione energetica da

solare a chimica ha raggiunto un massimo di 9%. (2) Un miscuglio di coke e acqua,

chiamato slurry, iniettato a getto continuo e risultato in conversioni di coke fino a

87% ed efficienze fino a 5%. Infine, (3) il residuo di vuoto (vacuum residue) liquido

con vapore acqueo iniettato coassialmente e risultato in conversioni fino a 50% ed

efficienze fino a 2%. L’efficienza piu debole degli ultimi due edotti usati e dovuta

all’eccesso di acqua necessario per la produzione dello slurry (2), o al deposito di

particelle per l’iniezione non ottimale del residuo di vuoto (3). La composizione tipi-

ca del gas di sintesi (syngas) prodotto e stato di 60% H2, 26% CO, 12% CO2, e 2%

CH4 per le particelle di coke e di 71% H2, 16% CO, 9% CO2 e 4% CH4 per il residuo

di vuoto. I risultati indicano la fattibilita tecnica della pirolisi e gassificazione a

iv Riassunto

vapore acqueo simultanea di particelle di coke con diametri tra i 2 e i 200 µm e di

residui di vuoto liquidi, utilizzando energia solare concentrata.

Per appoggiare la progettazione ingegneristica e stato utilizzato un modello

basato su parametri medi che considera lo stato stazionario del sistema e che collega

i processi di trasmissione della radiazione calorica alla cinetica della reazione chim-

ica. In una successiva simulazione piu dettagliata, il reattore e stato modellato per

mezzo di un flusso a due fasi che considera la trasmissione di calore per radiazione,

convezione, conduzione e la cinetica chimica per una sospensione di particelle di varie

dimensioni. Le equazioni di conservazione di massa ed energia sono state risolte me-

diante l’applicazione di tecniche Monte-Carlo e mediante il metodo dei volumi finiti.

La validazione e stata eseguita confrontando risultati calcolati numericamente per le

temperature, la composizione del gas prodotto e la conversione chimica con i valori

ottenuti in via sperimentale dai test con il reattore prototipo.

Il modello validato e poi stato utilizzato per ottimizzare le proporzioni geomet-

riche del reattore e i parametri operativi (dimensione delle particelle, rata d’immissione

degli edotti, energia solare incidente) per l’ottenimento di una conversione chimica

oppure di un’efficienza energetica massima. Sono stati considerati due reattori: da

un lato il prototipo testato a 5 kW e dall’altro un reattore simile progettato per un

input solare di 300 kW. Il rapporto vantaggioso volume a superficie del reattore a

300 kW e l’isolazione termica migliore sono risultati in un’efficienza energetica di

24%. Inoltre, l’aumento delle dimensioni della cavita consentono l’impiego di par-

ticelle di dimensioni piu grandi. Infine, una cavita a forma di tubo e risultata piu

adatta delle cavita a forma di barile comunemente impiegate, soprattutto perche la

maggior parte delle particelle e in tal modo direttamente esposta a radiazione solare

altamente concentrata.

Acknowledgments

First of all, I thank Prof. Dr. Aldo Steinfeld for supervising my doctoral studies at

the Professorship in Renewable Energy Carriers, ETH Zurich. The excellent research

environment he provided, his confidence in my independent way of working as well

as his lasting support were essential to my work.

I am very grateful to Prof. Dr. Michael F. Modest for critically reviewing this

manuscript and for acting as co-examiner.

The members of the Synpet project team deserve special acknowledgment. I

thank Dominic Trommer and Philipp Haueter for giving me insight in the fields of

chemical engineering and reactor design, respectively. I will keep in good memory

the weeks we spent struggling with our experimental setup. I am also indebted to

the technical staff at PSI for their support, they made the experimental campaigns

possible in the first place.

I thank all my colleagues and members of our research group. In particular,

my thanks go to Dr. Wojciech Lipinski and Dr. Jorg Petrasch for many fruitful

discussions on radiative transfer and numerical simulations. Hansmartin Friess

also deserves special thanks for his patient support in the development of the

gas-temperature measurement apparatus. I thank Dr. Thomas Osinga, Dr. Elena

Galvez, Dr. Peter von Zedtwitz, Tom Melchior, Gilles Maag, Viktoria von Zedtwitz-

Nikulshyna, Patrick Coray, Sophia Haussener, Nic Piatkowski and Lothar Schunk

for providing an inspiring working environment.

Last, but no least, my very special thanks go to those close to my heart. My

parents Erwin and Susanna for igniting my curiosity and for giving me the belief

that everything is possible. Fabia for the daily royal treatment.

vi Acknowledgments

Contents

Abstract i

Riassunto iii

Acknowledgments v

Nomenclature xi

1 Introduction 1

1.1 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2 Conventional Gasification . . . . . . . . . . . . . . . . . . . . . . . . 4

1.3 Solar Thermochemistry and Solar Gasification in Particular . . . . . . 6

I Experimental Work 9

2 Feedstock 11

2.1 Petrozuata Delayed Coke . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.1.1 Characterization of Coke-Water Slurries . . . . . . . . . . . . 16

2.2 Petrozuata Vacuum Residue . . . . . . . . . . . . . . . . . . . . . . . 18

3 Experimental Setup 19

3.1 Reactor Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

3.2 Feedstock Injection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3.2.1 Dry Powder Feeding . . . . . . . . . . . . . . . . . . . . . . . 22

3.2.2 Slurry Feeding . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

3.2.3 Liquefied Vacuum Residue . . . . . . . . . . . . . . . . . . . . 24

3.3 Peripherals and Facility . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.4 Key Process Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 27

viii Contents

3.4.1 Product Gas Composition . . . . . . . . . . . . . . . . . . . . 27

3.4.2 Chemical Conversion . . . . . . . . . . . . . . . . . . . . . . . 28

3.4.3 Energy Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . 29

3.4.4 Residence Time . . . . . . . . . . . . . . . . . . . . . . . . . . 30

4 Results 33

4.1 Coke — Dry Powder Feeding . . . . . . . . . . . . . . . . . . . . . . 33

4.2 Coke — Slurry Feeding . . . . . . . . . . . . . . . . . . . . . . . . . . 37

4.3 Vacuum Residue — Liquefied Injection . . . . . . . . . . . . . . . . . 42

4.4 Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

II Modeling 49

5 Simulation Framework 51

5.1 General Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . 51

5.1.1 Incoming Radiation . . . . . . . . . . . . . . . . . . . . . . . . 51

5.1.2 Reactor Walls . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

5.1.3 Quartz Window . . . . . . . . . . . . . . . . . . . . . . . . . . 56

5.2 Polydisperse Coke Particles . . . . . . . . . . . . . . . . . . . . . . . 58

5.2.1 Radiative Properties . . . . . . . . . . . . . . . . . . . . . . . 60

5.3 Chemical Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

6 System Analysis — Lumped Parameters Model 67

6.1 Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

6.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

7 Heat and Mass Transfer in the Reactor Cavity 77

7.1 Heat Transfer Modes in the Polydisperse Particle Suspension . . . . . 78

7.2 Governing Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

7.3 Numerical Implementation . . . . . . . . . . . . . . . . . . . . . . . . 82

7.3.1 Monte-Carlo Ray Tracer . . . . . . . . . . . . . . . . . . . . . 84

7.3.2 Fluid Flow Solver . . . . . . . . . . . . . . . . . . . . . . . . . 90

7.3.3 CFD-MC Coupling . . . . . . . . . . . . . . . . . . . . . . . . 91

7.4 Results and Validation . . . . . . . . . . . . . . . . . . . . . . . . . . 95

Contents ix

8 Optimization and Scale-up 105

8.1 Simulation Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

8.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

9 Conclusions 115

9.1 Experimental Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

9.2 Heat and Mass Transfer Modeling . . . . . . . . . . . . . . . . . . . . 116

9.3 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

III Appendices 119

A Gas Temperature Measurement in Radiating Environments 121

A.1 Additional Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . 121

A.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

A.3 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

A.4 Heat Transfer Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

A.5 Calibration Measurements . . . . . . . . . . . . . . . . . . . . . . . . 126

A.6 Furnace Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 128

A.7 CFD Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

A.8 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 133

B Measurement Accuracy of Qsolar 135

C Radiation Absorption in the Gas Phase 137

Bibliography 149

x Contents

Nomenclature

Acronyms

BC Boundary Condition

CCS CO2 Capture and Sequestration

CTL Coal to Liquids

CPC Compound Parabolic Concentrator

CCR Conradson Carbon Residue

ETH Eidgenossische technische Hochshule

IGCC Integrated Gasification Combined Cycle

FV Finite Volume

GCI Grid Convergence Index

LHV Lover Heating Value

PDF Probability Density Function

PSI Paul Scherrer Institut

RMS Root Mean Square

SEM Scanning Electron Microscope

VR Vacuum Residue

Latin Characters

A window absorptance

Ci concentration of species i

cp specific heat capacity, J/(kg ·K)

D particle diameter, m

Dc diffusion coefficient, m2/s

E emitted/absorbed power in the MC solver

EA apparent activation energy

xii Nomenclature

Eλb Planck’s blackbody spectral emissive power, W/(m2 · µm)

f particle size distribution function

fV solid volume fraction

h enthalpy, J/kg

h convective heat transfer coefficient, W/(m2 ·K)

I radiation intensity, W/(m2 · sr)i chemical species

K kinetic rate constant

k thermal conductivity, W/(m2K)

k imaginary part of the refractive index

k0 kinetic frequency factor

ln liters under normal conditions: 273 K and 1 bar

M molar mass, kg/mol

m mass, kg

mi mass flow rate of species i, g/min

Np number of particles

n number of moles

n real part of the refractive index

ni molar flow rate of species i, mol/min

p pressure, Pa

Qa,s,ext absorption, scattering and extinction efficiencies for a single sphere

Q heat rate, W

Qsolar incident solar radiation, W

q heat flux, W/m2

qsolar incident solar radiative flux, W/m2

R universal gas constant, J/(mol ·K)

Ri volumetric reaction rate of species i, kg/(m3 · s)Rw window reflectance

r radial coordinate, m

ri mass specific reaction rate of species i, mol/(g · s)S cross sectional area, m2

S surface area, m2

Si sensitivity

s distance along path

T temperature, K

Nomenclature xiii

Trw window transmittance

t time, s

U overall heat transfer coefficient, W/(m2 ·K)

u velocity, m/s

V volume, m3

V volumetric flow rate, ln/min

XC carbon conversion

XH2O steam conversion

x axial coordinate, m

Yi mass fraction of species i

yi molar fraction of species i

Greek Characters

β extinction coefficient, 1/m

χ fraction

∆HR reaction enthalpy, J/mol

ε emissivity

ε relative error

η particle effectiveness

ηchem solar-to-chemical energy conversion efficiency

ηproc solar thermal process efficiency

κ absorption coefficient, 1/m

µ dynamic viscosity, kg/(m · s)ρ density, kg/m3

Φ scattering phase function

φ energy source term, W/(m3 · s)σs scattering coefficient, 1/m

σ Stefan-Boltzmann constant 5.67051 · 10−8, W/(m2 ·K4)

τ relaxation time, s

τ mean residence time, s

ξ size parameter

xiv Nomenclature

Subscripts/Superscripts

0 initial

a absorbed

b blackbody

chem used by the chemical reaction

cond conduction

conv convection

g gaseous phase

i species

in inlet

rad radiative

rerad reradiation from the cavity

s solid phase

sk shrunk

T total

w window

λ spectral

Chapter 1

Introduction

Today’s energy sector is facing the important challenge of meeting the rising demand

for energy without increasing the emission of greenhouse gases and, in particular,

those of CO2. This increasing demand, combined with the predicted decline in

conventional oil production in 10-20 years [78], can be met by increased utilization

of coal, whose reserves are estimated to last for 250 years [82]. With the increase in oil

prices also the cost- and energy-intensive extraction of oil shales and tar sands [85], as

well as inter-fuel substitution (coal-to-liquids and coal-to-gas technology, CTL) [47,

58] have gained in importance. Unfortunately, these technologies do not contribute

to the reduction of CO2 emission. Clean coal technologies, including CO2 capture

and sequestration (CCS) [49] and increased plant efficiency, i.e., the use of coal-fired

combined cycle processes (IGCC and PFBC) [75], are proposed as countermeasures

to mitigate the ecological drawbacks of a coal-based energy economy [15].

A higher potential in meeting the mentioned challenges altogether is attributed

to the substitution of fossil fuels by clean renewable fuels on one side [46, 22] and

the enhancement of material and energy efficiency on the other [45]. Typical sub-

stite fuels are, among others, fuels produced from biomass (biofuels) [35, 51] and

fuels produced by means of solar energy (solar fuels), typically hydrogen [93]. The

complete substitution is, if ever achievable, a long term goal. Nonetheless, mid term

goals aimed at the development of hybrid solar/fossil technologies are of strategic

importance to the creation of a transition path toward solar fuels.

An important example of such a hybrid solar/fossil process, in which fossil fuels

are used exclusively as the chemical source for H2 production and concentrated solar

power is used exclusively as the energy source of process heat, is the endothermic

2 1. Introduction

steam-gasification of petroleum coke (petcoke) to synthesis gas (syngas). Petcoke is

a solid residue from the processing of heavy and extra heavy oils using delay-coking

and flexicoking technologies. The calorific value of the feedstock is upgraded by

solar power in an amount equal to the enthalpy change of the reaction. A Second-

Law analysis for generating electricity using solar gasification products indicates the

potential of doubling the specific electrical output and, consequently, halving the

specific CO2 emission, vis-a-vis conventional petcoke-fired power plants [112, 101].

The sun is an attractive source for primary energy. In fact, the solar energy re-

serve is virtually unlimited, freely available, and its utilization is ecologically benign.

On the other hand, the incident solar radiation on the earth surface is diluted (≈ 1

kW/m2), intermittent and unequally distributed. These intrinsic drawbacks of solar

energy, which affect in particular solar electricity generation, are overcome by solar

gasification and by solar thermochemistry, in general. Solar gasification is therefore

an attractive way to transform solar energy into a storable and transportable energy

carrier.

1.1 Thesis Outline

The present thesis is performed in the framework of a joint project between the

Swiss Federal Institute of Technology in Zurich (ETH) and the Paul Scherrer In-

stitut in Villigen (PSI), both Switzerland, the national research center ‘Centro de

Investigaciones Energeticas, Medioambientales y Tecnologicas’ in Spain (CIEMAT),

and the research and development center of Petroleos de Venezuela, S.A. (PDVSA /

INTEVEP). The main goal of the project is the development and demonstration of

a process and of the related technology required for the production of high quality

syngas from carbonaceous materials using a solar thermochemical process.

A previous PhD thesis was done by D. Trommer in the same framework [100].

His work focused on the thermodynamics of petroleum coke gasification, and on the

investigation of the chemical kinetic mechanisms involved. The work presented here

covers three additional tasks of the project: (1) the conception, design, manufactur-

ing and experimental operation of a lab-scale prototype reactor at a power level of 5

kW, (2) the numerical modeling of the process and of heat and mass transfer inside

the reactor, and (3) the design of a scaled-up version of the reactor to a power level

of 500 kW.

This thesis is divided into two main parts. Part I is dedicated to the experimen-

1.1. Thesis Outline 3

tal work (Chapters 2-4). The engineering design, the operation of such a reactor on

a solar furnace, the types and morphology of the employed feedstock and the ex-

perimental results are presented. Part II focuses on the modeling aspects (Chapters

5-8). Developed models were used to support the engineering design, to identify

parameters affecting the efficiency of the concept and to predict the performance of

the scaled-up reactor.

The characteristics and origin of the employed feedstock types are described

in Chapter 2. Elemental composition and particle size distributions are given for

Petrozuata Delayed Coke. Production and rheology of coke water slurry is shortly

addressed. Finally, the properties of vacuum residue are presented.

Chapter 3 focuses on the experimental setup used. The design of the cavity-

type reactor, the instrumentation and the outline of the concentrating facility are

described in detail. Furthermore, the key performance parameters of the process

are defined and their derivation and accuracy is discussed. These parameters are

the product gas composition, the energy efficiency, the chemical conversion and the

average reactant’s residence time in the cavity.

In Chapter 4 the results of three experimental campaigns are reported, each

one for a distinct feedstock: (1) Dry coke powder with steam fed separately (further

referred to as Campaign 1), (2) coke-water slurry injected into the cavity (Campaign

2) and (3) liquefied vacuum residue with steam fed separately (Campaign 3). Syngas

was successfully produced in all three campaigns. Differences in reactor performance

related to the feedstock type and injection system are addressed. Material from this

chapter has been published in [123, 122] and [121].

General consideration on the framework that supports the simulation models are

given in Chapter 5. The characteristics of the incoming radiation as well as the

conductive behavior of the reactor walls was determined by stand-alone preliminary

simulations. Wall and window radiative properties are presented. The concept of

‘equivalent monodisperse diameter’ is introduced to model the polydisperse feedstock

types, which have random size-distribution functions. Finally, the kinetic rate laws

for steam gasification of petcoke developed by Trommer [102] and employed in this

thesis are shortly presented.

In Chapter 6 a lumped parameters model — used for system analysis — is

described. It is used to support the early stages of the engineering design. The model

is validated with the experimental results. Good agreement is found especially for

the temperatures and carbon conversion rates of Campaign 1. Parameter studies

4 1. Introduction

are performed and the beneficial aspect of up-scaling are investigated.

In Chapter 7 a model for heat and mass transfer in the reactor’s cavity is pre-

sented. It solves simultaneously for convective, conductive and radiative heat trans-

fer as well as for mass transport and chemical kinetics. Focus is put on the radiative

heat transfer in participating media. The numerical implementation, including the

coupling of a simple finite-volume flow solver with a Monte-Carlo raytracer that

solves the equation of radiative transfer is described in detail. The model is suc-

cessfully validated with good agreement for the two campaigns considered (1 and

2 with petcoke as feedstock). Material from this chapter has been submitted for

publication [125].

Finally, in Chapter 8 the model developed and validated in Chapter 7 is used to

study in detail the parameters that affect the reactor’s performance. These are the

morphology of the feedstock, the feeding rates and the geometry of the cavity. The

same analysis is also performed for the scaled-up reactor. Better performance due

to lower heat losses as well as increased flexibility towards feedstock morphology are

predicted. Material from this chapter has been submitted for publication [126].

Appendix A is dedicated to an apparatus developed to measure gas temperatures

in highly radiating environments. It is based on a simple suction thermocouple

supplemented by a correction model that accounts for radiative, conductive and

convective heat transfer. This chapter has been published as [120].

1.2 Conventional Gasification

Gasification is a process that, under addition of a gasification agent, usually CO

or steam, converts any carbon-containing material, such as coal, petroleum, coke,

biomass or waste into a synthesis gas (syngas) composed primarily of carbon monox-

ide and hydrogen. This endothermic reaction takes place at temperatures above

approx. 1000 K. Since the 1850s gasification technologies have been applied com-

mercially. In between 1850 and 1940 gasification of coal was used to produce ‘town

gas’ for light and heat. Virtually all gas for fuel and light was produced by gasifi-

cation until the development of natural gas supplies and transmission lines in the

1940s and 1950s. During World War II German engineers used gasification to pro-

duce synthetic fuel from coal. This technology was exported to South Africa in

the 1950s, where it was further developed to produce liquid fuels and chemicals.

More recently, gasification has gained in importance as an environmentally-friendly

1.2. Conventional Gasification 5

disposal and conversion technology of residues from heavy crude oil processing. Con-

ventional gasification is also used in integrated gasification combined cycle (IGCC)

plants fired with low or negative-valued feedstocks for clean and efficient electricity

production [58, 79].

In conventional gasification the energy required to heat the reactants and to

run the chemical reaction is either produced by partial oxidation of the feedstock

(autothermal gasifiers) or provided externally (allothermal gasifiers). Autothermal

gasification has the advantage that any losses associated with the heat transfer

are avoided and the construction of the gasifier is simplified. Furthermore, the

temperature can be adjusted quickly and accurately by regulation of the amount

of oxygen injected. On the other hand, allothermic gasification usually results in

higher-quality syngas, because the products are not contaminated by combustion

byproducts. Gasification reactors are divided into three groups depending on how

the solid fuel is brought into contact with the gasification agent: fixed/moving bed,

fluidized bed or entrained flow. The type of reactor influences the residence time, re-

actor temperature and pressure, and certain characteristics of the produced gas. For

this reason each reactor type is suited for a specific type, rank, and size distribution

of solid feedstock.

Fixed/moving bed gasifiers contain a bed of lump fuel supported by a grate and

maintained at a constant height. The feedstock is fed from the top end and flows

countercurrent to the rising gas stream. A single particle moving through the bed

passes different zones including drying and preheating, devolatilization, gasification,

oxidation, and ash removal. Fixed bed gasification systems are simple, reliable and

offer high efficiency with respect to feedstock and energy consumption. Outlet gas

temperature are in the range 425–650 C, feedstock size is between 6 and 50 mm. As

a consequence of the moderate temperatures the product gas contains around 10%

hydrocarbons, mostly CH4. A typical example of a commercial fixed bed gasifier is

the dry ash Lurgi gasifier [76, 37]. Lurgi gasifiers were employed on a large scale for

South Africa’s SASOL complex [21].

Fluidized bed gasifiers accept feedstock as grains with a size in the range 6 to

10 mm. The solid feedstock is suspended on upward-blowing jets of the gasification

agent. The result is turbulent mixing of gas and solids. The tumbling action, much

like a bubbling fluid, provides more effective chemical reactions and heat transfer.

Reactors of this type are characterized by outlet gas temperatures in the range 900 to

1050 C, a high specific gasification rate and product uniformity. A typical example

6 1. Introduction

of the fluidized bed technology is the Winkler process [10, 37].

Entrained-flow gasifiers are operated with pulverized feedstock with particles of

less than 100 µm. The solid feedstock is entrained with the gasifying agent to react

in a concurrent flow having the form of a high temperature flame. The principal

advantages of this process are the ability to handle practically any coal as feedstock

and to produce a clean, tar-free and low-in-methane synthesis gas. Entrained-flow

reactors require relatively high flow rates leading to small residence times. In order

to obtain full conversion of the coke, entrained flow reactors operate at very high

temperatures (above 1400 C) and require, therefore, a comparably high oxidant

supply. Devolatilization products are released in the high temperature region and

thus further cracked and oxidized. This reactor type is usually operated at 20-70

bar [37, 58]. This technology has bee selected for the majority of commercial IGCC

plants. The two best-known types of entrained-flow gasifiers are the top-fired coal-

water-slurry feed gasifier, as used in the Texaco process [87] and the dry coal feed

side-fired gasifier as developed by Shell and Krupp-Koppers (Prenflo)[109, 37].

1.3 Solar Thermochemistry and Solar Gasifica-

tion in Particular

Solar thermochemistry refers to a number of process technologies that harness con-

centrated solar energy by absorbing sunlight in an endothermal chemical reaction

occurring at high temperatures. The products of these chemical reactions are usually

called solar fuels or solar energy carriers, as they carry solar energy in an amount

equal to the reaction’s enthalpy change. Overviews on the different technologies en-

visaged and on processes that were experimentally demonstrated are given in [52, 28]

and [94]. These processes can be divided into three groups based on the employed

feedstock and on the chemical reactions involved: (1) direct thermal splitting of

water or H2S, a byproduct from natural gas, petroleum or coal refining, (2) H2O

splitting thermochemical cycles and (3) upgrading and decarbonization of fossil fuels.

Despite being a simple and intuitive way to produce hydrogen, the direct thermal

splitting of water is difficult to achieve in practice. In fact, the temperatures required

for an efficient dissociation lie above 2500 K. Furthermore, the explosive mixture

of gaseous products (H2 and O2) needs to be separated at high temperatures to

avoid recombination. The feasibility of the process has been investigated among

1.3. Solar Thermochemistry and Solar Gasification in Particular 7

others by [27] and [54]. An alternative way to produce hydrogen by direct thermal

splitting is the decomposition of H2S, a highly toxic industrial byproduct recovered

in the sweetening of natural gas and in the removal of organically-bound sulfur

from petroleum and coal. The process runs at temperatures around 1800 K on a

Al2O3 surface. Separation of the products H2 and S2 occurs by condensation and

subsequent solidification of the sulfur [48].

In order to reduce the reaction temperature and eliminate the need for high-

temperature gas separation, various processes for the production of hydrogen by

multi-step thermochemical water splitting cycles have been proposed — typically

two-step processes based on metallic redox systems. In the first step the metal oxide

is partly or completely reduced in a solar high-temperature endothermic reactor; in

the second step the product from the first step is oxidized with water in an exother-

mal reaction releasing hydrogen (hydrolysis reaction) closing the cycle. An overview

of the most considered redox pairs is given in the following. The Fe3O4/FeO cycle

was recently experimentally investigated by [16]. Complete thermal dissociation of

Fe3O4 at 1973 K and 0.1 bar under inert atmosphere was reported, while up to

80% conversion at 850 K was obtained for the hydrolysis step. Systems based on

Mn3O4/MnO, Co3O4/CoO can be thermally decomposed in air at 1810 and 1175 K,

respectively. Unfortunately, the H2 yields obtained were only 0.002% and 4 · 10−7%

at 900 K for Mn3O4/MnO and Co3O4/CoO, respectively. High H2 yield of 99.7% at

900 K was obtained for Nb2O5/NbO2, but the thermal decomposition temperature

of 3600 K in air is prohibitively high [62]. Extensive research has been performed on

the ZnO/Zn cycle. Latest results for direct thermal dissociation of ZnO are reported

by [89]. The in-situ formation and hydrolysis of Zn nanoparticles was investigated

by [23]. The issues arising from the high temperatures required for the direct ther-

mal dissociation of ZnO (approx. 2000 K at 1 bar) are mitigated by the addition

of carbon to the feedstock at the expense of CO2 emissions. Solar carbo-thermal

reduction of ZnO was successfully demonstrated at a power level of 10 kW [70] and

at 300 kW [115]. The produced zinc was reported to be significantly more reac-

tive in the hydrolysis step than commercially available material [107]. Finally, a

three-step sulfur-iodine cycle is proposed by [81], in which sulphuric acid is split by

concentrated solar radiation.

The processes involved in upgrading and decarbonization of fossil fuels can be

grouped into three categories: (1) solar thermal cracking of hydrocarbons, in which

solid carbon and hydrogen is produced, (2) solar reforming, in which a gaseous

8 1. Introduction

hydrocarbon is reacted to a H2-CO mixture (syngas) in a steam or CO2 atmosphere,

and (3) solar steam- or CO2-gasification of solid carbonaceous materials [111]. High

temperature solar pyrolysis of coal and biomass was already reported in 1983 by [7]

and [4], respectively. The solar pyrolysis of oil shale and the recovery of the volatile

product was reported by [8]. Lately, research focused on thermal splitting of methane

into hydrogen and carbon black. Successful lab-scale experiments were reported by

[41], [18] and [1]. The solar reforming of methane was reported among others by [19]

without catalyst, by [53] on a catalytic Ru/Al2O3 surface, and by [13] on a catalytic

rhodium surface.

Successful solar gasification of carbonaceous materials was first reported in the

80’s: coal, activated carbon, coke, and coal/biomass mixtures were employed in a

fixed bed windowed reactor by [31]. Charcoal, wood and paper was gasified with

steam in a fixed bed reactor, and charcoal was gasified with CO2 in a fluidized

bed by [95]. More recently, [26] reported on gasification of oil shale and coal in a

fixed bed reactor, [65] on gasification of waste tyres and plastic (PET), and [112]

on gasification of coal in a fluidized bed reactor. Lately, extensive studies on the

steam gasification of petroleum coke and vacuum residue were performed. Chemical

kinetics were measured by [101] and three distinct feedstocks were processed in a

lab-scale entrained-flow reactor: dry coke particles [123], coke-water slurries [122]

and vacuum residue [121].

Part I

Experimental Work

Chapter 2

Feedstock

The materials used as feedstock for the solar gasification experiments presented in

this thesis are two typical residuals of heavy crude oil refining: a vacuum residue

(VR) and a delayed petroleum coke (coke). These are produced from the processing

of extra heavy crude oil from the Petrozuata oil field in the Orinoco belt (southern

strip of the eastern Orinoco river basin in Venezuela) and were provided by PDVSA,

the Venezuelan oil company.

gasnaphthagasolinekerosenedieselheavy gas oilat

mosp

eric

dis

tillation

vac

uum

dis

tillat

ion

crude oil

furnace

desalter

vacuum gas oil

lubricating oilfurnace

atmosphericresidue

coke

dru

m

coke

dru

m

furnace

frac

tinat

or

gas/gasoline

gas oil

residue recycle

vacuumresidue (VR)

coke

delayed cokingdistillation(fractionation)

Figure 2.1: Schematic representation of a refinery. Only units involved in the con-version of crude oil to vacuum residue and delayed coke are shown.

12 2. Feedstock

Figure 2.1 shows the main components of a refining facility that are involved in

the production of either VR or coke. Further processing of volatile hydrocarbons is

omitted, whereas the focus is put on the treatment of the residues of the distillation

and coking processes.

Crude oil is usually contaminated with water, inorganic salts, suspended solids,

and water-soluble trace metals. In order to reduce corrosion, plugging, and fouling

of equipment, and to prevent poisoning of the catalysts it is, therefore, desalted

(cleaned) prior to further processing. Crude oil refining is a complex process, which

involves numerous specific secondary processes. Nevertheless, it can be split into

two major steps, especially with respect to the production of vacuum residue and

coke.

Distillation Desalted crude oil is separated into various fractions by distillation

(fractionation), first in an atmospheric column and subsequently in a vacuum column

operated at 30–50 mbar. The temperatures required for distillation of the heavier

fraction of crude oil at ambient pressure are so high that cracking would occur.

These materials are, therefore, distilled under vacuum, since a lowering of the pres-

sure results in lowering of their respective boiling temperatures. Partial pressures

are further lowered by steam injection. The main fractions obtained have specific

boiling-point ranges and can be classified in order of decreasing volatility into gases,

light distillates, middle distillates, gas oils, and vacuum residuum (VR). In contrast

to the distillation step, where thermal cracking is avoided, further processing of the

VR aims at the destruction of the long hydrocarbon chains of the heavy molecules.

Coking Heavy residuals are upgraded into lighter products or distillates by a

severe method of thermal cracking (coking). Coking produces straight-run gasoline

(coker naphtha) and various middle-distillate fractions used as catalytic cracking

feedstock. The final residue is a form of carbon called coke. The two most common

processes are delayed coking and continuous (contact or fluid) coking.

In delayed coking the charge, typically atmospheric or vacuum residues, is heated

above the coking point (480–520 C) and subsequently fed to a large coke drum,

which provides the long residence time (24 hours) needed to allow the cracking

reactions to proceed to completion. Premature coking in the heater tubes can be

avoided by sufficiently high flow rates. Vapors from the drums are returned to

a fractionator where gas, naphtha, and gas oils are separated out. The heavier

2.1. Petrozuata Delayed Coke 13

hydrocarbons produced in the fractionator are recycled through the furnace. In

order to maintain continuous operation usually two coke drums are used, one on

stream and the other being emptied. The full drum is steamed to strip out uncracked

hydrocarbons, cooled by water injection, and decoked by mechanical or hydraulic

methods.

2.1 Petrozuata Delayed Coke

Table 2.1: Coke feedstock types used in the experiments.

type shipment Dmin–Dmax1, µm D30

2, µm

1 1 jet milled 0.51–17.4 2.212 1 ball milled 0.76–300 6.73 2 sieved, 80 µm screen 1.5–300 17.64 2 sieved, 200 µm screen 2.3–592 30.85 2 as received 3–678 40.01Values strongly affected by the detection limit of the Horiba LA950.2See eq. (5.10) for the definition of the mean diameter D30.

The raw coke provided by PDVSA was further processed to the five distinct types

of feedstock listed in Table 2.1. A first shipment consisted of relatively big particles

with characteristic diameters of a few centimeters. Feedstock types 1 and 2 were

obtained by grinding the raw material with a jet mill and a ball mill, respectively

(ARP GmbH, Loeben, Austria). A second shipment was composed by somewhat

smaller particles with a maximal diameter around 1 mm. This material was sieved

with a 80 µm and a 200 µm screen to produce feedstock types 3 and 4, respectively.

Type 5 describes the raw material as received.

Figure 2.2 shows the particle size distribution functions measured by laser scat-

tering with a Horiba LA950. Plotted are the number density f (D) and the respective

volume density f (D) ·D3 for the five types of feedstock listed in Table 2.1. Although

the relative number of big particles seems negligible from Fig. 2.2 (a), those parti-

cles carry an important share of the total mass as shown by the volume distribution

curves shifted to the right in Fig. 2.2 (b). SEM pictures of feedstock type 1 are

shown in Fig. 2.3. Frames (a) and (b) show the unreacted spherically-shaped par-

ticles with diameters in the micrometer range. Frames (c) and (d) show particles

after pyrolysis above 1400 K, in which release of volatile matter and thermal cracking

14 2. Feedstock

1 2 5 10 20

num

ber

den

sity

f(D

)

particle diameter D, µm1 2 10 100 500

volu

me

den

sity

f(D

)·D

3

particle diameter D, µm

1

2

3 45

(a)

12 3 4 5

(b)

Figure 2.2: Particle size distribution functions of the petcoke feedstock types listedin Table 2.1. (a) shows the number density (also called population density); (b)shows the volume density. Type 1: jet milled, type 2: ball milled, type 3: sievedwith a 80 µm screen, type 4: sieved with a 200 µm screen, type 5: as provided byPDVSA.

result in the formation of small structures on the particles surface. Finally, frames

(e) and (f) show particles after a typical experimental run from the first campaign

with combined pyrolysis and gasification and carbon conversion of XC=0.75. As

expected, the size of these particles is smaller compared to Fig. 2.3 (a) and (b).

Furthermore, and in contrast to the purely pyrolyzed feedstock, no formation of

small structures on the surface is observed. The elemental composition of the PD

coke is shown in Table 2.2. It mainly consists of carbon (62% molar) and hydrogen

(35% molar) with minor shares of nitrogen, oxygen and sulfur together with nickel,

vanadium and sodium in the ppm range. The lower heating value (LHV) was calcu-

lated as the sum of the elemental LHVs, since the exact composition of the bound

hydrocarbons is not known. The obtained value of 35.8 MJ/Kg is slightly higher

than the LHVs reported for coal [79] (28.5–35.3 MJ/Kg).


(a) (b)

1¹m 1¹m

(e) (f)

1¹m 1¹m

(c) (d)

1¹m 1¹m

Figure 2.3: SEM micrographs of petcoke samples used in Campaign 1 (feedstock type1). (a) and (b) show unreacted particles, (c) and (d) show particles after pyrolysisabove 1400 K, (e) and (f) show particles after a typical experimental run withcombined pyrolysis and gasification and carbon conversion XC=0.75. Magnification:(a),(c) and (e) 4’000 x; (b),(d) and (f) 10’000 x

16 2. Feedstock

Table 2.2: Approximate main elemental chemical composition (ultimate analysis),low heating value and molar ratios H/C and O/C for PD coke.

wt% mol%

carbon 88.21 62.06hydrogen 4.14 34.71nitrogen 2.28 1.37oxygen 1.46 0.77sulfur 4.16 1.10

nickel ppm 414vanadium ppm 2207sodium ppm 100

LHV kJ/kg 35876H/C mol/mol 0.5581O/C mol/mol 0.0124

2.1.1 Characterization of Coke-Water Slurries3

In the second experimental campaign presented in Section 4.2 coke particles-water

mixtures, so called slurries, were employed. Coal water slurries (CWS) are well

known as an alternative to petroleum fuels used in furnaces and kilns for building

material industry and in industrial and heating boilers [90]. Typical coal-water

slurries consist of 60–75% coal and 25–40% water, with particle sizes in the range

of 50–500 µm. Low fractions of water are desirable in order to keep the reduction

in the fuel’s LHV small. Low viscosity is ensured by the addition of typically 1% of

chemical additives [20]. Sedimentation stability, rheology and the effect of particle

size on the slurry properties were investigated among others by [105] and [97].

The negative effect of excess water is less important for the steam gasification

process analyzed in this thesis. In fact a minimum of 57% wt water is required to

obtain a stoichiometric H2O/C mixture of the reactants (additional water beyond

57% wt in turn reduces the efficiency). Two types of slurry where analyzed: sample

1 was made of coke particles with an average diameter of 2.21 µm compared to

17.6 µm for sample 2 (see Table 2.1). The water-to-carbon molar ratio was 2/1

and 1/1, corresponding to mass ratios of 73%/27% and 57%/43% for sample 1 and

3Material presented in this section has been developed in the framework of: F. Kritter. Injectionsystem of coal water slurries for solar reactors. Diploma thesis, ETH Zurich, 2005.


0 500 1000 15000

100

200

300

400

500

600

700

shear rate γ, 1/s

dyn

amic

visc

osity

µ,m

·Pa/

s(a)

sample 1sample 2

0 500 1000 15000

20

40

60

80

100

120

140

shear rate γ, 1/ssh

ear

stre

ssτ,Pa

(b)

Figure 2.4: Viscosity (a) and shear stress (b) as a function of the shear rate of awater-coke slurry, measured for two types of feedstock [57]. The arrows indicatethe upward and downward curves, obtained by increasing and subsequently decreas-ing the shear rate, respectively. Nicely visible is the hysteresis loop, typical forthixotropic fluids.

sample 2, respectively. The slurries were prepared without any chemical additives

by mechanical mixing for several minutes. Viscosity and shear stress were measured

by shear rheometry at CIEMAT in Madrid. Results are plotted in Fig. 2.4. Curves

for dynamic viscosity and shear stress as a function of the shear rate, first increased

from 0 to 1500 1/s and subsequently decreased to 0 1/s, are show in Fig. 2.4 (a)

and (b) respectively. The shape of the curves indicates non-Newtonian thixotropic

behavior of the slurries. This means that the slurries undergo a decrease in viscosity

for increased shear rates. For example, for a shear rate of 250 1/s the viscosity

was 270 and 98.5 m · Pa/s, whereas at a shear rate of 1000 1/s it was 106 and 40.5

m · Pa/s for sample 1 and sample 2, respectively. As expected, the smaller particles

form a more viscous slurry, even if the water-to-coke ratio is doubled. Uniform

slurries were easily produced with the bigger particles but a the cost of storage

instability. In fact, separation of the two phases was observed after a short period

of time, especially for sample 2. During the experimental runs in the solar furnace

the slurry tank was, therefore, constantly stirred.

18 2. Feedstock

2.2 Petrozuata Vacuum Residue

The vacuum residue delivered by PDVSA is a residue from processing of extra-heavy

crude oil from the Petrozuata fields in Venezuela. It is solid at ambient temperature,

has plasto-elastic behavior for moderate shear rates but breaks into little pieces at

high shear rates, for example when hit with a hammer, showing a brittle fracture

surface. The VR liquifies at around 120 C, and strong release of volatile matter is

observed above 200 C (visual observations). Coking, that is the transformation to

a solid by thermal cracking, is reported to take place approx. above 400 C and has

to be avoided by any means in the feeding system. The elemental composition of

vacuum residue is given in Table 2.3 (data from PDVSA/INTEVEP). Significantly

larger amounts of hydrogen are detected compared to PD coke (57% vs. 37 % molar),

thus resulting in an increased LHV (as for coke, the LHV was calculated as the sum

of the elemental LHVs). On the other hand, smaller amounts of nitrogen, oxygen,

sulfur and metals are detected.

Table 2.3: Approximate main elemental chemical composition (ultimate analysis),low heating value and molar ratios H/C and O/C for vacuum residue.

wt% mol%

carbon 86.19 42.52hydrogen 9.6 56.84nitrogen 0.85 0.36oxygen 0.16 0.059sulfur 3.1 0.57

nickel ppm 185vanadium ppm 774

LHV kJ/kg 40908H/C mol/mol 1.337O/C mol/mol 0.0014

Chapter 3

Experimental Setup

An experimental reactor was developed to demonstrate the feasibility of solar steam

gasification of extra heavy crude oil or derived residues such as vacuum residue or

coke. The reactor was successfully tested at PSI’s solar furnace for 3 distinct exper-

imental campaigns underlining the flexibility of the implemented design and serving

as valuable support to the process of up-scaling. Feedstock was injected as coke

powder (Campaign 1), coke-water slurry (Campaign 2) or liquefied vacuum residue

(Campaign 3). This chapter describes the general experimental setup, including the

reactor design, details about the injection systems and a description of the facility

were the experiments were performed.

3.1 Reactor Design

The engineering design of the prototype reactor was based on a set of lab-scale

preliminary studies. Semi-batch mode experiments (a batch of feedstock flushed

with a continues gas stream) were performed in a conventional electrical furnace [2]

and [92], with a simple quartz dome reactor [25] and with an opaque ceramic tube

[3] mounted in the focus of ETH’s solar simulator [42]. Evaluated reactor concepts

included also direct/indirect concentrated solar irradiation, packed/fluidized beds

and semibatch/continous feeding. The highest potential was identified for a cavity

type reactor, where the constantly fed particulate feedstock is directly irradiated

while it is entrained by the reactive flow.

The final reactor configuration — developed in a joint effort at ETH in coop-

eration with engineers from the project partners — is schematically shown in Fig.

20 3. Experimental Setup

tc K

tc K

tc K

tc K

tc K

tc K tc

S

tc K

pro

duct

outlet

multifunct

ional

inle

tpurg

ing

noz

zles

win

dow

reac

tor

liner

(Al 2O

3)

reac

tor

shel

l(I

nco

nel

)

reac

tor

insu

lati

on(A

l 2O

3 /

ZrO

2)

aper

ture

tange

nti

al inje

ctio

nnoz

zles

oil co

oled

fru

stru

m

wat

er c

oole

d w

indow

fitti

ng

ring

gap

tc S

Figure 3.1: Scheme of the solar chemical reactor configuration used for the steamgasification of carbonaceous materials at a power level of 5kW. The location of thethermocouples (tc) of types K and S are indicated by stars.

3.1. It consists of a cylindrical cavity-receiver, 210 mm in length, 120 mm inside

diameter, that contains a 5 cm-diameter opening — the aperture — to let in con-

centrated solar power. The cavity-type geometry is designed to effectively capture

3.1. Reactor Design 21

the incident solar radiation; its apparent absorptance is estimated to exceed 0.95

[68]. The cavity is made out of Inconel 601, lined with Al2O3, and insulated with

an Al2O3/ZrO2 ceramic foam. The aperture is closed by a 0.3 cm thick, clear fused

quartz window, mounted in a water-cooled aluminum ring that also serves as a shield

for spilled radiation. The window is actively cooled and kept clear from particles

and/or condensable gases by means of an aerodynamic protection curtain created

by tangential flow through four tangential nozzles combined to radial flow through a

circular gap. In front of the aperture, the cavity-receiver is equipped with a diverg-

ing conical funnel for mounting the window 6 cm in front of the focal plane, where

the radiation intensity is about 10 times smaller and dust deposition is unlikely to

occur. Since radiation spillage can reach flux concentration ratios greater than 1000,

corresponding to an incident flux of 1000 suns (1 sun = 1 kW/m2), this component is

actively oil-cooled and kept in the range 393–453 K to prevent steam condensation.

Steam and particles are injected separately into the reactor cavity, permitting

separate control of mass flow rates and stoichiometry. Steam is introduced through

several ports. Based on flow visualization experiments in Plexiglas models and CFD

simulations, best flow patterns in terms of residence time and flow stability were

obtained with two sets of 4 symmetrically distributed tangential nozzles, located in

planes 4 and 12 cm behind the aperture plane, as shown in Figure 3.1. The petcoke

feeding unit is positioned on the top of the reactor vessel with its inlet port located

at the same plane as the primary steam injection system, allowing for the immediate

entrainment of particles by the steam flow. Inside the cavity, the gas-particle stream

forms a vortex flow that progresses toward the rear along a helical path. Reactor

wall temperatures were measured in 12 locations with type K thermocouples, in-

serted in the Inconel walls and not exposed to direct irradiation (indicated by the

stars in Fig. 3.1). The temperature of the inner Al2O3 cavity was measured with a

solar-blind pyrometer that is not affected by the reflected solar irradiation because

it measures in a narrow wavelength interval around 1.39 µm where solar irradia-

tion is mostly absorbed by the atmosphere [103]. The nominal reactor temperature,

denoted Tcavity, was then calculated as the mean between the pyrometer reading (cor-

rected for window transmittance, Tpyrometer = Tpyrometer,reading · Tr−0.25w , where Trw is

the transmittance) and the thermocouple reading (corrected for conduction through

the Al2O3 liner with eq. (6.9)). The exit gas temperature was measured with a

type S thermocouple in the outlet pipe, while the temperature of the inlet port was

monitored with an additional type K thermocouple.


3.2 Feedstock Injection

The reactor was primarily designed for processing of dry petcoke particles in the

10–100 µm range, in which particles and steam are injected separately. The inves-

tigation of alternative feedstocks, such as water-coke slurries and liquefied vacuum

residue, led to substantial changes in the design of the injection systems in order

to accommodate for their specific properties, while still creating a particle-laden

entrained flow.

3.2.1 Dry Powder Feeding

steam

petcoke particles/argon

reactor liner(Al2O3)

reactor shell(Inconel)

reactor insulation(Al2O3 / ZrO2)

tangential injectionnozzles

steam

steam

steam

Figure 3.2: Feedstock injection setup used in Campaign 1. Petcoke particles are fedby means of a brush conveyer and further entrained by the steam injected from fourtangential nozzles.

Figure 3.2 shows the injection system used in the first experimental campaign.

The petcoke particles were stored in a tube and conveyed toward a rotating brush

by a piston. Argon was used to flush this rotating brush to ensure constant feeding

rates. The feeding unit was mounted on top of the reactor permitting the particles

to fall into the cavity. They were then entrained by the steam injected from four

tangential nozzles mounted at the same axial position. Smooth overall operation was

3.2. Feedstock Injection 23

experienced with this setup, and a homogenously distributed particle suspension was

obtained (visual observations), which lead to good reactor performance.

3.2.2 Slurry Feeding

petcoke/water slurry




particles/steaminjection (long)

particles/steaminjection (short)

Figure 3.3: Feedstock injection setup used in Campaign 2. The coke-water slurry ispreheated in the tubes cast into the reactor’s insulation, the water evaporates andthe particles are injected into the cavity.

Figure 3.3 shows the injection system used in the second experimental campaign.

The petcoke particles were mixed with water in a constantly stirred tank mounted

above the reactor. The coke-water slurry was then conveyed with a peristaltic pump

and introduced into the cavity trough an Inconel tube cast into the reactor liner. The

water in the slurry is thereby evaporated, the steam expands and the preheated feed-

stock is injected into to cavity. The longer original design of the injection tube had

to be substituted for a shorter one. In fact, due to the high temperatures achieved,

clogging by coking of the particles to the tube walls occurred. As a consequence of

the thixotropic behavior of the slurry the tubing had to be cleaned and filled with

water prior to each feeding cycle. In addition, clogging was prevented by the use of

bigger particles and excess water, which resulted in a lower slurry viscosity.


3.2.3 Liquefied Vacuum Residue

steam/argonliqui¯ed VR




VR droplets

release ofvolatiles and particle formation

deposition ofbig particles

Figure 3.4: Feedstock injection setup used in Campaign 3. The liquefied vacuumresidue is injected through a cold nozzle, the droplets are subsequently heated, py-rolyzed and gasified.

Figure 3.4 shows the injection system used in the third experimental campaign.

The vacuum residue is liquefied in a heated tank mounted above the reactor and

kept at approx. 423 K. The VR is conveyed by means of a gear pump, while both,

the tubing and the pump head are heated to approximately 393 K. Five different

reactant injection arrangements were tested, including a setup similar to the one used

in the slurry campaign. Clogging by coking was observed for temperatures above

673 K. The injection location was therefore moved out of the directly irradiated

region and further cooled by a coaxial steam/argon flow. In addition, the nozzle

was removed to ease the VR injection because of the insufficient pressure delivered

by the gear pump. After injection the VR droplets fall inside the cavity, are exposed

to the concentrated radiation and thereby release the volatile matter and break up to

form particles. Due to this low-level injection technology only a part of the resulting

particles was entrained by the gas flow, whereas the bigger particles deposited in the

cavity, thus lowering the reactor’s performance.

3.3. Peripherals and Facility 25

3.3 Peripherals and Facility

reactor parabolic concentratorheliostat shutter

Figure 3.5: Photograph of PSI’s high flux solar furnace. Main visible features are thesun tracking heliostat, the Venetian-blind type shutter used to control the incomingradiation and, inside the building, the parabolic concentrator and the reactor.

Experimentation was carried out at the PSI’s solar furnace [34]. This solar re-

search facility, shown in Fig. 3.5, consists of a 120 m2 sun-tracking heliostat in-axis

with an 8.5 m-diameter paraboloidal concentrator, and delivers up to 40 kW at peak

concentration ratios exceeding 5000 suns. A Venetian blind-type shutter located be-

tween the heliostat and the concentrator controls the power input to the reactor.

Radiative solar flux intensities were measured optically with a calibrated CCD cam-

era by recording the image of the sun on a water-cooled Al2O3-coated Lambertian

(diffusely reflecting) plate positioned at the focal plane. The reactor itself was sub-

sequently positioned with its aperture at the focal plane and intercepting the regions

of maximum solar flux intensity. Integration of the incident radiative flux over the

reactor’s aperture yielded the solar power input, Qsolar . The accuracy of the opti-

cal measurement combined with the reactor’s misalignment led to an error of ±913%

for Qsolar (for details see Appendix B). Figure 3.6 shows an overview of the most

important peripherals attached to the reactor. Inlet gas flows of Ar and H2O were

controlled using electronic flow meters (Bronkhorst HI-TEC). Solid particles were


concentratedsolar radiation

Ar

oil cooling circuit

water cooling circuit

H2O

Bronkhorst steam generator

H2O

Arevaporator

win

dow

purg

e

steam line

cyclone

filter /condenser

IR

GC

Siemenscalomat

SiemensIR

detector

Varianmicro GC

flar

e

solid

resi

due

feedstock/steaminjection

cokeparticles

H2O cokeparticles

liquefiedVR residue

gear pump

mixer

slurry

peristalticpump

brushfeeder

screwconveyor

Campaign 1 (2004) Campaign 2 (2005) Campaign 3 (2006)

optionalsolid

recirculation

pyrometer

productoutlet

Figure 3.6: Main components of the experimental setup at PSI’s high flux solar fur-nace. Three different feeding systems are shown for the three distinct experimentalcampaigns performed from 2004 to 2006.

removed from the outlet stream in a cyclone and excess moisture was condensed in a

counterflow heat exchanger. Product gases were analyzed on-line by gas chromatog-

raphy (Varian Micro GC, equipped with a Molsieve-5 and a Poraplot-U column; 1

ppm detection limit; 0.75 min−1 sampling rate). Better temporal resolution was ob-

tained by additional IR-based detectors for CO, CO2, and CH4 (Siemens Ultramat

23; 0.2% detection limit; 1 s−1 sampling rate), and a thermal conductivity-based

3.4. Key Process Parameters 27

detector for H2 (Siemens Calomat 6; 50 ppm detection limit; 1 s−1 sampling rate

and interfering gas correction). Finally, the product gas was burned on a torch

mounted outside of the building. The reactor’s pressure was monitored with pres-

sure transducers while a pressure safety valve prevented overpressure derived from

a five-fold volumetric growth due to gas formation and thermal expansion. Three

distinct feeding systems were employed for each specific feedstock employed. The

feeding rate of dry coke particles was determined by measuring the weight differ-

ence of the particle conveyor. The rates for slurry and liquefied vacuum residue

were determined by calibration of the feeding pumps before the experiments. An

accuracy of ±5% was estimated. Steam was generated externally in a Bronkhorst

steam generator that allows H2O concentrations in Argon up to 100%. A combined

heater/cooler was used to temperate the oil flowing trough the diverging frustum.

The spilled radiation impinging on the window mounting was removed by water

cooling. Representative solid product samples collected at the filter downstream of

the reactor were examined by scanning electron micrography.

3.4 Key Process Parameters

3.4.1 Product Gas Composition

The measuring equipment used to determine the gas composition at the outlet pro-

vides only relative values, namely the molar fractions yi of species i. The molar

production rate of a specific species ni is then given by:

ni = ntot · yi (3.1)

The total molar flow rate ntot at the exit of the reactor was calculated by means of

the known argon input flow rate nAr:

ntot =nAr

yAr

(3.2)

where yAr is the molar fraction of argon in the product gas estimated with yAr =

1 −∑

yi, with i = H2, CO, CO2, CH4 and O2 under the assumption that argon is

the only non-negligible gas species not detected by the equipment.


A more accurate technique was introduced in Campaigns 2 and 3 to overcome

potential inaccuracies. Nitrogen was injected at a defined rate nN2 into the product

gas stream after the reactor. The total molar flow rate reaching the measurement

equipment is then n′tot = ntot + nN2 , and can be calculated as n′

tot = nN2/yN2 . The

total molar flow rate at the exit of the reactor ntot, required in eq. (3.1), can then

be expressed as:

ntot =nN2

yN2

− nN2 = nN2 ·(

1

yN2

− 1

)(3.3)

where yN2 is directly measured with the Varian Micro GC. For the purpose of cross-

checking a modified version of eq. (3.2) is used in Campaigns 2 and 3, which accounts

for the additional nitrogen flow:

ntot =nAr

yAr

− nH2 (3.4)

For example, comparison of the values for Campaign 2 revealed an average difference

of 6% with peak relative errors up to 22%. Note that the automatic interfering gas

correction of the Calomat does not work for nitrogen. The reading for the hydrogen

was therefore corrected manually by 0.09 % H2 for each % N2.

Other gas species, such as H2S, C2H2, C2H4 and C2H6 were not considered here

since their concentrations are negligible for the calculations of energy efficiencies and

chemical conversions.

3.4.2 Chemical Conversion

The share of carbon in the feedstock consumed by the gasification reaction, further

referred to as carbon conversion, was calculated by means of a C balance:

XC =nCO + nCO2 + nCH4

n0C

(3.5)

where n0C denotes the initial amount of carbon fed into the reactor. Similarly, water

conversion is obtained from the oxygen balance:

XH2O =nCO + 2 · nCO2

n0H2O

(3.6)

where n0H2O denotes the molar amount of water fed by the steam generator and

water contained in the coke-water slurry fed in Campaign 2. Note that the excess


steam, condensed right at the reactor outlet, can then be estimated as nH2O =

(1−XH2O) · n0H2O.

3.4.3 Energy Efficiency

The reactor’s thermal performance is described by two energy conversion efficiencies.

The solar-to-chemical energy conversion efficiency ηchem is defined as the portion of

solar energy input stored as chemical energy:

ηchem =XC · n0

C ·∆HR |298K

Qsolar

(3.7)

where XC · n0C describes the amount of carbon gasified and ∆HR |298K= 131 kJ/molC

is the reaction enthalpy. The solar thermal process efficiency ηproc is defined as the

portion of solar energy stored both as LHV in the product gas and sensible heat

— which potentially can be recovered — and takes into account the heat required

for steam generation (evaporation and superheating to 423 K) and the LHV of the

feedstock:

ηproc =nH2LHV H2 + nCOLHV CO +

∑speciesi

∫ Treactor

473knicp,i(T ) dT

Qsolar + Qsteam + mfeedstockLHV feedstock

(3.8)

Theoretical maximal values for ηchem are calculated assuming perfect insulation

(Qcond = 0), complete chemical conversion (XC = 1) and stoichiometric steam-

to-carbon feeding rates. The useful energy available to heat and react the feed-

stock is then given by the solar power input diminished by the reradiation losses:

Qsolar · (1− σT 4

CI), where C is the average concentration at the aperture and I is the

incident solar radiation intensity. The maximal molar feeding rate of carbon, nC,

heated to a given reactor temperature T and completely reacted results then in

nC =Qsolar · (1− σT 4

CI)

∆HR |298K +∫ T

T∞(cp,C + cp,H2O + χ · cp,H2) dT

(3.9)

where χ is the hydrogen-to-carbon ratio 0.5 · H/C of the feedstock given in Tables

2.2 and 2.3 for coke and vacuum residue, respectively.

Table 3.1 lists theoretical maximal solar-to-chemical energy conversion efficien-

cies as a function of the cavity temperature for coke and vacuum residue and for

the 5 kW prototype reactor and its 300 kW scale-up. Typical values lie in the range


Table 3.1: Theoretical maximal solar-to-chemical energy conversion efficiency ηchem

for coke and vacuum residue for the 5 kW prototype reactor and its 300 kW scale-upas a function of the cavity temperature.

Tcavity, K

feedstock Qsolar, kW Daperture, mm 1300 1500 1700

ηchem, %coke 5 50 64 56 48VR 5 50 61 53 45coke 300 500 62 52 41VR 300 500 58 49 38

38–64 %. As expected, higher cavity temperatures lead to lower efficiencies as a con-

sequence of the increased reradiation losses. Values for vacuum residue are slightly

smaller that those for coke since more H2 bound in the feedstock has to be heated

(χVR > χcoke). Finally, the scale-up reactor performs worse than the prototype re-

actor as a consequence of the lower concentration C at the aperture (C5kW ≈ 2500

suns, C300kW ≈ 1500 suns).

3.4.4 Residence Time

The average residence time of the reactants in the reactor, τ , is influenced by two

phenomena: (1) thermal expansion at the entrance of the reactor and (2) increase

of the molar flow due to the chemical reaction. Assuming a plug flow reactor and

first-order reaction the carbon conversion as a function of time t is given by [59]

X (t) = 1 − e−k·t, where k is the rate constant. Imposing the boundary conditions

X (0) = 0 and X (τ) = XC it results:

X (t) = 1− eln(1−XC)

τ·t (3.10)

where XC is the carbon conversion measured at the reactor’s outlet defined in

eq. (3.5). The total molar flux of gases n (t) in the reactor is:

n (t) = nAr + nH2,feedstock + n0H2O + n0

C ·X (t) (3.11)

where nH2,feedstock accounts for the hydrogen bound in the coke released by pyrolysis

at the entrance of the reactor and the last term accounts for gas generated by


the chemical reaction. Assuming ideal gas behaviour the fluid velocity is given by

dx/dt = u (t) = (n (t) RT ) / (pS), where S is the reactor cross-section. Integration

with respect to time and imposing the boundary conditions x (0) = 0 and x (τ) = L,

L being the length of the reactor, leads to:

τ =p · V

R · Tcavity

· 1

nAr + nH2,feedstock + n0H2O + n0

C ·(1 + XC

ln(1−XC)

) (3.12)

where V is the reactor volume, XC is the carbon conversion at the outlet, and all ni

are meant as initial feeding rates.

Chapter 4

Results

4.1 Coke — Dry Powder Feeding1

Operational conditions and a summary of measurements taken under approximate

steady-state conditions for 23 solar experimental runs are listed in Table 4.1. All

runs were performed with PD coke particles of type 1 with 2.21 µm mean initial

diameter. Reactants were continuously fed at a mass flow rate in the range of 2.0–

4.8 g coke/min and 6.0–9.0 g H2O/min, and at ambient temperature and 423 K for

petcoke and steam, respectively. The average residence time of the reactants in the

reactor’s cavity, as calculated from eq. (3.12), varied between 0.7 and 1.3 s. Solar

power input through the aperture was in the range of 3.3–7.0 kW. The nominal

reactor temperatures varied between 1289 K and 1719 K. Chemical conversions for

petcoke and steam after a single pass, as defined by eqs (3.5) and (3.6), reached up

to 87% and 68%, respectively. Temperatures and gas compositions for a represen-

tative solar run (run 7) are shown in Fig. 4.1. In this run, the reactor was first

heated to above 1450 K under an argon flow. Thereafter, reactants were introduced

during an interval of 12 min at a rate of 3.5 g/min of coke and 6 g/min of H2O,

corresponding to a H2O/C molar ratio of 1.3. Average values at approximately

steady state condition are indicated by gray bars: Qsolar=4.3 kW, Treactor = 1421

K, Tcavity=1201 K, nH2=0.2 mol/min, nCO = 0.094 mol/min, nCO2=0.04 mol/min,

and nCH4=0.007 mol/min. The concentration of H2S — a gasification byproduct

1Material from this section has been published in: A. Z’Graggen, P. Haueter, D. Trommer, M.Romero, J. C. de Jesus, and A. Steinfeld. Hydrogen production by steam-gasification of petroleumcoke using concentrated solar power — II. reactor design, testing, and modeling. InternationalJournal of Hydrogen Energy, 31:797–811, 2006.

34 4. Results

0 10 20 300

10

20

30

40

pro

duct

gas

com

pos

itio

n,%

0 10 20 30

3

4

5

1000

1200

1400

1600

time, min

Qsolar,kW

tem

per

ature

,K

Tcavity

Tshell

Qsolar

H2

CH4

COCO2

start ofcoke feeding

Figure 4.1: Temperatures, solar power input and product gas composition for atypical run of Campaign 1 (run # 7 in Table 4.1). The dashed vertical line indicatesthe start of the petcoke feeding, the two vertical dotted lines delimit the intervalconsidered for the steady state calculations.

—, determined in separate measurements, reached up to 0.05% [101]. Its removal

may be accomplished downstream, for example by scrubbing. The syngas quality,

characterized by the molar ratios H2/CO ≈ 2.2 and CO2/CO ≈ 0.42, was in the

range of the values reported for conventional gasification [79].

The solar-to-chemical efficiency, ηchem, defined in eq. (3.7) ranges from 5% to

9%; the process efficiency, ηproc, defined in eq. (3.8) ranges from 22% to 35%. The

complete energy balance for each experimental run, ordered by increasing Tcavity, is

shown in Fig. 4.2. Indicated are the power lost by reflection and attenuation by

the window and reradiation from the cavity Qrerad, power lost by heat conduction

trough the reactor walls Qcond, power used to heat the reactants from 293 K to

Tcavity, Qheating, and change in enthalpy due to the chemical reaction Qchem (details

on the calculation are described in Chapter 6). Also given is the solar power input

Qsolar and its estimated accuracy bounds (see also Appendix B). In the average,

heat losses are principally due to attenuation and emission by the window and re-

radiation through the aperture (≈ 17% ), and conduction through the reactor walls

(≈ 66%). As expected, the radiative losses are strongly temperature dependent. To

some extent they can be minimized by augmenting the input solar power flux, e.g.

4.1. Coke — Dry Powder Feeding 35

1

2

3

4

5

6pow

er,kW

1300

1400

1500

1600

1700

tem

per

ature

,K

experimental runs ordered by increasing Tcavity

Qrerad Qcond Qchem Qheating

Qsolar

Tcavity

Figure 4.2: Breakdown of the total input power into heating of the reactants Qheating,chemical reaction enthalpy Qchem, reradiation losses Qrerad and conduction lossesQcond for the 23 runs of Campaign 1. Also plotted is the measured solar powerinput, Qsolar with its respective accuracy bounds.

by means of secondary optics — typically compound parabolic concentrators (CPC)

[114] — allowing the use of a smaller aperture for capturing the same amount of

energy. Increasing the reactor temperature further results in a higher reaction rate

and degree of chemical conversion, which in turn results in higher energy conversion

efficiencies.

36 4. Results

Table 4.1: Steam and petcoke feeding rates, composition of the product gas, nom-inal temperatures and performance parameters for 23 valid runs of experimentalcampaign 1 conducted with dry petcoke particles.

mco

ke

mH

2O

nA

rn

H2

nC

On

CO

2n

CH

4Q

sola

rT

cavity

Tsh

ell

XC

XH

2O

η chem

η pro

cτ

#g/

min

mol

/min

kWK

K–

––

–s

12.

36.

00.

310.

220.

090.

046

0.00

44.

514

9512

760.

830.

550.

070.

291.

02

2.0

6.0

0.23

0.17

0.07

0.03

50.

005

5.0

1390

1193

0.75

0.42

0.05

0.22

1.2

32.

46.

00.

270.

240.

100.

046

0.00

84.

615

0512

630.

870.

580.

070.

311.

04

2.1

6.0

0.27

0.18

0.08

0.03

80.

008

4.4

1510

1203

0.78

0.46

0.06

0.27

1.0

52.

36.

00.

270.

120.

040.

026

0.01

13.

312

8910

670.

470.

280.

050.

231.

36

2.2

6.0

0.27

0.15

0.06

0.03

10.

007

4.3

1432

1099

0.62

0.38

0.05

0.23

1.1

73.

56.

00.

270.

200.

090.

040

0.00

74.

314

2112

100.

540.

520.

070.

261.

18

2.9

6.0

0.27

0.23

0.11

0.04

10.

006

5.3

1542

1221

0.74

0.58

0.07

0.27

1.0

93.

56.

00.

270.

280.

130.

044

0.00

86.

115

9812

080.

710.

660.

070.

280.

910

3.6

6.0

0.27

0.29

0.14

0.04

30.

004

5.7

1454

1201

0.71

0.68

0.07

0.28

1.0

113.

96.

00.

270.

270.

130.

040

0.00

74.

614

5812

270.

630.

640.

090.

311.

012

4.5

6.0

0.27

0.29

0.15

0.03

80.

004

6.7

1475

1219

0.59

0.68

0.06

0.25

1.0

134.

56.

00.

270.

280.

130.

037

0.00

85.

915

1012

080.

540.

620.

070.

260.

914

3.4

7.0

0.27

0.26

0.12

0.04

70.

007

5.6

1694

1214

0.71

0.56

0.07

0.29

0.8

153.

58.

00.

270.

260.

130.

049

0.00

75.

817

1912

240.

720.

500.

070.

290.

716

3.5

9.0

0.27

0.26

0.12

0.05

30.

008

6.6

1671

1231

0.70

0.45

0.06

0.26

0.7

173.

79.

00.

270.

300.

120.

057

0.01

05.

316

5012

300.

700.

480.

080.

320.

718

3.9

9.0

0.27

0.31

0.14

0.05

60.

012

5.7

1618

1220

0.71

0.50

0.08

0.32

0.7

194.

89.

00.

270.

330.

150.

057

0.01

17.

016

1412

130.

610.

530.

070.

280.

720

3.5

9.0

0.27

0.32

0.13

0.06

40.

009

4.6

1466

1266

0.78

0.51

0.09

0.35

0.8

213.

59.

00.

270.

290.

120.

058

0.01

15.

014

3912

230.

750.

480.

080.

320.

822

3.5

9.0

0.27

0.23

0.09

0.04

70.

014

4.3

1355

1137

0.59

0.37

0.08

0.29

0.9

232.

99.

00.

270.

260.

110.

055

0.00

44.

615

1313

250.

790.

440.

080.

320.

8

4.2. Coke — Slurry Feeding 37

4.2 Coke — Slurry Feeding2

Table 4.2 lists a summary of 29 experimental runs taken under approximate steady

state conditions. All runs were performed with PD coke. Three different particles

sizes were used: 1) ball milled, with an average particle diameter of 8.5 µm 2) sieved,

with particle diameters < 80 µm; and 3) sieved, with particle diameters < 200 µm,

corresponding to feedstock types 2, 3 and 4 of Table 2.1, respectively. The slurry

was prepared by stirring petcoke with demineralized water for a desired stoichio-

metric molar ratio (H2O/C)slurry in the range 1–3. The total water-to-coke molar

ratio (H2O/C)total, which includes the additional steam used to purge the window,

varied in the range 2.1–10.8. The petcoke mass flow rate was between 0.3 and 3.6

g/min. For better control purposes, an Ar flow of 0.13 mol/min was added to the

steam flow protecting the window. The resulting average residence time for the coke

particles was calculated using eq. (3.12), and τ was found to be in the range 0.9–2.3

seconds. Solar power input was varied over the range of 3.2–5.1 kW, resulting in

a nominal reactor temperature in the range of 1392–1566 K. Chemical conversion

for steam and petcoke after a single pass reached up to 28% and 87%, respectively.

The variation of temperatures and gas compositions during a representative solar

experimental run (run 10 of Table 4.2) is shown in Fig. 4.3. In this run, the reac-

tor was first heated to above 1500 K under an Argon flow. Thereafter, reactants

were continuously introduced during an interval of 14 minutes at a mass flow rate

of 2.9 g/min of slurry with (H2O/C)slurry = 2. Additionally, 4 g/min of H2O were

injected to protect the window, corresponding to (H2O/C)total = 5.8. Main product

gases were H2, CO, and CO2. Relatively smaller amounts of CH4 were derived from

the pyrolysis of PD coke [102], believed to occur immediately after exposure to the

high-flux irradiation as a result of slurry heating rates exceeding 500 K/s. Average

values under approximate steady-state conditions, in the interval indicated by ver-

tical dashed bars of Fig. 4.3, were: Treactor=1536 K, nH2=0.05 mol/min, nCO=0.02

mol/min, nCO2=0.01 mol/min, and nCH4=0.001 mol/min. H2S concentration, de-

termined in separate measurements, reached up to 0.05% [101]; its removal may be

accomplished downstream, e.g., by scrubbing. Particle deposition or residual ashes

was not observed inside the reactor cavity. Previous thermogravimetric studies re-

2Material from this section has been published in: A. Z’Graggen, P. Haueter, G. Maag, A.Vidal, M. Romero, and A. Steinfeld. Hydrogen production by steam-gasification of petroleum cokeusing concentrated solar power — III. reactor experimentation with slurry feeding. InternationalJournal of Hydrogen Energy, 32:992–996, 2007.

38 4. Results

0 5 10 150

5

10

15

20

25

pro

duct

gas

com

pos

itio

n,%

0 5 10 15

3

4

5

1000

1200

1400

1600

time, min

Qsolar,kW

tem

per

ature

,K

Tcavity

Tshell

Qsolar

H2

CH4

CO

CO2

start ofslurryfeeding

Figure 4.3: Temperatures, solar power input and product gas composition for atypical run of Campaign 1 (run # 10 in Table 4.2). The dashed vertical line indicatesthe start of the slurry feeding, the two vertical dotted lines delimit the intervalconsidered for the steady state calculations.

vealed a residuum of only 1.1 % weight after completion of the gasification reaction

[102]. The reactor’s thermal performance is described by the efficiencies ηchem and

ηproc defined in eq. (3.7) and eq. (3.8), respectively.

Efficiencies up to 5% and 23% were achieved for ηchem and ηproc, respectively.

An energy balance for all experimental runs is shown in Fig. 4.4. Also plotted is

the value of the solar power input Qsolar. Main power losses are due to re-radiation

Qrerad and conduction Qcond, with average shares of 17% and 78%, respectively.

Qrerad, which takes into account attenuation by the spectrally selective window and

re-radiation through the aperture, was calculated using the radiosity method [124].

The power lost by conduction through the insulation Qcond was calculated assuming

a simplified cylindrical reactor geometry and using material properties provided

by the manufacturers. The power delivered for the chemical reaction Qchem was

calculated from the reaction enthalpy change at 298 K, whereas that stored in the

form of sensible heat of products was calculated from the species enthalpy change

between 473 K and the corresponding reaction temperature. Qsteam accounts for the

evaporation and superheating of H2O from 298 to 423 K (details are given in Chapter

6). Figure 4.5 shows the effect of particle size and slurry stoichiometry on the


1

2

3

4

5

6

pow

er,kW

1300

1400

1500

1600

tem

per

ature

,K



Qsolar Tcavity


0

0.2

0.4

0.6

0.8

1

carb

onco

nve

rsio

nX

C

0

5

10

15

20

25

effici

ency

, %XC

type 3type 4

type

2

3:1 2:1 1.5:1 2:1 1.5:1 1:1

slurry molar ratio H2O/C

ηprocηchem

Figure 4.5: Carbon conversion and efficiencies for the 29 experimental runs of Cam-paign 2 grouped by feedstock type and slurry molar ration H2O/C. Feedstock types2, 3 and 4 correspond to initial mean particle diameters of 6.7, 17.6 and 30.8 µm,respectively.

40 4. Results

petcoke chemical conversion and energy conversion efficiency. Higher (H2O/C)slurry

favors the reaction kinetics, but at the expense of higher mass flow rates and shorter

residence times, resulting in lower XC. It also results in a decrease of ηchem because

of the energy wasted to evaporate and superheat excess water. The opposite is

true for ηproc, assuming the sensible heat of excess steam exiting the reactor is

recovered. Best results were obtained for (H2O/C)slurry = 1. Plugging of the feeding

systems was observed for (H2O/C)slurry < 1 and 8.5 µm-particles. In principle, small

particles offer high specific surface area and, consequently, enhance the reaction rate.

Unreacted particles could be recycled in an industrial application.


Table 4.2: Water and petcoke feeding rates, composition of the product gas, nom-inal temperatures and performance parameters for 29 valid runs of experimentalcampaign 2 conducted with a coke-water slurry.

D30

mco

ke

mH

2O

(H2O

/C

) slu

rry

nA

rn

H2

nC

On

CO

2n

CH

4Q

sola

rT

cavity

Tsh

ell

XC

XH

2O

η chem

η pro

cτ

#µm

g/m

inm

ol/m

olm

ol/m

inkW

KK

––

––

s

16.

71.

29.

63.

00.

130.

040.

010.

020

0.00

05.

015

6612

880.

290.

080.

010.

121.

02

6.7

2.1

10.3

3.0

0.14

0.02

0.01

0.00

50.

002

5.1

1516

1066

0.10

0.03

0.01

0.10

1.0

330

.81.

78.

62.

00.

140.

050.

010.

022

0.00

24.

215

1411

210.

250.

110.

020.

131.

14

30.8

1.9

7.7

1.5

0.14

0.08

0.04

0.02

10.

002

4.1

1478

1227

0.44

0.19

0.03

0.17

1.2

530

.81.

36.

51.

50.

140.

050.

030.

014

0.00

23.

614

3111

880.

450.

150.

030.

151.

46

30.8

3.4

10.8

1.5

0.14

0.07

0.03

0.02

30.

006

4.1

1393

1151

0.25

0.13

0.03

0.16

1.0

730

.80.

95.

81.

50.

140.

040.

020.

011

0.00

13.

715

3012

680.

530.

140.

020.

131.

58

30.8

0.3

4.7

1.5

0.13

0.01

0.00

0.00

40.

000

3.3

1542

1275

0.31

0.04

0.00

0.08

1.8

930

.81.

97.

71.

50.

130.

080.

040.

021

0.00

23.

715

3312

880.

420.

180.

030.

181.

210

17.6

0.8

6.1

2.0

0.13

0.05

0.02

0.01

10.

001

3.8

1536

1283

0.51

0.12

0.02

0.14

1.5

1117

.61.

88.

72.

00.

130.

100.

040.

021

0.00

24.

015

1112

530.

500.

170.

040.

211.

112

17.6

1.1

6.1

1.5

0.13

0.06

0.02

0.01

40.

001

4.3

1540

1295

0.44

0.14

0.02

0.13

1.4

1317

.62.

08.

11.

50.

130.

110.

040.

025

0.00

34.

715

0712

640.

450.

200.

030.

181.

114

17.6

3.6

11.2

1.5

0.13

0.16

0.06

0.03

40.

005

4.7

1489

1244

0.38

0.21

0.05

0.23

0.9

1517

.61.

46.

81.

50.

160.

080.

030.

018

0.00

24.

614

6611

950.

490.

180.

020.

141.

316

17.6

1.4

4.8

1.5

0.16

0.08

0.03

0.01

60.

002

4.7

1475

1209

0.51

0.25

0.02

0.14

1.6

1717

.61.

45.

81.

50.

160.

090.

040.

019

0.00

24.

814

8412

200.

600.

240.

030.

151.

418

17.6

0.6

4.8

1.0

0.13

0.04

0.01

0.00

90.

001

3.2

1547

1309

0.49

0.11

0.01

0.14

1.7

1917

.62.

47.

21.

00.

130.

110.

050.

020

0.00

33.

815

1012

490.

380.

210.

040.

201.

220

17.6

1.4

5.8

1.0

0.13

0.08

0.03

0.01

50.

002

4.1

1502

1234

0.52

0.20

0.03

0.16

1.5

2117

.61.

44.

81.

00.

130.

080.

040.

015

0.00

14.

015

3312

660.

530.

250.

030.

161.

622

17.6

1.4

3.8

1.0

0.13

0.07

0.03

0.01

30.

001

3.9

1543

1276

0.48

0.28

0.03

0.15

1.8

2317

.60.

64.

81.

00.

130.

040.

020.

009

0.00

03.

614

9612

450.

690.

150.

020.

131.

724

17.6

0.6

4.8

1.0

0.09

0.05

0.02

0.01

10.

001

3.7

1514

1257

0.80

0.17

0.02

0.14

1.9

2517

.60.

63.

81.

00.

090.

060.

030.

012

0.00

23.

715

1812

620.

870.

230.

020.

142.

226

17.6

0.6

2.8

1.0

0.14

0.05

0.02

0.00

90.

001

3.8

1525

1269

0.73

0.26

0.02

0.12

2.3

2717

.61.

45.

81.

00.

090.

080.

030.

014

0.00

23.

815

1212

580.

470.

180.

030.

161.

628

17.6

1.4

4.8

1.0

0.09

0.07

0.03

0.01

20.

002

3.8

1499

1244

0.44

0.21

0.03

0.15

1.8

2917

.62.

49.

21.

00.

090.

130.

050.

024

0.00

34.

414

5511

980.

440.

200.

040.

211.

1

42 4. Results

4.3 Vacuum Residue — Liquefied Injection 3

Table 4.3 lists the operational parameters and results of 12 solar experimental runs

taken under approximate steady-state conditions. Mass flow rate of VR mVR was

in the range 0.65–4.17 g/min. Mass flow rate of steam mH2O — injected coaxially

— was in the range 2–6 g/min, corresponding to a H2O/C molar ratio between 1.12

and 7.18. Ar was injected through purging nozzles to protect the window at a mass

flow rate in the range 0.27–0.36 mol/min. The resulting average residence time for

the coke particles was calculated using eq. (3.12), where nH2,feedstock denotes the

hydrogen bound in the VR, which is released by pyrolysis at the entrance of the

reactor. Residence time τ for all runs was in the range 0.8–1.8 seconds. The solar

power input Qsolar varied in the range 3.6–6.8 kW, which corresponds to a mean solar

flux concentration ratio over the reactor’s aperture in the range 1833–3463 suns (1

sun = 1 kW/m2). The nominal reactor temperature Treactor was in the range 1420–

1567 K. Typical heating rates exceeded 500 K/s. VR was firstly pyrolyzed, releasing

volatile hydrocarbons such as C2H2, C2H4, and C2H6 as reported previously [73, 6].

The remaining solid particles, typically about 15–24 %wt of the original VR —

denoted Conradson carbon residue (CCR, [30]) — underwent steam-gasification to

produce primarily H2, CO, and CO2, consistent with the thermodynamic equilibrium

predictions. Chemical conversion for steam XH2O and carbon XC reached up to 44%

and 50%, respectively.

The variation of temperatures and gas compositions during a representative so-

lar experimental run (run 12 of Table 4.3) is shown in Fig. 4.6. In this run, the

reactor was first heated to above 1550 K under an Ar flow. Thereafter, reactants

were continuously introduced during an interval of 10 minutes at mVR=1.2 g/min

and mH2O=2 g/min, corresponding to H2O/C molar ratio of 1.3. Average molar

flow rates of gaseous products, measured at Treactor=1567 K under approximately

steady-state conditions in the interval indicated by vertical dashed bars of Fig. 4.6,

were: nH2=0.08 mol/min, nCO=0.03 mol/min, nCO2=0.01 mol/min, and nCH4=0.002

mol/min. Negligible amounts of volatile hydrocarbons (C2H4 and C2H2 at 11 and

60 ppm, respectively) were derived from the pyrolysis of VR, occurring immediately

after exposure to the high-flux solar irradiation. Approximately 40%wt of the CCR

3Material from this section has been published in: A. Z’Graggen, P. Haueter, G. Maag, M.Romero, and A. Steinfeld. Hydrogen production by steam-gasification of carbonaceous materialsusing concentrated solar energy — IV. reactor experimentation with vacuum residue. InternationalJournal of Hydrogen Energy,33:679–684, 2008.

4.3. Vacuum Residue — Liquefied Injection 43

0

5

10

15

pro

duct

gas

com

pos

itio

n,%

3

4

5

1000

1200

1400

1600

Qsolar,kW

tem

per

ature

,K

0 2 4 6 8 10 12time, min

Tcavity

Tshell

Qsolar

H2

CH4

CO

CO2

start ofVRfeeding

Figure 4.6: Temperatures, solar power input and product gas composition for atypical run of Campaign 1 (run # 12 in Table 4.3). The dashed vertical line indicatesthe start of the petcoke feeding, the two vertical dotted lines delimit the intervalconsidered for the steady state calculations.

1

2

3

4

5

6

pow

er,kW

1300

1400

1500

1600

tem

per

ature

,K



Qsolar

Tcavity


44 4. Results

0

0.2

0.4

0.6

0.8

1

experimental runs ordered by increasing mVR

rela

tive

orig

inof

pro

duct

H2 χH2,pyrolysis

χH2,gasification

mVR

mH2O

0

1

2

3

4

5

6

7

m,g

/min

Figure 4.8: Fractions of hydrogen in the product gas derived by either pyrolysis orgasification, and the feeding rates of reactants VR and steam for experimental runsof Campaign 3.

deposited as solid agglomerates inside the cavity (size up to 5 mm), which contained

90% carbon, 3.5% sulfur, and 5% metallic fraction constituted by 75% V and 18%

Ni. This composition is comparable to that of Petrozuata Delayed coke [101]. The

rest consisted of particles with a carbon content of 96.5%, a BET specific surface

area of 20.5 m2/g (determined with Micromeritics TriStar Analyzer), and a mean

particle size of 6.5 µm(determined by laser scattering with a Horiba LA950). Energy

conversion efficiencies up to 2% and 17% were achieved for ηchem and ηproc, respec-

tively. Main heat losses are due to re-radiation through the aperture and conduction

through the insulation, with average shares of 17% and 73%, respectively, as shown

in Fig. 4.7. Re-radiation losses, which take into account attenuation by the spec-

trally selective quartz window and emission through the aperture, were calculated

using the radiosity method [124]. Conduction heat losses, which take into account

heat bridges through the oil-cooled components, were calculated using the thermal

conductivities of the insulation materials, as provided by the manufacturer (for de-

tails see Chapter 6). Optimization for minimizing heat losses and maximizing the

solar energy conversion efficiency was outside the scope of this study. Figure 4.8

reveals the origin of the H2 in the product gas. Plotted for all experimental runs are

the fractions of hydrogen derived by either pyrolysis or gasification, χH2,gasification and

χH2,pyrolysis, and the feeding rates of reactants, mVR and mH2O (ordered by increasing

4.3. Vacuum Residue — Liquefied Injection 45

mVR). Average values were χH2,pyrolysis=63% and χH2,gasification=37%. As expected,

a higher mVR resulted in a higher χH2,pyrolysis since H2 contained in the VR is com-

pletely released during the pyrolysis step. In contrast, mH2O had negligible effect

because H2O was fed in excess and only partly consumed by the gasification process.

Ideally, complete conversion of carbon bound in the vacuum residue would result in

χH2,pyrolysis=40% and χH2,gasification=60%.

46 4. Results

Table 4.3: Steam and VR feeding rates, composition of the product gas, nominaltemperatures and performance parameters for 12 valid runs of experimental cam-paign 3 conducted with liquefied VR.

mV

Rm

H2O

nA

rn

H2

nC

On

CO

2n

CH

4C

C2H

4C

C2H

6C

C2H

2Q

sola

rT

cavity

Tsh

ell

XC

XH

2O

η chem

η pro

cτ

#g/

min

mol

/min

ppm

kWK

K–

––

–s

11.

03.

00.

290.

050.

020.

010

0.00

514

629

2238

5.6

1420

1096

0.44

0.23

0.01

0.09

1.5

24.

26.

00.

360.

170.

030.

008

0.01

314

6729

1943

5.6

1488

1160

0.16

0.13

0.02

0.18

0.8

30.

76.

00.

360.

050.

010.

011

0.00

232

021

44.

214

7212

320.

500.

100.

010.

131.

04

1.2

6.0

0.36

0.07

0.01

0.01

00.

005

880

619

3.8

1552

1274

0.31

0.10

0.02

0.17

1.0

52.

26.

00.

360.

110.

020.

010

0.00

830

82

1381

4.1

1543

1271

0.24

0.12

0.02

0.19

0.9

62.

26.

00.

270.

110.

020.

013

0.00

523

12

999

5.4

1534

1233

0.27

0.15

0.02

0.15

1.0

70.

74.

00.

270.

040.

010.

008

0.00

253

034

43.

914

9112

640.

410.

120.

010.

111.

48

1.2

4.0

0.27

0.06

0.01

0.00

80.

003

800

570

3.6

1501

1298

0.27

0.13

0.01

0.14

1.3

91.

74.

00.

270.

080.

020.

009

0.00

488

159

44.

215

4012

900.

240.

160.

020.

141.

210

0.7

2.0

0.27

0.03

0.01

0.00

70.

001

160

124

4.1

1503

1320

0.34

0.20

0.01

0.08

1.8

111.

22.

00.

270.

050.

010.

007

0.00

540

83

1345

6.8

1513

1191

0.32

0.27

0.01

0.07

1.7

121.

22.

00.

280.

080.

030.

010

0.00

211

060

4.9

1567

1332

0.48

0.44

0.02

0.12

1.5

4.4. Comparison 47

4.4 Comparison

Table 4.4: Average operational parameters and results of the solar experimentalcampaigns conducted with dry coke powder, cokewater slurry, and VR.

feedstock dry coke powder4 coke-water slurry VR4

mfeedstock, g/min 3.3 1.47 1.50mH2O, g/min 7.17 6.61 4.25(H2O/C), mol/mol 1.71 4.0 2.74Tcavity, K 1514 1506 1510XC, % 68.9 47.5 33.1XH2O, % 52.6 17.3 17.6

Qsolar, W 5177 4079 4683Xchem, % 7.0 2.5 1.4Xproc, % 28.3 15.2 13.2τ , s 0.92 1.40 1.274Steam fed separately

Table 4.4 summarizes the main average process parameters for three experimental

campaigns performed with comparable solar power inputs and reactor temperatures,

and with three different feedstocks: (1) dry coke powder (with steam fed separately),

(2) coke-water slurry, and (3) VR (with steam fed separately). The campaign with

dry coke powder outperforms the other two in terms of chemical conversion and en-

ergy efficiency. For example, ηproc is 28.3%, 15.2% and 12.3% for dry coke powder,

slurry, and VR, respectively. In the slurry campaign, the overall performance was

diminished by excess water fed as a result of the high H2O/C molar ratios required

for the slurry. In the VR campaign, besides using excess water, the formation of

solid agglomerates affected negatively the chemical conversion. The implementation

of a spray nozzle may facilitate dispersion into smaller particles and, consequently,

enhance the reaction kinetics. The aerodynamic protection of the window worked

well for the duration of all experimental runs, each about 45 minutes. The win-

dow was maintained clean and clear from particle deposition or condensable gases.

The oil-cooled frustum served as a buffer zone between the window and the cavity

(reaction chamber), and prevented steam condensation. In a scale-up application

for a solar tower concentrating facility, this annular component may be heated with

spilled radiation from the heliostat field. Argon should be completely substituted

by steam to avoid the energy and cost of recycling an inert gas.

48 4. Results

Part II

Modeling

Chapter 5

Simulation Framework

The simulations presented in the second part of this thesis share a common set

of boundary conditions and models for the chemical kinetics and the polydisperse

particulate medium. The characteristics of the incoming solar radiation as well as the

conductive behavior of the reactor walls were determined by stand-alone preliminary

simulations. Wall and window radiative properties are presented. The concept of

‘equivalent monodisperse diameter’ is introduced to model the polydisperse feedstock

types, which have random size-distribution functions. Finally, the kinetic rate laws

for steam gasification of petcoke developed by Trommer [102] are presented.

5.1 General Boundary Conditions

5.1.1 Incoming Radiation

An experimentally validated 3D Monte-Carlo ray-tracer was used the determine the

exact properties of the concentrated radiation incident on the reactor’s aperture.

The simulation setup is shown in Fig. 5.1. The location and direction of incident

rays on the aperture was recorded taking into account the geometrical characteristics

of the concentrating facilities, specular reflection errors, and non-parallelism of the

sun rays. Figure 5.2 shows the solar power through the aperture Qsolar and the

solar flux qsolar as a function of the aperture’s radius for (a) PSI’s solar furnace

(nominal Qsolar=5 kW), and (b) CIEMAT’s solar tower (nominal Qsolar=300 kW,

data from CIEMAT). Shown in 5.3 is the averaged angular solar flux distribution at

the aperture, used as input to the solar reactor. Finally, the solar spectrum of the

incoming radiation is approximated by Planck’s blackbody emission at 5780 K.

52 5. Simulation Framework

30

20

100

10

0

10

10

5

0

5

10

reactoraperture

parabolic concentrator

heliostat

incident solarradiation

reactor

x, m y, m

concentratedradiation

z, m

Figure 5.1: Simulation setup used to determine the angular and radial distributionsof the incoming concentrated solar radiation at PSI’s solar furnace [34].

The same simulation setup was used in the early stages of the engineering design

to estimate the radiative flux intensities to be expected inside the cavity. Figure 5.4

shows lines of equal incident solar flux qsolar in kW, projected on planes parallel to

the aperture. An aperture diameter of 50 mm and a nominal power through the

aperture of 5 kW were considered, while no interaction with the cavity walls was

modeled. Maximal qsolar>3050 W/m2 is reached on a small spot at the center of

the aperture. Peak values of 2959, 2815, 2088 and 1318 W/m2 are predicted for

planes located 1, 2, 3 and 4 cm behind the aperture. Also indicated is the cavity of

the prototype reactor presented in Chapter 3 (gray). In that reactor particles are

injected on a plane 4 cm away from the aperture in order to take advantage of the

still highly concentrated radiation (1318 kW/m2), on the one hand, and to prevent

the particles to exit the reactor through the aperture on the other.

5.1. General Boundary Conditions 53

0 0.025 0.050

1

2

3

Daperture, m

q solar,M

W/m

2

0

1

2

3

4

5

Qsolar,kW

(a)

0 0.25 0.50

0.5

1

1.5

2

2.5

Daperture, m

q solar,M

W/m

20

50

100

150

200

250

300

Qsolar,kW

(b)

Figure 5.2: Solar radiative flux (left axis) and solar power (right axis) as a functionof the aperture’s diameter for (a) PSI’s solar furnace, and (b) CIEMAT’s solar tower.

0

π/4−π/4

π/8−π/8

incidentradiation

PSI’s solarfurnaceCIEMAT’ssolar tower

Figure 5.3: Angular distribution of the concentrated solar radiation at the aperturefor PSI’s solar furnace and CIEMAT’s solar tower.


0 0.02 0.04 0.06 0.08 0.1

-0.05

-0.025

0

0.025

0.05

x, m

r,m

incomingradiation

50 mmaperture

250

500

10002000

28003050

100

Figure 5.4: Lines of equal projected incident solar flux qsolar in kW/m2 on planesparallel to the aperture. Results are given for a 50 mm aperture and a nominalpower trough the aperture Qsolar=5 kW. Also indicated is the shape of the prototypereactor’s cavity (gray).

5.1.2 Reactor Walls1

Figure 5.5 shows the heat fluxes at a cavity wall element considered in the simulation.

Conductive heat transfer was preliminarily simulated with a 2D axi-symmetric FE

solver. The exact reactor geometry shown in Fig. 3.1 was considered, including the

inner ceramic Al2O3 liner (k = 25 W/mK), the Inconel 601 shell (k = 11.2 W/mK),

the outer insulation (Insulform 160, k=0.09–0.3 W/mK), the oil-cooled aluminium

frustum (k = 186 W/mK) as well as the graphite insulation between the shell and

the frustum and the 1 mm gap between the Al2O3 liner and the shell. Typical

radiative loads on the liner were calculated with the raytracer presented in Section

5.1.1 and for measured Qsolar, the heat removed by the oil cooling was estimated

from the measured temperature differences between inlet and outlet. Finally, natural

convection for a cylinder was assumed on the outside. Figure 5.6 shows the average

overall heat transfer coefficient extracted from 23 simulation runs based on the

measurements of Campaign 1. Values range between 12 W/ (m2K) at the center of

the cavity to 180 W/ (m2K) close to the aperture due to heat bridges. The overall

1The FE simulation used in this section has been set up in the framework of: L. Donati,Hydrogen production by steam-gasification of petroleum coke using concentrated solar power —Thermal Simulation of the Synpet Solar Reactor with ANSYS, Semester Thesis, ETH Zurich, 2005.


insideoutside

reactor liner (Al2O3)

reactor shell (Inconel)reactor insulation(Al2O3 / ZrO2)

Twall

Tshell

natural convection

T1

condicq

q

σεT 4

wall

Figure 5.5: Schematic of heat fluxes at the wall. Radiation emitted by particlesand other cavity wall elements, qic, is incident from the inside (right), while the wallitself emits radiation toward the inside. On the outside (left) the reactor is cooledby natural convection. Finally, conduction occurs through the three material layersof the wall.

heat transfer coefficient U is related to the oil temperature on the front cone and to

the ambient temperature T∞ for the rest of the cavity.

Diffuse gray surfaces (ε = 0.8) are assumed for the reactor’s inner walls because

of deposited petcoke particles. The impact of the uncertainty of ε on Tcavity and

XC was assessed by sensitivity analysis (see Section 7.4). In the average, a change

of 10% in ε induced a change of 0.4 % in Tcavity and XC. Convective heat transfer

from the walls to the fluid is found by CFD to be two orders of magnitude smaller

than the heat transfer by radiation and conduction, and is therefore neglected. An

energy balance at the wall leads then to the following implicit expression for the

temperature: ∫ ∞

0

qic,λελ dλ− σεT 4wall − qcond = 0 (5.1)

where qic is the incident thermal radiation, the second term accounts for radiation

emitted from the wall, and the third term accounts for the conduction heat loss

through the wall:

qcond = U · (T∞ − Twall) (5.2)


0 5 10 15 20 250

40

80

120

160

front cone

cylinderrear cone

U,W/(

m2K

)

x, cm

Figure 5.6: Overall heat transfer coefficient U for conductive losses trough the reactorwalls as a function of the location along the reactor axis. Also indicated is thelongitudinal cross-section of the reactor’s cavity.

5.1.3 Quartz Window

The spectral directional reflectivity ρλ (θ) of the quartz window surface is calculated

with the Fresnel equations [68] and the complex refractive index n + ik from [72],

while the absorption coefficient of the window is given by κλ = 4πkλ/λ. The ‘one

pass’ transmittance trough the window τ is then given by τ = e−κ·s, where s is

the distance traveled in the quartz, determined by the window thickness and the

incidence angle θ. Assuming similar refractive indexes on both sides of the win-

dow, the total transmittance, reflectance and absorptance used in the simulation are

expressed by (subscripts λ omitted for clarity) [68]:

Trw =(1− ρ)2 τ

1− ρ2τ 2(5.3)

Rw = ρ

(1 +

(1− ρ)2 τ 2

1− ρ2τ 2

)(5.4)

Aw =(1− ρ) (1− τ)

1− ρτ(5.5)

Figure 5.7 shows the complex refractive index of quartz and the total transmittance

Trw of a 3 mm thick window for selected incident angles θ = 0, 45 and 60 as a


0.1 1 100

0.5

1

1.5

2

2.5

λ, µm

n;k

n

k

0

0.2

0.4

0.6

0.8

1

Tr

w,θ

Trw,0Trw,45

Trw,60

Figure 5.7: Complex refractive index of quartz n + ik [72] and transmittance Trw

for a 3 mm thick window and for three selected incident directions θ as a functionof wavelength λ.

function of the wavelength λ. Strong selective behavior of the window is observed,

with virtually zero transmittance below 0.2 µm and above 4 µm and values higher

than 0.8 in between. The influence of the incidence angle is marginal; for example

average values of 0.93, 0.91 and 0.83 are found for the normal direction (0), for the

maximal acceptance angle of the facility (45) and for 60, respectively.

Figure 5.8 shows the heat fluxes considered at the window. The window temper-

ature is assumed constant over it’s thickness of 3 mm. Convective heat losses on the

outside are then given by qconv = h · (T∞ − Tw) where h = 263 W/(m2K) is obtained

from the empirical correlation Nux = 0.0308 · Re4/5x · Pr1/3 for turbulent flow over

a flat plate [44], as the window is cooled externally by a ventilator. The impact

of the uncertainty of h on Tcavity and XC was assessed by sensitivity analysis (see

Section 7.4). In the average, a change of 10% in h induced a negligible change of

0.03 % in Tcavity and XC. Similarly to eq. (5.1) heat balance for the window yields:∫ ∞

0

(qic,λ + qsolar,λ) Aw,λ dλ− 2σAw (Tw) T 4w − qconv = 0 (5.6)


insideoutside

window

icsolar

µ = 45º

ventilator

conv

Tw

T1

q

q q

σAwT 4

wσAwT 4

w

Figure 5.8: Schematic of heat fluxes at the window. Concentrated solar radiationqsolar is incident from the outside of the reactor (left) whereas radiation emitted bythe cavity walls and particles, qic, is coming from the inside (right). The window itselfemits radiation on both sides. Finally, the window is cooled by forced convection onthe outside.

5.2 Polydisperse Coke Particles

The polydisperse medium is characterized by its population density, i.e. the particle

size distribution function f describing the number of particles found in an infinites-

imal interval around diameter D. For a sample containing a discrete number of

particles Np with respective diameters Di the continuous population density f is

given by the sum of each particle’s contribution:

f (D) =1

Np

Np∑i=1

δ (D −Di) =1

Np

Np∑i=1

w (D −Di, ∆D) (5.7)

where the Dirac delta function δ is substituted for a symmetric kernel function w

around 0 with bandwidth ∆D and with∫∞

0w dD = 1 (typically, w is a Gaussian

error distribution curve). The solid lines in Figure 5.9 show f (D) ·D3, the volume

density measured by laser scattering, for feedstocks type 1 and 3 considered here

(see also Fig. 2.2). As a consequence of the chemical reaction the particles shrink by

a factor ηsk = Di (XC > 0) /Di (XC = 0), thus influencing the distribution function:

fsk (D) =1

Np

Np∑i=1

w(D −D0

i · ηsk, ∆D)

(5.8)

5.2. Polydisperse Coke Particles 59

1 10 100

f(D

)·D

3

D, µm

XC = 0.0XC = 0.5XC = 0.75

type 1type 3

Figure 5.9: Volume density of the polydisperse particles used in Campaigns 1 (type1) and 2 (type 3). Solid lines are measured values for the unreacted samples, dashedand dashed-dotted lines are calculated results at carbon conversions of 0.5 and 0.75,respectively.

where ηsk is a function of the particle diameter and of the extent of the reaction (see

Section 5.3 for details about the differences in reactivity as a function of particle

size).

Two examples of the modified distribution function are given in Fig. 5.9, for a

total reaction extent XC=0.5 and 0.75:

XC = 1−∫∞

0fsk · (D · ηsk)

3 dD∫∞0

f ·D3 dD(5.9)

As expected, bigger particles tend to react slower than the small ones leading to a

shape change of the distribution function as the chemical reaction progress, observ-

able especially for feedstock type 3. For simplification, the polydisperse medium can

be described by a set of equivalent monodisperse diameters defined generally as:

Dpq =

(∫∞0

f (D) Dp dD∫∞0

f (D) Dq dD

)( 1p−q )

(5.10)

where p and q are integer numbers between 0 and 3. D30, D31, D32 are applied for


0 0.75 0.95 0.99 1

0.1

1

10

100

XC

D,µm

Dη

D32

D31

D30

type 1

type 3

Figure 5.10: Equivalent monodisperse diameters calculated with eqs (5.10) and(5.26) of the polydisperse particles used in Campaigns 1 (type 1) and 2 (type 3)as a function of the carbon conversion.

volume-based, volume-to-diameter-ratio based, and volume-to-surface-ratio based

phenomena, respectively (see also eqs (5.13) and (7.13) for examples of application).

Figure 5.10 shows the variation of the equivalent monodisperse diameters as a func-

tion of the reaction extent XC for feedstock types 1 and 3 with mean initial sizes D30

of 2.21 and 17.58 µm, respectively. The difference between the equivalent diameters

indicate the importance of considering the polydisperse nature of the particulate

medium, especially for the feedstock type 3, where the difference reaches one order

of magnitude. Note that eqs (5.10) and (5.9) lead to a simple relation for the mean

diameter of the polydisperse medium:

D30 = D030 · (1−XC)

13 (5.11)

5.2.1 Radiative Properties

Values for the complex refractive index of coke are given in Table 5.1 [14, 29]. For the

solid volume fractions fV < 10−3 considered in this study the independent scattering

regime is valid [91]. Continuum absorption, scattering, and extinction coefficients

for the polydisperse medium are then found as a function of the corresponding


Table 5.1: Complex refractive index of coke [14, 29].

λ, µm n k

0.308 1.96 0.8691.35 2.04 0.9012.40 2.11 0.9343.44 2.19 0.9664.49 2.27 0.9985.53 2.35 1.03

properties of single spheres [68]:

κλ, σsλ, βλ =3fV

2·∫∞

0Qa,λ, Qs,λ, Qext,λ f (D) D2 dD∫∞

0f (D) D3 dD

(5.12)

For size parameter ξ = πD/λ in the range 0.1 − 103 considered here, Mie theory

applies [11]. The efficiency factors for absorption, scattering and extinction of a

single sphere, Qa,λ, Qs,λ, and Qext,λ are calculated based on the complex refractive

index of the solid material using the routine BHMIE [11]. For ξ > 5, geometrical

optics applies and eq. (5.12) simplifies to:

κλ, σsλ, βλ =3

2

fV

D32

· Qa,λ, Qs,λ, Qext,λ (5.13)

Comparison of the values calculated with eq. (5.12) and with the simplified approach

of eq. (5.13) showed negligible differences for the particle size distributions considered

in this study (note that for the comparison Qa,λ, Qs,λ, and Qext,λ in eq. 5.13 were

evaluated at D32). κλ, σsλ and βλ of the polydisperse particle cloud are therefore

calculated by means of eq. (5.13) based on the equivalent diameter for radiation,

D32.

Figures 5.11 (a) and (b) show the spectral absorption and scattering coefficients

of the particulate medium in Campaigns 1 (feedstock type 1) and 2 (type 3), respec-

tively, for a typical initial volume fraction fV = 5 · 10−5, and for carbon conversions

XC = 0, 0.5, and 0.75. The smaller particles of Campaign 1 absorb more than 10

times better that those of Campaign 2.

Analogous to eq. (5.12) the scattering phase function of the polydisperse particle


0.5 1 2 4

20

40

80

120

κλ;σ

λ,s,1/

m

λ, µm

σλ,s

κ λ

XC = 0.0

0.5

0.75

(a)

0.5 1 2 4

1

2

4

8

κλ;σ

λ,s,1/

m

λ, µm

σλ,s

κ λ

XC = 0.0

0.5

0.75

(b)

Figure 5.11: Spectral distribution of the absorption and scattering coefficients, cal-culated for a typical volume fraction of 5 · 10−5 and carbon conversions of 0.0, 0.5and 0.75, for the feedstock used in (a) Campaign 1, and (b) Campaign 2.

cloud is found by a weighted sum of the phase functions of each particle size [68]:

ΦTλ (ω0) =

∫∞0

Φλ (D, ω0) ·Qs,λ · f (D) D2 dD∫∞0

Qs,λ · f (D) D2 dD(5.14)

where Φλ is calculated with the routine BHMIE [11]. Because of computational time

restraints, the Henyey-Greenstein approximation [36] is introduced,

ΦTλ (D, θ0) =1− g2

Tλ

(1 + g2Tλ − 2gTλ · cos θ0)

3/2(5.15)

where the asymmetry factor gTλ is calculated similarly to eq. (5.14):

gTλ =

∫∞0

gλ (D) ·Qs,λ · f (D) D2 dD∫∞0

Qs,λ · f (D) D2 dD(5.16)

and the asymmetry factor gλ for a single particle size is defined as

gλ = 14π

∫4π

Φλ (D, θ) sin θ dΩ.

The scattering phase functions for particles of Campaign 1 (feedstock type 1)

are shown in Figs 5.12 (a) and (b) for carbon conversions 0.0 and 0.75, respec-


10−1

100

101

102

103

θ, rad

Φ

Mie for D32 = 3.134 µmMie, polydisperse at XC = 0.0HG, polydisperse at XC = 0.0

(a)

λ = 0.5 µm

λ = 2.0 µm

0 π/2π/4

θ, rad

Φ

λ = 0.5 µm

λ = 2.0 µm

Mie for D32 = 1.99 µmMie, polydisperse at XC = 0.75HG, polydisperse at XC = 0.75

(b)

0 π/2π/410

−1

100

101

102

103

Figure 5.12: Scattering phase function for the equivalent diameter D32, for thepolydisperse medium (eq. (5.14)), and for the Henyey-Greenstein approximation(eq. (5.15)), calculated for carbon conversions of (a) 0.0; and (b) 0.75, for thefeedstock used in Campaign 1. Directions θ = π/2 – π not shown in plot, becauseno backward scattering peak was observed.

tively. Two selected wavelengths λ=0.5 and 2 µm are considered, corresponding to

the locus of maximum emissive power Eλb for radiation emitted at 5780 and 1450

K, respectively. Curves are plotted for a monodisperse particulate medium with

equivalent diameter D32, for the polydisperse particulate medium of Fig. 5.9, and


for the Henyey-Greenstein (HG) approximation (eq. (5.15)). Preference for forward

scattering is observed. As expected, the polydisperse particle cloud has a smoother

scattering phase function, which is well approximated by the HG equation.

5.3 Chemical Kinetics

Table 5.2: Arrhenius parameters of the kinetic rate constants for the steam gasifi-cation of coke [100].

EA, J/mol k0, 1/s

K1 2.707 · 105 1.158 · 103

K2 1.615 · 102 4.978 · 10−1

K3 −9.404 · 104 1.149 · 10−7

K4 7.068 · 103 4.033 · 10−8

K5 4.551 · 102 8.152 · 10−6

The chemical kinetics of PD coke steam-gasification were extensively investigated

by D. Trommer [102]. This section summarizes the findings pertinent to the present

study. The overall chemical conversion can be represented by the simplified net

reaction:

CHzOy + (1− y) H2O =(z

2+ 1− y

)H2 + CO (5.17)

where z and y are the elemental molar ratios of H/C and O/C in the coke, re-

spectively. The enthalpy change of the endothermic reaction is ∆HR |298K= 131

kJ/molC. Table 2.2 lists the approximate main elemental chemical composition, the

low heating value (LHV), and elemental molar ratios of H/C and O/C for Petrozu-

ata Delayed (PD) coke used in the experimental campaign. The gasification kinetic

model is based on the oxygen-exchange mechanism describing reversible O-transfer

surface reactions followed by an unidirectional gasification step, and on reversible

steam sorption as OH/H groups and irreversible surface chemistry [102]. A set of

kinetic rate laws of the Langmuir-Hinshelwood type are formulated to describe the

formation and consumption of each gas species [102]:

rH2O =−K2 · pH2O −K2K3 · pH2OpCO

1 + K4pCO2 + K5pH2O

(5.18)

5.3. Chemical Kinetics 65

rH2 = −rH2O (5.19)

rCO =2K1 · pCO2 + K2 · pH2O −K2K3 · pH2OpCO

1 + K4pCO2 + K5pH2O

(5.20)

rCO2 =−K1 · pCO2 + K2K3 · pH2OpCO

1 + K4pCO2 + K5pH2O

(5.21)

rC = − (rCO + rCO2) =−K1 · pCO2 −K2 · pH2O

1 + K4pCO2 + K5pH2O

(5.22)

ri is the reaction rate of species i ( i = H2, H2O, CO, CO2 and C) for heterogeneous

surface reactions defined as:

− ri =1

mC

dni

dt(5.23)

where mC is the mass of coke. In order to account for mass transfer limitations

inside the solid particle for varying particle diameter, the particle effectiveness η,

experimentally determined by [100], is introduced,

η (D) = 0.571e−1.29·104·D + 0.429e−9.56·102·D (5.24)

η converges to 1 for D approaching 0 and diffusion becoming instantaneous. The

temperature dependence of each Ki is determined by imposing the Arrhenius law

Ki (T ) = k0 exp(−EA

RT

). Apparent activation energies and frequency factors, cal-

culated by linear regression of experimental data obtained by thermogravimetric

measurements, are listed in Table 5.2 [100]. Comparison of the gasification rates be-

tween the experimental runs performed in the thermogravimeter (packed bed with

heating rates of 0.1–0.3 K/s by predominantly convective/conductive heat transfer)

and the experimental runs performed in the solar reactor (gas-particle entrained

flow with heating rates of 104–106 K/s by predominantly radiative heat transfer, see

Chapter 7) led to the introduction of an empirical Arrhenius-type proportionality

constant ksolar = 498 · exp (−5.71 · 104/RT ) to resolve the mass/heat transfer dif-

ferences between both set-ups and the fact that the release of volatiles by pyrolysis

occurs immediately at the entrance of the solar reactor. The volume-specific reaction

rate for species i and for a single particle diameter D can then be written as:

Ri = ksolar · η (D) · ri ·M · YC · ρ (5.25)

For the polydisperse medium η is evaluated at the equivalent monodisperse diameter


Dη,

Dη = η−1

(∫∞0

f (D) D3 · η (D) dD∫∞0

f (D) D3 dD

)(5.26)

where η−1 denotes the inverse function of eq. (5.24).

Chapter 6

System Analysis — Lumped

Parameters Model1

The engineering design of the reactor presented in the first part of this thesis was

supported by performance calculations based on a simple lumped-parameters model.

Figure 6.1 shows the mass and energy flows that are considered. The system bound-

ary is defined by the reactor shell. Coke and Argon are fed at ambient temperature,

whereas H2O is first evaporated and fed at Tsteam=423 K. Inside the cavity the re-

actants are heated to the reaction temperature Tcavity and eventually react to form

a mixture of Ar, H2, CO, CO2 and residual steam and carbon. The products exit

the reactor at Tcavity and are cooled down to Thx=473 K in a heat exchanger, where

a part of the sensible heat is recovered. A fraction of the incident solar power Qsolar

is lost either by radiation, Qrerad, which accounts for reradiation from the cavity as

well as for emission and reflection from the window, or by conduction through the

cavity walls Qcond.

6.1 Governing Equations

A black box system delimited by the inner cavity walls is considered. The temper-

ature at any location inside the cavity and on the cavity walls is equal to Tcavity.

1Material from this section has been published in: A. Z’Graggen, P. Haueter, D. Trommer, M.Romero, J. C. de Jesus, and A. Steinfeld. Hydrogen production by steam-gasification of petroleumcoke using concentrated solar power — II. reactor design, testing, and modeling. InternationalJournal of Hydrogen Energy, 31:797–811, 2006.

68 6. System Analysis — Lumped Parameters Model

Ar

coke

H2O

Tsteam

T1

steam

coke

Ar

H2O

Tcavity

chem

heatingsolar

rerad

sensible

(1-XC) C

(1-XH2O) H2O

Ar

XC C

H2O

CO,H2

C

Ar

Tshell

cond

window

insulation

Thx

reactoroil cooling, Toil

Q

Q

Q

Q Q

Q

Q n

n

n

n

n

n

m

Figure 6.1: Diagram of the considered process and it’s energy and mass flows usedin the simulation.

Formulation of steady state energy conservation for the cavity results then in:

0 = Qsolar − Qrerad − Qcond +∑

species

ni,in · hi (Tin)−∑

species

ni,out · hi (Tcavity) (6.1)

The enthalpy difference described by the last two terms of eq. (6.1) can also be

expressed as the sum of the power required to heat the feedstock from Tin to Tcavity,

Qheating, and the enthalpy change due to the chemical reaction, Qchem:

0 = Qsolar − Qrerad − Qcond − Qheating − Qchem (6.2)

where the conduction heat losses are calculated with the overall heat transfer coef-

ficient U presented in 5.1.2:

Qcond = Scone · Ucone · (Tcavity − Toil) + Scavity · Ucavity · (Tcavity − T∞) (6.3)

6.1. Governing Equations 69

and Ucone = 65 and Ucavity = 22 W/(m2K) are overall heat transfer coefficients

relative to Toil and T∞, respectively. Values given are averages over Scone and Scavity,

the inner surface of the reactor’s front cone and the surface of the rest of the cavity,

respectively.

The power lost by radiation Qrerad is given by:

Qrerad = Qsolar ·Rw (Tsun)

+ Saperture · σT 4cavity · Trw (Tcavity)

+Qsolar · Aw (Tsun) + Saperture · σT 4

cavity · Aw (Tcavity)

2(6.4)

where the first therm accounts for reflected incoming radiation, the second term

for radiation reradiated from the cavity and the third term for emission from the

window. The total window absorptance Aw, reflectance Rw and transmittance Trw

are calculated from the spectral values presented in eqs (5.3)–(5.5) as:

Aw, Rw,Trw =

∫∞0

Eλb (T ) · Aλ,w, Rλ,w,Trλ,w dλ∫∞0

Eλb (T ) dλ(6.5)

where Eλb (T ) is the Plank’s blackbody spectral emissive power at Tsun = 5780 K or

Tcavity depending on the origin of the radiation. The factor 1/2 in the third therm of

eq. (6.4) takes into account emission from both sides of the window. Note also that

the radiative flux incident on the window from the inside of the cavity, used in the

second and third term of eq. (6.4), is approximated simply by σT 4cavity. In fact the

apparent absorptions of the cavity is close to one and furthermore the radiation is

mostly absorbed/emitted by the particles. This is in contrast the method presented

previously in [123] where the radiosity method was used to determine incident fluxes

on the window.

The heat absorbed by the chemical reaction Qchem reads:

Qchem = XC · nC ·∆HR (Tcavity) (6.6)

where the enthalpy of the reaction ∆HR is evaluated at the temperature of the

cavity. The gas composition at the outlet and the carbon conversion are found by

numerical integration of the system of differential equations given in eqs (5.18)–

(5.22), assuming a plug flow reactor. Similarly to Section 3.4.4 the fluid velocity is


given by u (t) =(∑

species ni (t) RT)

/ (pS) and the integration is stopped at t = τ ,

where the mean residence time τ is given by the implicit relation L =∫ τ

0u (t) dt,

and L is the length of the reactor. The sensible heat that can be recovered from the

product gases is then defined as:

Qsensible =∑

species

∫ Tcavity

Thx

ni · cp,i(T ) dT (6.7)

and the heat used to produce the steam reads:

Qsteam = nH2O ·(

(373.15 K− T∞) cp,H2O(l)+ ∆Hvap +

∫ Tsteam

373.15 K

cp,H2O(g)(T ) dT

)(6.8)

where ∆Hvap is the evaporation enthalpy for water.

Finally, eq. (6.2), which is implicit in Tcavity, is iteratively solved using the Nelder-

Mead Simplex method [69]. The temperature of the outer surface of the Inconel shell

Tshell, is calculated as:

Tshell = Tcavity −Ucavity

Ushell

· (Tcavity − T∞) (6.9)

where the overall heat transfer coefficient between cavity wall and outer Inconel shell

Ushell = 108 W/(m2K) is calculated assuming 1D cylindrical conduction:

Ushell =1

Dcavity ·∑

layerslog(Di+1/Di)

ki

(6.10)

where ki and Di describe the thermal conductivity and the diameter of the different

material layers involved, respectively.

6.2 Results

Two parameter studies were performed for the baseline simulation parameters listed

in Table 6.1: (1) the optimal volume and coke feeding rates were identified for a

reactor operating at 5 kW and with a 50 mm in diameter aperture, (2) the potential

of up-scaling the ideal configuration found in (1) was investigated up to a nominal

power of 1 MW.

Figure 6.2 shows the results for parameter study (1): the carbon conversion

6.2. Results 71

Table 6.1: Baseline model parameters for the studies conducted for the 5 kW labscale reactor and it’s scale-up.

Parameter study: lab scale scale-up

Aperture diameter Daperture, m 0.05 0.08–0.40Cavity volume Vcavity, dm3 0.002–2.2 5.7–794

Feeding rates mcoke, g/min 0.6–20 31–1986mH2O, g/min 1.3–53 82–5257

VAr, ln/min 2–66 0.0Overall stoichiometric ratio H2O/C, – 2.0 2.0

Solar power input Qsolar, kW 5 40–1000

Static pressure p0, Pa 105

Inlet equivalent particle diameter D30,in, µm 2.21D32,in, µm 3.13

Overall heat transfer coefficient U , W/(m2K) 22/65 10Inlet temperature Tin, K 423.15Temperature after the HX Thx, K 473.15Oil temperature Toil, K 393.15 –Surroundings temperature T∞, K 293.15

at the exit of the reactor (a), defined in eq. (3.5), the solar-to-chemical efficiency

(c), defined in eq. (3.7), and the temperatures of the cavity (b) and of the Inconel

shell (d) as a function of the coke feeding rate and the reactor volume. A lower

petcoke feeding rate leads to a higher chemical conversion due to longer residence

times, but at the expense of a lower net solar energy absorbed. In contrast, a higher

coke feeding rates lead to a larger portion of energy used for heating the feedstock

but not converted into chemical energy. A similar effect has an increase in the

reactor volume, resulting in longer residence times and higher chemical conversion,

but at the expense of higher conduction losses. Optimal parameters determined for

maximum energy conversion efficiency were mcoke = 7.5 g/min and Vreactor = 0.54

dm3. Unfortunately, these parameters resulted in very high temperatures for the

reactor liner Tcavity=1771 K as well as for the Inconel shell Tshell=1418 K. In order

to reduce the thermal load of the reactor components, the prototype reactor was,

therefore, designed with a cavity volume of 1.6 dm3 resulting in a slight decrease in

maximal ηchem from 20% to 17%.

The effect of scaling-up the reactor is elucidated in Fig. 6.3, where ηchem and


0

4

8

12

16XC

mcoke,g/m

in

0.1

0.25

0.5

0.9

1

(a)

Tcavity, K

1400

1500

1600

18002000

22002400

(b)

0 0.5 1 1.5 20

4

8

12

16 ηchem

Vcavity, dm3

mcoke,g/

min

0.05 0.05

0.1

0.14

0.170.19

0.140.1

0.05

(c)

0 0.5 1 1.5 2

Tshell, K

Vcavity, dm3

1200

13001400

1600

1800

(d)

Figure 6.2: Carbon conversion (a), cavity (b) and shell (d) temperatures and chem-ical efficiency (c) as a function of the carbon feeding rate and the reactor volume forthe 5 kW prototype reactor and the baseline parameter listed in Table 6.1.

ηproc together with the related Tcavity and XC are plotted as a function of Qsolar and

mcoke, using as baseline the optimum operating conditions found in Fig. 6.2. The

aperture diameter and the volume were adjusted to keep a constant Qsolar/Aaperture

ratio (average flux of 2300 suns) and constant Qsolar/V2/3cavity, respectively. Further,

the oil cooling of the reactor front is omitted and an improved ceramic insulation is

assumed (U = 10, W/ (m2K)). There is a remarkable positive effect of scaling-up the

reactor as a result of the relatively lower conduction losses through a smaller area-

to-volume ratio. For example, for an optimum petcoke feeding rate, the predicted

ηchem of a 40, 200, and 1000 kW reactor are 12%, 32%, and 35%, respectively,

whereas the predicted ηproc are 49%, 73%, and 74%, respectively. The locations of

maximal efficiency lie on straight lines of constant mcoke/Qsolar in the range 0.13–0.15

6.2. Results 73

0

20

40

60

80

100

120

XC

mcoke,kg

/h

0.00

1 0.1

0.5

0.9

1

(a)

Tcavity, K

1000

1200

1400

1600

1800

2000

2200

2400

(b)

200 400 600 800 10000

20

40

60

80

100

120

ηchem

Qsolar, kW

mcoke,kg

/h

0.01

0.1

0.2

0.3

0.34

0.34

0.3

0.2

0.1

(c)

200 400 600 800 1000

ηproc

Qsolar, kW

0.2 0.4 0.

6 0.7 0.

73

0.73 0.7

0.6

0.4

0.2

(d)

Figure 6.3: Predicted carbon conversion (a), cavity temperature (b), chemical effi-ciency (c) and process efficiency (d) as a function of the carbon feeding rate and thesolar power for a scaled-up reactor. The baseline parameter are listed in Table 6.1.

kg/ (hr · kW) with typical related XC in the range 0.9–1.0 and Tcavity in the range

1700–1900 K.

Finally, the simulation model was run for the operational parameters of each

experimental run listed in Tables 4.1 and 4.2 for Campaign 1 and Campaign 2, re-

spectively. Figure 6.4 shows numerically calculated and experimental measured data

for the (a) 23 solar experimental runs of Campaign 1; and (b) 20 solar experimental

runs of Campaign 2, ordered by increasing Qsolar. Shown are the cavity and shell

temperatures, the carbon conversion and the steam conversion. Full circles indicate

numerically calculated values; open circles indicate the experimentally measured

data. The error bars indicate the propagated inaccuracy of the input parameters,

namely ±913 % and ±5 % for the solar input power Qsolar and the feeding rate mcoke,


800

1200

1600

T,K

temperature, input power

Tcavity

Tshell

RMSTshell= 0.05

RMSTcavity= 0.016

4

4.5

5

5.5

6

Qsola

r,kW

(a)

0

0.2

0.4

0.6

0.8

1carbon conversion

XC

RMS = 0.31

1 5 9 13 17 210

0.2

0.4

0.6

0.8

1steam conversion

XH

2O

RMS = 0.47

experiment #

800

1200

1600

T,K


Tcavity

Tshell

RMSTshell= 0.045

RMSTcavity= 0.018

4.2

4.3

4.4

4.5

Qsola

r,kW

(b)

0

0.2

0.4

0.6

0.8

1carbon conversion

XC

RMS = 0.63

1 5 9 13 170

0.2

0.4

0.6

0.8

1steam conversion

XH

2O

measuredcalculated

RMS = 1.3

experiment #

Figure 6.4: Numerically calculated (lumped parameters model) and experimentalmeasured data for the (a) 23 solar experimental runs of Campaign 1; and (b) 20solar experimental runs of Campaign 2, ordered by increasing Qsolar. Shown are theaverage cavity and wall temperatures, the carbon conversion and, the chemical andprocess efficiencies. Full circles indicate numerically calculated values; open circlesindicate the experimentally measured data; the error bars indicate the propagatedinaccuracy of the input parameters.

respectively (see Section 3.3). The simulation was able to successfully predict the

reactor temperatures. In fact, relative RMS errors below 5% were found for the

shell temperatures Tshell and the cavity temperatures Tcavity and for both campaigns.

6.2. Results 75

In Campaign 1, most predictions for XC and XH2O lie inside the accuracy bounds.

Nevertheless, the considerable width of these accuracy bounds shows the fuzziness

of the results. For Campaign 2, finally, the simulation fails to predict the chemical

conversion, especially for H2O, the RMS being 130%. In fact, the simple model used

does not consider the dependency between particle size and radiation absorption ef-

ficiency, thus overestimating the performance of experimental runs performed with

poorly absorbing particles (experimental Campaign 2 with feedstock type 3).

Chapter 7

Heat and Mass Transfer in the

Reactor Cavity1

A model for heat and mass transfer inside the cavity of the reactor presented in

Section 3.1 was developed. It solves for radiative, conductive, and convective heat

transfer, fluid flow and reaction kinetics. The model is implemented as two separate

solver modules: (1) a finite volume solver (CFD) that solves for coupled radiative,

conductive and convective heat transfer, fluid flow and reaction kinetics and (2) a

Monte-Carlo (MC) raytracer that provides the source term for radiation used in (1).

The intrinsic statistical variance induced by the MC solver is reduced by means of

Gaussian kernel smoothing and adaptive underrelaxation when coupled to the finite

volume solver. The main boundary conditions were previously presented in Chapter

5 as well as the treatment of the walls and the window.

The two-phase medium considered is composed by a solid particulate polydis-

persion suspended in a gas mixture. The particle diameters of the solid phase are in

the range 1–100 µm. The particle suspension is modeled as a non gray absorbing,

emitting, and scattering participating medium subjected to concentrated thermal

radiation. Typical radiative equilibrium temperatures are in the range 1500–1800

K. Both solid and gas phases are involved in the heterogeneous chemical reaction

taking place predominantly on the outer surface of the solid particles and, to some

extent, on the inner porous surface of the particles. As the gasification reaction

1Material from this chapter has been submitted for publication as: A. Z’Graggen, and A.Steinfeld. Heat and Mass Transfer Analysis of a Suspension of Reacting Particles subjected toConcentrated Solar Radiation — Application to the Steam-Gasification of Carbonaceous Materials.International Journal of Heat & Mass Transfer, 2007.

78 7. Heat and Mass Transfer in the Reactor Cavity

progresses, the solid particles shrink and their thermal properties vary, as well as

the optical properties of the particle suspension.

The scheme developed here is able to deal with concentrated thermal radiation

input, polydisperse particulate media with spectral and directional optical proper-

ties, and temperature-dependent chemical kinetics and fluid properties. Validation

is accomplished by comparing numerical and experimental results.

7.1 Heat Transfer Modes in the Polydisperse Par-

ticle Suspension

Typical relaxation times are compared for assessing the relative importance of the

different modes of heat transfer, namely conduction, convection, and radiation. The

reference time scale for the fluid flow, τflow = L/u , equals 0.1 s. For conductive heat

transfer inside the particles, the relaxation time of the core temperature of a sphere

with respect to a sudden change in its surface temperature is given by [44]:

τcond =ρs · cp,s ·D2

4 · ks

(7.1)

Values 10 to 105 times smaller than τflow are found for particles of 100 to 1 µm,

respectively. Uniform temperature is therefore assumed for a single particle. The

relaxation time for convective heat transfer of the temperature of a sphere submerged

in a fluid at a different temperature reads:

τconv =ρs · cp,s ·D

6 · h(7.2)

Particles are assumed to be entrained by the gas flow, as justified by Stokes numbers

Stk< 10−2, thus the Nusselt number for a sphere is Nu=2 [44] and eq. (7.2) simplifies

to:

τconv =ρs · cp,s ·D2

12 · kg

(7.3)

The resulting τconv is comparable to τflow for 100 µm particles. The energy conserva-

tion equation is therefore solved for each phase separately. Finally, the temperature

T of a particle homogenously irradiated is described as a function of time by:

πD3ρscp,s

6· ∂T

∂t= πD2σε

(T 4∞ − T 4

)(7.4)


where T∞ is the temperature of the surroundings. Linearization of T 4 around a

typical particle temperature T0 leads to:

T 4 = T 40 + 4T 3

0 T (7.5)

The relaxation time of the temperature of a particle homogenously irradiated is

then:

τrad =ρs · cp,s ·D

24 · ε · σ · T 30

(7.6)

Calculated τrad are 1 to 3 orders of magnitude smaller than τflow. The temperature

field is therefore mainly influenced by the local divergence of the radiative flux and

only marginally by the heat transported by advection.

7.2 Governing Equations

The solid particles are assumed to be entrained in the gas phase, as justified by

the Stokes number in the range of 10−2–10−6. Conduction and diffusion in the flow

direction are neglected, as justified by the Peclet numbers for conduction (Pe =

uL/α) and for diffusion (Pe = uL/Dc) in the order of 102. Steady-state mass

conservation for the mixture is then expressed by:

∇ · (ρ · u) = 0 (7.7)

and for a single species i by:

∇ · (ρYi · u) = Ri (7.8)

where Ri denotes the volumetric rate of production/consumption of species i as

defined in Section 5.3. Steady-state energy conservation for the solid phase yields:

∇ · (ρshs · u) = φrad − φconv − φchem (7.9)

where the radiation source φrad is given by the negative divergence of the radiative

flux −∇ · qr , the convection source φconv is given by the convective heat exchange

between the two phases, and the chemistry source φchem account for enthalpy carried

by mass changing its phase. Steady-state energy conservation for the gas phase

yields,

∇ · (ρghg · u) = φconv + φchem (7.10)


The mean absorption coefficient of a typical gas composition (CO, CO2, H2, H2O)

at 1 bar, calculated line-by-line using the HITEMP/HITRAN database [83], is ap-

prox. 0.06 and 0.006 m−1 for radiation emitted at 1800 and 5780 K, respectively.

Absorption by the gas is therefore neglected vis-a-vis absorption by the particles

(see Appendix C for details).

Energy source for convection Convective heat transfer between the polydis-

perse solid and gas phases is described by Newton’s law of cooling [44] as the sum

of each particle’s contribution:

φconv =

∑Np

h · πD2

V(Ts − Tg) (7.11)

further using h = Nu · kg/D and fV = Npπ

6V

∫∞0

fD3 dD leads to:

φconv =6 · fV · Nu · kg

∫∞0

fD dD∫∞0

fD3 dD(Ts − Tg) (7.12)

Finally, the Nusselt number for a stationary sphere is equal to 2 [44] and the defini-

tion of equivalent diameters (eq. (5.10)) is used:

φconv =12 · fV · kg

D231

· (Ts − Tg) (7.13)

Energy source for chemistry The mass consumed, generated, or transformed

during the chemical reaction contributes to the enthalpy change, evaluated at the

temperature of the originating phase:

φchem =n∑

i=1

hi (T ) ·Ri (Ts, Yi, f) (7.14)

Note that the rate of reaction Ri depends on the temperature of the solid phase,

Ts, the gas composition, and the particle size distribution f . At the high tem-

peratures considered, only two reactions are of importance, namely gasification

C + H2O → H2 + CO and water-gas-shift CO + H2O → H2 + CO2. Equation (7.14)

can then be explicitly written as:


φchem = hH2O (Tg) ·RH2O (7.15)

+hH2 (Ts) ·RH2

+hCO (Ts) · (RCO −RCO2) + hCO (Tg) ·RCO2

+hCO2 (Tg) ·RCO2

where the reaction rates Ri are calculated with equations (5.18)–(5.22). Note that

the temperature field is related to the enthalpy by the implicit equation:

h (T ) = href +

∫ T

Tref

cp (T ) dT (7.16)

where href is the reference enthalpy at Tref = 273 K, and temperature-dependent

values of cp for all species are calculated with the empirical correlations given in

[88].

Energy source for radiation The divergence of the radiative flux is obtained

from the difference between emitted and absorbed radiation

∇ · qr (s) = 4π

∫ ∞

λ=0

κλ (s)

(Iλb (s)− 1

4π

∫ 4π

ω=0

Iλ (s, ω) dω

)dλ (7.17)

Using the Planck mean absorption coefficient κP eq. (7.17) can be simplified to:

∇ · qr (s) = 4κPσT 4 −∫ ∞

λ=0

κλ (s)

∫ 4π

ω=0

Iλ (s, ω) dωdλ (7.18)

where the first term accounts for emission from the volume and the second term for

absorption in the volume. The latter is further referred to as φrad,a. The radiative

intensity Iλ, required in the second term, is given by the equation for radiative

transfer for a participating medium

dIλ

ds= −βλ (s) Iλ (s) + κλ(s)Iλb +

σsλ (s)

4π

∫ 4π

ωic=0

Iλ (s, ωic) Φλ (ω, ωic) dωic (7.19)

solved by Monte-Carlo raytracing (see Section 7.3.1). Note that the solution of

eq. (7.19) in a bounded simulation domain also indirectly yields values for the radia-

tive power absorbed on the walls∫∞

0qic,λελ dλ and by the window

∫∞0

qic,λAw,λ dλ,

as introduced in eqs (5.1) and (5.6), respectively.


7.3 Numerical Implementation

selective quartz window:- absorbing, emitting and transmitting radiation- conduction losses

ceramic insulation:- absorbing, emitting and reflecting radiation (diffuse-gray)- conduction losses

two-phase medium:- polydisperse- chemically reacting- absorbing, emitting and scattering

concentrated solar power:- solar spectrum- angular distribution- radial distribution

solid and gaseous feedstock at Tin- initial composition- initial velocity

productoutlet

x

r

1

5432

cylindricalcompartments

Figure 7.1: Axis-symmetric model domain, featuring five concentric cylindrical com-partments. Indicated are also the boundary and inlet conditions. The grid spacing inx-direction is adapted to the expected temperature gradients, shown is an examplefor the Monte-Carlo solver.

The simulation domain and the main boundary conditions are shown in Fig. 7.1.

Preliminary CFD simulations showed an axis-symmetric flow pattern with negligible

radial velocity components. However, in order to account for the radial dependency

of the incoming solar radiation (see Fig. 5.2), the domain is subdivided into con-

centric cylindrical compartments, as indicated in Figs 7.1 and 7.2, where a typical

finite volume is described. As shown is Section 7.1 advective heat transport is of sec-

ondary importance. Thus, a two-phase fluid flow with neglected angular and radial

components of the velocity vector is assumed. This simplification further enables

to better elucidate the physical phenomena involved in the interaction of radiation

with the chemical reacting flow. The homogenous grid spacing in the radial di-

rection (number of compartments) is equal for both, the MC and the CFD solver.

In contrast, the grid spacing in the x-direction is adapted to the requirements of

each solver separately. Each volume element is considered isothermal (each phase

separately), with homogenously distributed species and particle size fractions. The

following boundary/initial conditions and material properties are set:

7.3. Numerical Implementation 83

x

r

w e

advection

radiation

compartment j

compartment j 1

compartment j+1

u, ½, Yi, h

D, T, fVPSj

Sj+1

Sj1

¢x

Figure 7.2: Typical finite volume considered in the simulation with faces ‘w’ and ‘e’,centerpoint ‘P’, finite length ∆x, and cross-section Sj. Also shown are the respectiveface-centered and cell-centered variables. Advection is solved in x-direction for eachcompartment j separately, whereas radiative exchange is considered along both, thex and the r directions.

• Concentrated solar radiation The angular and radial distribution of the in-

coming radiation were presented in Section 5.1.1. The spectral distribution is

approximated by Planck’s blackbody emission at 5780 K.

• Quartz window Spectral values for the transmittance, absorptance and re-

flectance of the window, as well as the overall heat transfer coefficient for

conduction/convection losses were given in Section 5.1.3. In contrast to the

experimental reactor, where the window is mounted on a diverging frustum

away from the aperture, in the model the window is located at the aperture

for two reasons: (1) the radiation BC is given at the aperture and (2) the com-

putational domain is reduced. This simplification does not affect the results

since no radiation-particle interactions occur in the diverging frustum.

• Reactor walls Overall heat transfer coefficients for conduction in the walls as

well as radiative properties were given in section 5.1.2. The walls are considered

diffuse gray (ε = 0.8) due to particle deposition (visual observations). The

impact of the uncertainty of ε on the results was assessed by sensitivity analysis


(see Section 7.4).

• Feedstock injection The inlet axial velocity is assumed to be uniformly dis-

tributed over the cross section, as justified by preliminary CFD simulations

and visual observation. Further assuming ideal gas, the initial velocity is given

by:

uin =(nH2O + nAr + nH2,feedstock) ·R · Tin

p0 · S(7.20)

where p0 is the ambient pressure, S is the cavity cross-sectional area and

nH2,feedstock is the hydrogen bound in the feedstock released by fast pyrolysis

at the entrance of the reactor.

• Solid particles Petrozuata delayed coke was used as feedstock (Chapter 2).

Particles size distribution and radiative properties were given in Section 5.2.

Equivalent monodisperse diameters were used to describe the changing prop-

erties of the polydisperse medium as a consequence of the chemical reaction.

Reaction kinetics were presented in Section 5.3.

7.3.1 Monte-Carlo Ray Tracer

An overview of common approaches used to deal with radiative heat transfer in

participating media is given in [43] and [96]. The Monte-Carlo (MC) method used

in this study was chosen for it’s simplicity and flexibility. It permits to virtually

model any complex physical phenomenon that affects radiative transfer, including

non-isothermal, non-gray and anisotropically-scattering media. Furthermore, com-

plex geometries, coupled convection and conduction heat transfer as well as spa-

tially varying medium properties can be easily implemented. Previous pertinent

MC modeling studies include pulverized coal furnaces [33] and solar reactors for

coal gasification [60, 110] and CH4 decomposition [40].

The source term for radiation absorbed in the volume, φrad,a (second term of

eq. (7.18) describing the divergence of the radiative flux), as well as the radiation

absorbed by the walls qic is calculated with a statistical simulation in which the his-

tory of a large number of rays either emitted in the volume and on the boundaries

or incident through the window is tracked. Emission wavelength and direction, as

well as absorption and scattering of these rays is determined by means of probabil-

ity density functions (PDF) based on local radiative properties and using random


select ray starting point

compute penetration

length: l(βλ,s) or ln(βλ,s)

compute distance to next

face: Lf

hit

inside element?

l<Lfyesno

compute absorption

by the gas phase:

Ea,g(κλ,g,l)

update qray

select ray direction: r

update qray

face is

boundary?

no

yes

ray starting

point

=

intersection

with face

compute absoption

at the boundary

update

dimensionless

penetration

length

update qray

ray starting

point

=

intersection

with

boundary

compute

reflection

direction

ray starting

point

=

location of

particle

interaction

qray>qmin

yes

qray>qmin

yes

qray>qminyes no

compute absorption

by the solid phase:

Ea,s(ωλ)

update qray

qray>qmin yes

compute

scattering

direction rs

new ray: qray, λ

compute absorption

by the gas phase:

Ea,g(κλ,g,l)

Figure 7.3: Flowchart of the Monte-Carlo raytracer used the calculate the radiativesource term in a domain composed of finite volumes. Note that the volumes aredelimited by either interfacial faces or/and domain boundaries. Computation stepsthat involve one or several random numbers are show by the bold boxes.


numbers R drawn from a uniformly distributed set between 0 and 1. The sum of

all ray’s interactions with the two phase medium and the walls yields then a spatial

distribution of absorbed radiation:

qic =

∑rays Ea,wall

S(7.21)

φrad,a =

∑rays Ea,volume

V(7.22)

where Ea,wall and Ea,volume describe each ray’s contribution to the power absorbed

on a wall element dS and in a volume element dV , respectively. The simulation

domain shown in Fig. 7.1 is dived in hexahedral elements; due to the rotational

symmetry of both the flow field and the radiative BC only a slice of the reactor

is considered, while the cutting planes are modeled as perfect reflecting mirrors.

Figure 7.3 shows the flowchart of the Monte-Carlo approach used [113, 24, 119].

Rays are generated by emission from the walls and the volumes (gas and solid phase)

as well as by transmission through the window (incident qsolar). Each ray is described

by its power qray, wavelength λ, starting point and direction r. A ray traveling in

a volume dV can undergo four different interactions: (1) continuous absorption in

the gas phase, (2) local absorption and scattering in the solid phase, (3) absorption,

reflection and transmission on a wall element dS and (4) transmission to the adjacent

volume element. For each absorption interaction the ray’s power is diminished by

the amount Ea related to local radiative properties. The ray history is terminated

when the residual power qray = q0ray −

∑Ea is smaller than the cutoff power qmin or

when it leaves the reactor through the window. Note that the treatment of radiation

absorption in the gas phase is presented for completeness, although it is neglected

in the simulation results presented later on. The equations describing the behavior

of the above mentioned interactions are divided into three groups.

Interaction with a wall The total power of the rays emitted from a wall element

is given by:

E = S · ε (T ) σT 4wall (7.23)

where ε is the total emissivity of the surface. The ray starting points are homoge-

nously distributed on the surface element dS while the direction of the emitted rays

is given by the azimuthal angle θ and the circumferential angle φ related to the face


normal n [91] (see also Fig. 7.4):

φ = 2 · π ·Rφ (7.24)

Rθ =2π∫ θ

0

( ∫∞0

ελ (λ, θ′) Iλb (λ, Twall) sin θ′ cos θ′ dλ)

dθ′

εσT 4wall

(7.25)

where ελ (λ, θ) is the spectral directional emissivity of the surface. The wavelength

λ of the emitted ray is given by [91]:

Rλ =2π∫ λ

0

(Iλb (λ′, Twall)

∫ π/2

0ελ (λ′, θ) sin θ cos θ dθ

)dλ′

εσT 4wall

(7.26)

The power absorbed by the wall when hit by a ray is given by Ea = qray · ελ (λ, θ).

The direction of the reflected ray, which carries the remaining power qray − Ea, is

either calculated with eqs (7.24) and (7.25) for a diffuse surface or by rreflected =

r− 2 (r · n)n for a perfectly specular surface while λ does not change.

Interaction with a window The window is treated similarly to the walls. The

spectral directional emissivity ελ (λ, θ) is simply substituted for the spectral direc-

tional absorptance Aw (λ, θ) of the window when calculating emission and absorp-

tion. The amount of power reflected is calculated with the spectral directional

reflectance of the window Rw (λ, θ). In addition, radiation incident from the outside

of the domain is considered. The radial, directional and spectral distribution of the

incident solar power is given in Section 5.1.1. A part of the solar power is absorbed

and reflected (qsolar · Aw (Tsun) and qsolar · Rw (Tsun)). The remaining power enters

the cavity while the angular distribution shown in Fig. 5.3 is weighted by Trw (λ, θ)

calculated with eq. 5.3.

Interaction with a volume The gas and particulate phases are assumed perfectly

mixed and homogeneous over the whole volume. The total emitted power is given

by:

E = Eg + Es = 4V ·(κP,g(Tg) · σT 4

g + κP,s(Ts) · σT 4s

)(7.27)

where κP is the Planck mean absorption coefficient defined as∫∞

0κλEλb dλ/σT 4 and

the absorption coefficient κλ of the solid phase is given in Section 5.2. The emission


n

rµ

r

rs

µ

Á

Á

l

Lfboundary element:"¸(µ), Tw¸(µ)

volume element:·¸,g¯¸,s=·¸,s+¾s¸,s

emission fromthe wall ortransmission ofincident solarradiation:qray, ¸

transmission toadjacent element

intersection with theboundary if no scattering hadoccured

absorption andscattering by the solid phase

absorption bythe gas phase

Figure 7.4: Typical volume element used in the MC solver. A ray emitted from aboundary face (gray) with initial power qray and wavelength λ undergoing absorptionand scattering is shown (dotted arrow). The participating medium is described byκλ,g, κλ,s, and σsλ,g the absorption coefficients for the gas and solid phase, and thescattering coefficient of the solid phase.

direction is given by

φ = 2 · π ·Rφ (7.28)

θ = cos−1(1− 2 ·Rθ) (7.29)

while the wavelength is calculated as:

Rλ =1

κP · σT 4

∫ λ

0

κλ Eλb dλ′ (7.30)

Figure 7.4 shows a typical volume element considered in the MC simulation. A

ray traveling trough the volume is either absorbed by the gas phase or absorbed

and scattered by the solid phase. Absorption by the gas is assumed significantly

smaller than by the particles. The penetration length of a ray, i.e., the distance

until extinction by the solid phase occurs, l, is found from:

−∫ l

0

βλ,s (s) ds = ln Rl (7.31)


ray-particleinteraction

L1 L2 Ln

¯1 ¯2 ¯n

...

ray

ln

Figure 7.5: Schematic of a ray passing trough several elements of width Ln andextinction coefficient βn. Also shown is the residual penetration length ln in the lastelement.

where the extinction coefficient βλ,s = κλ,s+σsλ,s of the solid phase is given in Section

5.2. For a homogenous volume, eq. (7.31) simplifies to l = − ln (Rl) /βλ,s. The power

absorbed by the gas phase along distance l is then given by:

Ea,g = qray(1− e−κλ,g·l) (7.32)

while the power absorbed by the solid phase at a distance l from the emission point

is expressed as [113]:

Ea,s = qray(1− ωλ) (7.33)

where the scattering albedo ωλ is defined as ωλ = σsλ,s/ (κλ,s + σsλ,s). Finally, the

direction of the ray after scattering is given by:

Rθ =1

2

∫ θ

0

Φ (θ′) sin θ′ dθ′ (7.34)

φ = 2πRφ (7.35)

where θ is the angle between the forward direction of the incident ray r and the

direction of the scattered ray rs, φ is given on a plane perpendicular to r, and the

phase function Φ (θ) of the polydisperse solid phase is given in Section 5.2.

In a medium with a small extinction coefficient or when very small volumes are

considered a ray is likely to pass trough multiple volume elements before experiencing

the first interaction with the solid phase (see Fig. 7.5). In this case eq. (7.31) is solved

for piecewise constant properties. Assuming absorption in the nth element it reads


then:

L1β1 + L2β2 + · · ·+∫ l

∑n−1i=1 Li

βn (s) ds = − ln (Rl) (7.36)

The remaining path length in element n, ln = l −∑n−1

i=1 Li is then given by:

ln =− ln (Rl)− L1β1 − L2β2 − · · · − Ln−1βn−1

βn

(7.37)

where the numerator describes the dimensionless penetration length, − ln (Rl), de-

creased by the sum of the optical thicknesses of the traversed layers. Note that a new

dimensionless penetration length is calculated each time that a ray hits a particle

or a wall but not when it hits a specular symmetry boundary.

7.3.2 Fluid Flow Solver

The fluid flow is solved in a array of concentric cylindrical compartments as shown

in Figs 7.1 (5 compartments) and 7.2. Furthermore, the angular component of the

flow is neglected, thus eqs (7.7)–(7.10) — solved for each concentric comportment j

separately — are reduced to the sole axial dimension x. The resulting flow field is

of parabolic type (one way coordinate) [74] and is therefore solved directly volume-

by-volume starting from the inlet. Equations (7.7)–(7.10) are discretised into finite

volumes, taking advantage of the relation∫

V∇·F dV =

∮δV

Fn dS. The 1D volumes

are described by a cross-section S and finite length in the flow direction ∆x and are

delimited by the face ‘w’ at the inlet, and the face ‘e’ at the outlet. Mass conservation

of the mixture results then in (index j omitted for clarity):

ue =(ρu)w

ρe

(7.38)

while mass conservation for each species i yields:

(ρuYi)e − (ρuYi)w =1

S·∫

V

Ri dV = ∆x ·Ri (7.39)

A downwind scheme is used to estimate the average reaction rate Ri over the control

volume since the reaction rate has a negative exponential behavior. Best convergence

was found for Ri = 3/8 · Ri,w + 5/8 · Ri,e. The mass fraction of species i at face ‘e’


is then:

Yi,e =(ρuYi)w + ∆x(3

8Ri,w + 5

8Ri,e)

ρeue

(7.40)

Energy conservation for the solid and the gas phase are treated similarly:

he,s =(ρsuhs)w + ∆x

(φrad − φconv − φchem

)(ρsu)e

(7.41)

he,g =(ρguhg)w + ∆x

(φconv + φchem

)(ρgu)e

(7.42)

where the average source terms for radiation and convection are calculated by the

central scheme φrad = 0.5 (φrad,w + φrad,e) and φconv = 0.5 (φconv,w + φconv,e), while

the average source term for chemistry is calculated similarly to the average reaction

rate: φchem = 3/8 · φchem,w + 5/8 · φchem,e.

Equations (7.38)–(7.42) are solved iteratively for each finite volume consecutively.

The step length ∆x is adapted to ensure conversion after 8 iterations. Due to the

stiffness of the system of differential equations that describe the chemical reaction

(eqs (5.18)–(5.22)) typical finite volumes for the CFD solver are 10 times smaller

(∆x ≈ 0.1 mm) than those used by the Monte-Carlo raytracer. Furthermore, when

particles in the 1 µm range are considered, particle-gas convection becomes virtually

instantaneous (τconv τflow) and volumes that are up to 105 smaller than those used

by the Monte-Carlo raytracer are required to ensure convergence.

7.3.3 CFD-MC Coupling

Figure 7.6 shows the flowchart of the coupled simulation setup. The iterative process

is started with a guessed temperature and mass distribution inside the cavity. The

MC solver calculates then the source terms φrad,a (x) and qic (x) required by the

CFD solver. The latter solves for heat and mass conservation returning values for

Tg (x), Ts (x) D32 (x) and fV (x) to the MC solver. This procedure is repeated until

the relative change in the results between two consecutive iterations is smaller than

10−5.

Note that, because of the statistical noise, the time required to perform a single

iteration at a reasonable accuracy for the MC module is several orders of magnitude

larger than that for the CFD module. For example, a simulation with 10 com-

partments run on an Intel R© Pentium R© 4 CPU at 2.53 GHz requires 1.2 seconds


guess temperature and

mass distribution

solve radiative transfer

equation by MC raytracing

calculate divergence of

radiative flux in the volume

calculate incident radiative

power on the walls

re-calculate

temperature

distribution of the walls

re-calculate temperature and mass

distributions, and equivalent

diameters in the volume

apply smoothing filters and

adaptive underrelaxation

iterate

solve coupled energy and mass

conservation equations

Figure 7.6: Flowchart of the coupled solver. The bold box describes the MC ray-tracer described in Fig. 7.3.

for one iteration of the CFD module. The MC module of the same simulation —

solved on a grid composed by 1577 volume elements and 5663 faces — requires 4,

29, and 297 seconds for 104, 105, and 106 primary rays (each primary ray results

in approximately 150 follow-up rays until its power is completely absorbed) with a

related standard deviation for the radiative source term of 0.12, 0.04, and 0.014, re-

spectively. To some extent, this problem is alleviated by Gaussian kernel smoothing

with adaptive bandwidth. For the incident radiation on the walls, the local mean

estimator is expressed by [12]:

q′ic (x) =

∑ni=1 w (xi − x, hi) · qic,MC (xi)∑n

i=1 w (xi − x, hi) ·(7.43)

where qin,MC denotes the result of the MC module, w is the Gaussian kernel de-

fined as w (x, hi) = exp(−x2

2h2i

)/(hi

√2π)

, and the bandwidth hi is proportional

to the local grid spacing. The second term of the divergence of the radiative flux,

φrad,a is estimated analogously to eq. (7.43), where a two-dimensional kernel of the


0 100 200 30010

–5

10–4

10–3

10–2

10–1

100

RM

Srel;

γ

0 100 200 3000

0.2

0.4

0.6

0.8

1

RMSrel

γemitted power

i0

i

Pem

it/m

ax

(Pem

it)

Figure 7.7: Relative RMS error for the temperature and underrelaxation parameter γ(left axis) and total emitted power, scaled by its maximal value, used as an indicatorfor system convergence (right axis) as function of the iteration number i. Values forrun #12 of Campaign 1 with ξ = 32.

form w (xi − x, hx) · w (ri − r, hr) is used with bandwidths hx and hr proportional

to the local grid spacing in the x and r direction, respectively [84]. Gaussian kernel

smoothing is not intrinsically conservative, the resulting radiative source terms q′ic

and φ′rad,a are therefore corrected:

qic = q′ic ·∫

wallsqic,MC∫

wallsq′ic

(7.44)

φrad,a = φ′rad,a ·∫

volumeφrad,a,MC∫

volumeφ′rad,a

(7.45)

An adaptive underrelaxation scheme for the radiative source term is introduced to

ensure convergence and reduce the overall computational time:

φi =(1− γi

)φi−1 + γiφi

MC (7.46)

where i is the iteration step, φiMC denotes the result of the MC module (either the

source term in the volume or on the walls), and γ is the adaptive underrelaxation

factor. Initially, γ is set to γ0 = 1 and the radiative source term is calculated


with low accuracy but quickly. The total power emitted from the volumes and

the walls by the MC solver Pemit =∑

volumes E +∑

walls E is used to monitor the

system convergence. Once Pemit as a function of the iteration number i approaches

a constant value, convergence is reached and γ is subsequently reduced according

to:

γi = γ0 exp

(i0 − i

ξ

)(7.47)

where i0 is the iteration step for the system convergence and ξ is a positive real

number. Figure 7.7 shows the development of the relative RMS error for the particle

temperature:

RMSrel =

Nvolumes∑j

(T i+1

j − T ij

T i+1j

)2/Nvolumes

0.5

(7.48)

A MC simulation with only 4 · 104 primary rays results in convergence after 150

iterations, yielding a relative standard deviation of 0.11 and 0.07 for the raw and

the smoothed φirad,a, respectively. Significant increase in accuracy is accomplished by

reduction of γ beyond iteration step i0 (ξ = 32) as observed in the reduction of the

RMSrel and in the flattening out of Pemit. Alternative variance reduction techniques

used in computer graphics, as for example Quasi Monte Carlo Sampling [67, 56]

and Tiled Blue Noise Samples [38], were applied but did not enhance the simulation

results for radiative transfer calculations.

Two distinct grid convergence studies were performed to asses the discretization

error. Operational parameters of a typical run from experimental Campaign 1 were

used. In the first study grids composed by 8 concentric compartments (radial direc-

tion) and 10, 20, 40, 80 and 160 elements along the axial direction were considered.

In the second study grids composed by 2, 4, 8, 16 and 32 concentric compartments

and 80 elements in the axial direction were considered. Figure 7.8 shows relative

errors for XC and XH2O as a function of subsequent grid refinement in axial direction

εx and in radial direction εr defined as:

ε =|Xi −Xi,finest|

Xi,finest

(7.49)

were Xfinest describes the result for 160 axial elements and 32 radial compartment

for studies one and two, respectively. Discretization errors smaller than 10−3 are

7.4. Results and Validation 95

1 2 4 8

104

103

102

grid refinement factor

ε

XC

XH2O

εxεr

Figure 7.8: Relative grid convergence errors in radial direction, εr, and in axialdirection, εx, for carbon and steam conversions at the outlet as a function of thegrid refinement factor. The grid refinement factor 1, 2, 4 and 8 corresponds to 10,20, 40, 80 elements in x-direction and to 2, 4, 8 and 16 elements in r-direction.

obtained for grids with more than 40 elements in axial and 10 compartments in

radial direction. A grid convergence index (CGI) of 0.07% (Fs = 3) and an order

of convergence p ≈ 2 is found for the axial direction [80]. In order to account

for differing operational conditions (in particular with respect to the absorption

coefficient) the minimal grid size in axial direction is doubled to 80 elements.

7.4 Results and Validation

Sets of 23 and 29 solar experimental runs were carried out for Campaigns 1 and 2,

respectively (Chapter 4). The 9 experiments performed with slurry made of feed-

stock types 2 and 4 for Campaign 2 are not considered here, because, due to the

occasional clogging of the feeding system, no reliable values for mcoke were obtained.

Numerically simulated and experimentally measured cavity and reactor shell tem-

peratures, and coke/steam conversions at the reactor outlet are shown in Figs 7.9

(a) and (b) for Campaign 1 and Campaign 2, respectively. Tcavity for the experi-

mental runs is the mean between the pyrometer reading corrected for the windows’

transmittance and the thermocouple reading Tshell corrected for conduction through


800

1200

1600

2000

T,K


Tcavity

Tshell

RMSTshell= 0.056

4

4.5

5

5.5

6

Qsola

r,kW

(a)

0

0.2

0.4

0.6

0.8

1

XC

carbon conversion

RMS = 0.2

1 5 9 13 17 210

0.2

0.4

0.6

0.8

1

experiment #

XH

2O

steam conversion

RMS = 0.18

1000

1200

1400

1600

1800

T,K


Tcavity

Tshell

RMSTshell= 0.013

4.2

4.3

4.4

4.5

Qsola

r,kW

(b)

0

0.2

0.4

0.6

0.8

1

XC

carbon conversion

RMS = 0.25

1 5 9 13 170

0.2

0.4

0.6

0.8

1

experiment #

XH

2O

steam conversion

measuredcalculated

RMS = 0.39

Figure 7.9: Numerically calculated and experimental measured data for the (a) 23solar experimental runs of Campaign 1; and (b) 20 solar experimental runs of Cam-paign 2, ordered by increasing Qsolar. Shown are the average cavity and wall tem-peratures, and the carbon and steam conversions at the outlet. Full circles indicatenumerically calculated values; open circles indicate the experimentally measureddata; the error bars indicate the propagated inaccuracy of the input parameters.

the Al2O3 liner (eq. (6.9)). Tcavity for the numerical simulation is the weighted mean

over all finite volumes of the cavity, Tcavity =∑

TiViρi

/∑Viρi, while the lower and

upper error bars describe the 25th and 75th percentile of Ti, respectively. Steam and


coke conversions are determined from the oxygen and carbon mass balance (Section

3.4.2). The error bars for the calculated values of XC, XH2O and Tshell are due to

the propagated error of the input parameters, namely ±5% for the coke feeding rate

mcoke and ±9%13% for the solar power input Qsolar (Section 3.3). The relative RMS

errors between measured and calculated values are indicated (assuming unbiased

input parameters). RMS errors were 5.6%, 20% and 18% for the shell temperature

and for steam and carbon conversion in Campaign 1, and 5.4%, 25% and 39% in

Campaign 2. However, overlapping of the accuracy intervals (error bars) of cal-

culated and measured data points is observed for the majority of the experiments

(error bars for ±5% accuracy for the measured XC and XH2O not shown in figure).

Finally, calculated temperature ranges in the cavity volume, Tcavity, are found to

be significantly higher than those experimentally estimated (based on the measured

wall temperatures), especially for Campaign 1, in which peak temperatures are ex-

pected to occur at the center of the cavity due to strong radiation absorption by the

small particles.

Table 7.1 lists the selected baseline model parameters of two typical solar ex-

Table 7.1: Baseline model parameters for two representative solar experimental runsfor Campaigns 1 and 2.

Experimental campaign # 1 2

Molar inlet flow rates nC, mol/min 0.254 0.102nH2

2, mol/min 0.142 0.057nH2O, mol/min 0.446 0.264nAr, mol/min 0.266 0.165

Overall H2O/C ratio - 1.76 2.6

Inlet velocity uin, m/s 0.065 0.088

Solar power input Qsolar, kW 5.7 4.7

Static pressure p0, Pa 105

Inlet equivalent particle diameter D30,in, µm 2.21 17.58D32,in, µm 3.13 47.14

Reactor cross sectional area S, m2 8 · 10−3

Overall heat transfer coefficient U , W/(m2K) 12–180Inlet temperature Tin, K 423.15 1003.5Surroundings temperature T∞, K 293.152Bound in feedstock


104

103

102

101

500

1000

1500

2000

x, m

T,K

0

50

100

150

200

250φ

rad,M

W/m

3

φrad

Ts

Tg

r = 0.025 m

r = 0.0 m

(a)

104

103

102

101

500

1000

1500

2000

x, m

T,K

0

1

2

3

4

5

φrad,M

W/m

3

φrad

Ts

Tg

r = 0.025 m

r = 0.0 m

(b)

Figure 7.10: Numerically calculated temperatures profiles for the gaseous and solidphases and the radiative source term along the reactor at two radial positions: center(r = 0.0 m) and close to the wall (r = 0.025 m). The baseline parameters listed in7.1 have been employed for (a) Campaign 1; and (b) Campaign 2.

perimental runs for each experimental campaign presented in Sections 4.1 and 4.2.

Figures 7.10 (a) and (b) show the temperature profiles for the gaseous and solid

phases and the radiative source term along the reactor calculated for the baseline

parameters listed in Table 7.1 at two radial positions: center (r = 0.0 m) and close


to the wall (r = 0.025 m). Note that the x-axis scale is logarithmic for a bet-

ter appreciation of the rapid heating rate — of the order of 105 K/s — across the

first centimeters after the inlet plane, where particles reach temperatures of 2144

and 1861 K for Campaigns 1 and 2, respectively. Thus, very efficient radiative

heat transfer to the reactants is attained by direct concentrated irradiation of the

gas-particle flow. Temperatures peak after 1 and 5 cm for Campaigns 1 and 2,

respectively, and decrease toward the rear of the reactor as a result of the endother-

mic reaction and conductive heat losses. As expected from the time scales τconv and

τrad, the temperature difference between the solid and gas phases is almost zero for

Campaign 1 with smaller particles, and less than 50 K for Campaign 2 with bigger

particles. The mean temperature of the metallic shell Tshell is in the range 1108–

1301 K and significantly below that of the coke particles because of the Al2O3 liner

and because the particle suspension serves as a radiation shield, as corroborated

experimentally in a similar gas-particle reactor configuration [110]. The radiation

source term undergoes a fast decrease as the reactants flow along the reactor because

of two phenomena: (1) the absorption coefficient of the particle suspension rapidly

decreases with increasing temperatures due to volumetric expansion and particle

shrinkage as the reaction progresses and (2) the mentioned increase of the particle

temperatures leads to strong emission thus lowering the net radiative source term.

A strong radial temperature gradient is induced by the angular and radial distribu-

tion of the incoming solar radiation. But this gradient disappears after approx. 10

cm behind the aperture as a result of absorption by the particles and emission from

particles and walls.

The progress of the chemical reaction is shown in Figs 7.11 (a) and (b), where the

variation of the chemical composition (chemical species’ molar fractions) is plotted

along the reactor at two radial positions: center (r=0.0 m) and close to the wall

(r=0.025 m). The baseline parameters of Campaigns 1 and 2, listed in Table 7.1,

were employed. Note that the x-scale is shown linear — and not logarithmic as

in Figs 7.10 (a) and (b) — because the chemical reaction rates are significantly

slower than the radiative heat transfer rates. Main product gas components are

H2 and CO, with less than 5% CO2, as predicted by thermodynamic equilibrium

[101]. Simultaneous fast pyrolysis and steam-gasification occurs immediately after

the entrance of the reactor. The conversion of steam and petcoke is significantly

lower close to the walls (r=0.025 m) as a result of the relatively low temperatures

existing there. In addition, the reaction rates decrease toward the exit of the reactor,


0 0.05 0.1 0.15 0.2 0.250

0.1

0.2

0.3

0.4

0.5

x, m

yi

r = 0.0 mr = 0.025 m

(a)

H2O

H2

C

CO

CO2

0 0.05 0.1 0.15 0.2 0.250

0.1

0.2

0.3

0.4

0.5

x, m

yi

r = 0.0 mr = 0.025 m

H2O

C

H2

CO

CO2

(b)

Figure 7.11: Variation of the chemical composition (chemical species’ molar frac-tions) along the reactor at two radial positions: center (r = 0.0 m) and close to thewall (r = 0.025 m). The baseline parameters listed in 7.1 have been employed for(a) Campaign 1; and (b) Campaign 2.

primarily because of the lower temperatures attained for a decreasing radiation

source, as indicated in Fig. 7.11 (b). As expected, the gasification proceeds at a

higher rate in Campaign 1, due to the higher temperatures and particle effectiveness

η of the feedstock.


0 0.05 0.1 0.15 0.2 0.25

-0.05

0

0.05

-0.05

0

0.05

x, m

r,m

T , K

XC∂XC

∂t, 1/s

, 1/s

0 0.05 0.1 0.15 0.2 0.25

-0.05

0

0.05

-0.05

0

0.05

x, m

r,m

T , K

XC∂XC

∂t

0.980.9

0.50.25

0.010.0250.1

0.5 0.75

1800

18501900

1950

2050

0.75

0.99

0.96

(a)

2000

0.80.9

0.5 0.050.25

0.60.2

1250

1500

1000 1700

0.75

0.95

(b)

0.15

1650

16001725

Figure 7.12: 2D contour map of the carbon conversion XC and reaction rate dXC/dt(upper plot) and corresponding temperatures (lower plot). The baseline parameterslisted in 7.1 have been employed for (a) Campaign 1; and (b) Campaign 2.

This is also corroborated in the 2D contour maps of Figs 7.12 (a) and (b), showing

the reaction rate dXC/dt and the carbon conversion XC in the upper plot — scaled

with its maximal value — and the corresponding temperature profiles in the lower

plot. In Campaign 1, the peaks for the temperature (> 2050 K) and for the reaction


rate 1.4 · 105 1/s are pronounced and located within the first centimeters past the

inlet plane, followed by a decrease to 1800 K and 6 · 10−6 1/s toward the exit of

the reactor. In Campaign 2, a more uniform temperature and reaction rate field is

observed, with maximum values attained 5 cm after the inlet plane. The difference

between the two campaigns is attributed to the differences in the particle sizes of

their feedstock. The smaller particles used in Campaign 1 have a higher absorption

coefficient, as observed from Fig. 5.11, augmenting radiative heat transfer at the

entrance region but preventing radiation to penetrate deeper into the rear of the

cavity. The smaller H2O/C molar ratio and, consequently, the higher solid volume

fraction, further contribute to this effect. For example, at the location of maximum

temperature the solid volume fractions are 1.83 · 10−5 and 9.75 · 10−6 for Campaigns

1 and 2, respectively, and the corresponding absorption coefficients are 13.3 and 0.28

1/m, respectively. At the exit of the reactor, after condensing the unreacted and

excess H2O, the molar composition of the syngas was 0.66 H2/ 0.25 CO/ 0.09 CO2

for Campaign 1 and 0.68 H2/ 0.21 CO/ 0.11 CO2 for Campaign 2, which compares

well with the gas composition measured experimentally.

Finally, a sensitivity analysis was performed to elucidate the effect of a given

input parameter Z by computing the system derivative Si = dY/dZ, where Y is

the output of interest [86]. The input parameters are the solar power input Qsolar,

the coke feeding rate mcoke, the inlet temperature Tin, the emissivity of the walls

εwall, and the heat transfer coefficient at the window hw. The outputs of interest

are the mean cavity temperature Tcavity and the coke conversion XC at the reactor

outlet. Results are listed in Table 7.2, where the local relative sensitivity Si,relative

is found by dividing the numerator and denominator of the system derivative Si by

the respective absolute values of Y and Z:

Si,relative =

(Y(Z + ∆Z

2

)− Y

(Z − ∆Z

2

))/Y

∆Z/Z(7.50)

Qsolar is the input parameter with the strongest impact on the output, especially on

XC for both campaigns. For Campaign 1, increasing mcoke has a negative impact

on XC and Treactor. Interestingly, for Campaign 2, its impact on Tcavity is positive

because the feedstock is introduced at a much higher temperature after preheating.

Tin has a positive impact for both campaigns, but stronger for Campaign 2. An

increase in the inlet temperature compensates for the weaker radiation absorption of


Table 7.2: Relative sensitivity of input/output parameters for both experimentalcampaigns.

Campaign Z → Qsolar mcoke Tin εwall hw

1 Si,relative (Tcavity) 0.42 −0.090 0.026 −0.008 −0.0041 Si,relative (XC) 1.46 −0.41 0.073 −0.009 −0.003

2 Si,relative (Tcavity) 0.31 0.014 0.11 −0.015 −0.0012 Si,relative (XC) 1.81 −0.16 0.40 −0.11 −0.004

the feedstock used in Campaign 2 because it reduces the energy required to heat the

reactants to the reaction temperature and, consequently, increases the energy left to

drive the chemical reaction. The emissivity of the walls, εwall, was found to generally

have a marginal impact on the reactor performance. As expected, the sensitivity to

εwall is higher for Campaign 2, because — due to less efficient radiation absorption

by the particles — more radiation is incident directly on the walls. Finally, the

convective heat transfer coefficient at the window, hw, was found to have negligible

impact on the reactor performance.

Chapter 8

Optimization and Scale-up1

In the first part of this thesis a solar chemical reactor was designed and a 5 kW

prototype was fabricated and experimentally demonstrated for steam-gasification of

petroleum coke (petcoke) in a high-flux solar furnace. The present chapter focuses

on the optimization and scale-up aspects of such a high-temperature chemical reac-

tor. The tool employed is the two-phase reactor model based on the fundamental

analysis of heat and mass transfer in a directly-irradiated suspension of reacting par-

ticles presented in Chapter 7. Simulations are carried out to determine the optimal

process parameters of the 5 kW prototype reactor and that of the up-scaled 300 kW

reactor. Special emphasis is placed on the feedstock characteristics and the reactor

geometrical configuration, and their effect on the extent of carbon conversion and

solar energy conversion efficiency.

8.1 Simulation Setup

Table 8.1 lists the dimensions and baseline operational conditions of the two solar

reactors analyzed: Synpet5, for a solar power input of 5 kW, and Synpet300 for a

solar power input of 300 kW. Synpet5 was tested at PSI’s solar furnace (Section

3.3); Synpet300 will be tested at CIEMAT’s solar tower at the Plataforma Solar

de Almeria [77]. A scheme of the Synpet300 reactor configuration is depicted in

Fig. 8.1. In contrast to the reactor concept presented in the Chapter 3 the oil cooled

1Material from this chapter has been submitted for publication as: A. Z’Graggen, and A. Stein-feld. Hydrogen Production by Steam-Gasification of Carbonaceous Materials using ConcentratedSolar Energy — V. Reactor modeling, optimization, and scale-up. International Journal of Hydro-gen Energy, 2008.

106 8. Optimization and Scale-up

Table 8.1: Baseline operational parameters for the two reactors analyzed.

Reactor type Synpet5 Synpet300

Cavity volume m3 0.0016 1.4Aperture diameter Daperture, m 0.05 0.5

Feeding rates mcoke, g/min 2.5 500mH2O, g/min 6.6 1320

VAr, ln/min 4.1 0Overall molar H2O/C ratio mol/mol 2 2

Solar power input Qsolar, kW 5 300

Static pressure p, Pa 105

Overall heat transfer coefficient U , W/(m2K) 12–180 10Inlet temperature Tin, K 573.15 773.15Ambient temperature T∞, K 293.15

product outletpurging nozzleswindow

reactor liner(Al2O3 / SiO2)

reactor shell (Inconel)aperture

tangential injection nozzles

Figure 8.1: Scheme of the up-scaled chemical reactor configuration for the solarsteam-gasification of coke, featuring a continuous gas-particle vortex flow confinedto a cavity receiver and directly exposed to concentrated solar radiation.

frustum was substituted for a ceramic cone. Furthermore, the thickness of both, the

insulation and the reactor liner were significantly increased to reduce heat losses by

conduction and in order to lower thermal loads on the reactor shell.

The solar spectrum of the incoming radiation is approximated by Planck’s black-

8.2. Results 107

body emission at 5780 K. Figure 5.2 shows the solar power through the aperture as

a function of the aperture’s radius for (a) PSI’s solar furnace, and (b) CIEMAT’s

solar tower. Shown in Figures 5.3 is the angular solar flux distribution at the aper-

ture, used as input to the solar reactor. Spectral values for total transmittance,

reflectance, and absorptance of the 3 mm-thick quartz window are presented in

Fig. 5.7. The overall heat transfer coefficient U , calculated by a FE-solver with

thermal conductivities provided by the manufacturer (Section 5.1.2) ranges from 12

W/m2K at the center of the cavity to 180 W/m2K close to the aperture due to

heat bridges for Synpet5, and averages 10 W/m2K for Synpet300. The inlet tem-

perature Tin is set to 573 K and 773 K for Synpet5 and Synpet300, respectively,

corresponding to an augmented pre-heating of the feedstock for the up-scaled re-

actor. Petrozuata Delayed petroleum coke is employed as a feedstock (Table 2.2).

Most of the optimization calculations were performed with monodisperse particle

clouds with diameters in the range 0.5–1000 µm. In addition, for the purpose of

comparison with the experimental results, three polydisperse particle clouds were

considered: type 1, obtained by subsequent ball and jet milling; type 3, obtained by

80 m-screen sieving; and type 5, the raw feedstock as received from the refinery (see

Table 2.1). Figure 2.2 shows the volume density distribution for the three types of

particles, as determined by laser scattering (Horiba LA950). Mean initial diameters

are 2.21, 17.58 and 40.0 µm for types 1, 3, and 5, respectively. Finally, for the

purpose of computational speedup, convection between the two phases is assumed

instantaneous (Tg=Ts) and eqs (7.9) and (7.10) simplify to:

∇ · (ρh · u) = φrad (8.1)

In fact, as observed in Fig. 7.10, the temperature difference between the gas and

solid phase vanishes after the initial heating-up, thus having a negligible impact on

the carbon conversion XC at the outlet of the reactor.

8.2 Results

The baseline simulation parameters used for all calculations are listed in Table 8.1.

The effect of varying four parameters was examined: 1) particle size, 2) feeding rate,

3) input power, and 4) geometry of the cavity’s longitudinal section at a constant

cavity volume. The reactor performance is characterized by the extent of feedstock


conversion determined from a carbon mass balance (eq. (3.5)):

XC =nCO + nCO2 + nCH4

n0C

(8.2)

and by the solar-to-chemical energy conversion efficiency ηchem, defined as the portion

of solar energy input stored as chemical energy (eq. (3.7)):

ηchem =XC · nC ·∆HR |298

Qsolar

(8.3)

Particle size Figure 8.2 shows (a) the temperature profiles (average values in the

radial direction), and (b) reaction extents along the reactor axis. The calculation was

performed for the baseline parameters of Synpet5 and with monodisperse particle

clouds of initial diameters 0.57, 0.96, 4.6, 62 and 176 µm, corresponding to curves 1,

2, 3, 4 and 5, respectively. Small particles induce a high volumetric absorption of the

incoming solar radiation, resulting in extremely fast heating rates of 2.9 · 105 K/s,

but at the expense of preventing further penetration of radiation into the cavity.

0 0.1 0.2

1000

1200

1400

1600

1800

2000

1

2

3

4

5

x, m

T,K

(a)

0 0.1 0.20

0.2

0.4

0.6

0.8

1

1

2

3

4

5

x, m

XC

(b)

Figure 8.2: (a) Temperature profiles (average values in the radial direction), and(b) reaction extents along the reactor axis, for the baseline parameters of Synpet5(Table 8.1) and with monodisperse particles of initial diameters 0.57, 0.96, 4.6, 62,and 176 µm, corresponding to curves 1, 2, 3, 4, and 5 respectively.

8.2. Results 109

0.5 1 2 5 10 20 50 200

0

0.2

0.4

0.6

0.8

1

type 1 type 3 type 5

D, µm

XC

mcoke/mcoke,baseline

0.51.0

2.0

4.0

(a)

2 4 10 40 100 400

0

0.2

0.4

0.6

0.8

1

type 1 type 3 type 5

D, µmX

C


0.5

1.0

2.0

4.0

(b)

Figure 8.3: Reaction extent at the exit of the reactor as a function of the feedstock’sinitial particle diameter for (a) Synpet5, and (b) Synpet300. Curves are plotted forcoke and water feeding rates corresponding to 0.5, 1, 2, and 4 times the baselinevalues of Table 8.1. Indicated are the equivalent diameters for the polydispersefeedstock types 1, 3, and 5.

This effect of shadowing is clearly visible for curves 1 and 2. For example, 0.57

µm particles peak at 1920 K in the first centimeter followed by rapid decrease to

1150 K toward the rear of the cavity, leading to a reaction rate slow down. In

contrast, bigger particles absorb the incident radiation only partially, resulting in

slower heating and lower peak temperatures, as observed for curves 4 and 5. For

example, 176 µm particles are heated at 2.1 · 103 K/s and peak at 1670 K after

12 cm from the entrance of the reactor. Curve 3 corresponds to a more uniform

distribution of the radiative absorption over the cavity volume, leading to higher

and more homogenously distributed temperatures, an thus to peak values for XC at

the reactor’s outlet.

Figure 8.3 shows the reaction extent at the exit of the reactor as a function of

the feedstock’s initial particle diameter for (a) Synpet5, and (b) Synpet300. Curves

are plotted for coke and water feeding rates corresponding to 0.5, 1, 2, and 4 times

the baseline values of Table 8.1. For maximum XC, optimal particle diameters

in the range 2–7 µm were obtained for Synpet5, and in the range 11–35 µm for

Synpet300 because of the larger cavity dimensions. Also indicated are the equivalent


diameters D32 for the polydisperse feedstock types 1, 3, and 5. Synpet5 exhibits good

performance with feedstock types 1 and 3, whereas types 3 and 5 are better suited

for Synpet300. Carbon conversions up to 65% and 99% are predicted for Synpet5

and Synpet300, respectively, using feedstock type 5.

Feeding rate Figure 8.4 shows the reaction extent and solar-to-chemical energy

conversion efficiency as a function of the coke mass flow rate (normalized to the

baseline rate) for the three feedstock types 1, 3, and 5, and for (a) Synpet5; and

(b) Synpet300. Also indicated (solid curve) are the results for the optimal initial

particle diameters, as determined in Figure 8.3. In both (a) and (b) cases, complete

reaction extent is achieved for very low feeding rates (mcoke/mcoke,baseline < 0.39

and < 0.57 for Synpet5 and Synpet300, respectively) at the expense of low energy

conversion efficiency. A moderate increase of the feeding rate to about 1.1 times

mcoke,baseline results in an increase in ηchem coupled to a slight decrease in XC. In this

case, the available heat is used efficiently to run the chemical reaction. A further

increase of mcoke leads to a drop of the temperatures inside the reactor because of

0

5

10

15

ηchem

,%

0.5 1 2 4 8

0

0.2

0.4

0.6

0.8

1


XC

optimal Dtype 1type 3type 5

ηchem

XC

(a)

0

5

10

15

20

25

ηchem

,%

0.5 1 2 4

0

0.2

0.4

0.6

0.8

1


XC

optimal Dtype 1type 3type 5

ηchem

XC

(b)

Figure 8.4: Reaction extent (left axis) and solar-to-chemical energy conversion ef-ficiency (right axis) as a function of the coke mass flow rate (normalized to thebaseline rate) for (a) Synpet5; and (b) Synpet300. The feedstock types are 1, 3, and5. The solid line is for optimal initial particle diameters in the range 2-7 and 11-35µm for Synpet5 and Synpet300, respectively.

8.2. Results 111

the heat consumed to heat the reactants, with a consequent decrease of both XC and

ηchem. Optimal feeding rates are in the 0.06–0.4 kg/h range and 17–44 kg/h range

for Synpet5 and Synpet300, respectively. As a consequence of the advantageous

volume-to-surface ratio for the scaled-up reactor and of its lower conduction losses

through a thicker insulation, ηchem can be more than doubled. Values of 10% and

24% are predicted for Synpet5 and Synpet300, respectively.

Input solar power Figure 8.5 shows the (a) solar-to-chemical energy conversion

efficiency, and (b) syngas yield as a function of input solar power for Synpet300 using

feedstock type 5. The syngas composition is characterized by H2/CO molar ratios

in the range 2.6–9.7 and CO2/CO molar ratios in the range 0.6–3.2. The dotted,

dash-dotted, and dashed lines represent the performance for selected reaction extents

of 0.8, 0.95, and 0.99, respectively. The solid curve in both figures represents the

locus of maximum ηchem and the corresponding reaction extents (Fig. 8.5a) and

petcoke feeding rates (Figure 8.5b). Evidently, higher XC is obtained at the cost of

200 300 400 500 6000

5

10

15

20

25

0.52

0.64

0.710.75

0.77

ηchem

,%

Qsolar, kW

200 300 400 500 6000

10

20

30

40

50

60

17

31

45

58

72

m,kg

/h

Qsolar, kW

optimalXC = 0.8XC = 0.95XC = 0.99

XC

(a)

mcoke, kg/h

(b)

H2

CO

Figure 8.5: Solar-to-chemical energy conversion efficiency (a), and syngas yield (b),as a function of input solar power for Synpet300 using feedstock type 5. The dotted,dash-dotted, and dashed lines represent the performance for selected reaction extentsof 0.8, 0.95 and 0.99, respectively. The solid curve in both figures represents thelocus of maximum ηchem. Indicated are selected corresponding values of XC in (a)and mcoke in (b).


lower ηchem. For example, for Qsolar=464 kW (spring equinox), mcoke=35 kg/h, and

XC=95% yields mH2=9.1 kg/h and mCO=39.8 kg/h at ηchem=20% (H2/CO=3.2 and

CO2/CO=0.8). An analogous calculation performed for feedstock type 3 yielded, as

expected, superior performance: ηchem=24%, mH2=10.8 kg/h and mCO=46.4 kg/h

for mcoke=41 kg/h (H2/CO=3.3 and CO2/CO=0.9). However, feedstock type 3

requires an additional processing step.

Geometry The seven cavity cross-sections shown in Fig. 8.6 were examined for

(a) Synpet5 and (b) Synpet300, where the radius and the length of the reactor are

varied while the cavity volume is kept constant; The geometries are labeled from

‘–3’ to ‘3’, corresponding to radii from 9.4 to 2.5 cm and overall reactor lengths

from 9.5 to 90 cm for Synpet5, and to radii from 121 to 25 cm and overall reactor

lengths from 71 to 686 cm for Synpet300. ‘0’ corresponds to the baseline design

shape of Synpet5 and Synpet300. Figure 8.7 shows the contour map of reaction

extents as a function of geometry and particle diameter for the baseline parameters

listed in Table 8.1 and for (a) Synpet5 and (b) Synpet300. Interestingly, Fig. 8.7a

0 0.2 0.4 0.6 0.8–0.1

0

0.1

x, m

r,m

–3–2–1 0 1 2 3

(a)

0 1 2 3 4 5 6 7

–1

0

1

x, m

r,m

–3–2

–1 0 1 2 3

(b)

Figure 8.6: Reactor cross-sections of constant volume where geometry ‘0’ corre-sponds to the baseline dimensions (Table 8.1).

8.2. Results 113

0.9

0.85

0.75

0.6

0.9

0.85

0.750.60.4

0.94XC

(a)

geometry #

D,µm

–3 –2 –1 0 1 2 3

0.5

1

2

5

10

20

50

D,µm

0.9

0.850.750.5

0.25

0.1

0.94

XC

(b)

–3 –2 –1 0 1 2 31

2

5

10

20

50

100

geometry #

Figure 8.7: Contour maps of reaction extent as a function of geometry (see Fig. 8.6)and particle diameter for the baseline parameters listed in Table 8.1 and for: (a)Synpet5, and (b) Synpet300.

shows two local maxima. The first, located at D = 5 µm for geometry #–1, is due

to a minimum in the conductive heat losses — 56% compared to a maximum of

62% of Qsolar —, while the radiative losses stay constant at 19% of Qsolar. Since

geometry #–1 approaches a sphere, it has the smallest surface-to-volume ratio. A

second maximum is found at D = 2 µm for geometry #3, as the cavity’s radius


approaches the aperture’s radius. In this case, all injected particles are exposed to

the peak flux and reach up to 2090 K compared to 1922 K for the first maximum.

As a consequence, the chemical reaction rate is faster and takes place in the 1st

half of the reactor. Conduction losses become less important (37% of Qsolar) due to

the somewhat lower wall temperatures in the 2nd half of the reactor, as products

exit at 582 K compared to 1743 K for the first maximum. In contrast, re-radiation

losses increase dramatically (48% of Qsolar) due to the high temperatures reached

in the front of the cavity. Figure 8.7b shows one maximum located at D = 8 µm

between geometry #2 and #3. In contrast to the calculations performed for Synpet5

(Fig. 8.7a), Synpet300 does not show a second maximum as it approaches a spherical

geometry because, due to the increased insulation thickness, conduction losses play

a lesser role.

Chapter 9

Conclusions

This thesis work was performed in the framework of a joint project between the

Swiss Federal Institute of Technology in Zurich (ETH) and the Paul Scherrer Insti-

tut in Villigen (PSI), both in Switzerland, the national research center ‘Centro de

Investigaciones Energeticas, Medioambientales y Tecnologicas’ in Spain (CIEMAT)

and the research and development center of Petroleos de Venezuela, S.A. (PDVSA

/ INTEVEP). The main goal of the still-ongoing project is the development and

demonstration of a process and of the related technology required for the produc-

tion of high quality syngas from carbonaceous materials using a solar thermochemical

process. ETH’s and PSI’s working packages included six tasks: (1) chemical thermo-

dynamics analysis, (2) second-law (exergy) analysis, (3) chemical kinetics analysis,

(4) engineering design and operation of a 5 kW prototype reactor (Synpet5), (5)

engineering design of the scaled-up 300 kW reactor (Synpet300) and (6) heat and

mass transfer modeling. Tasks (1)–(3) were performed by D. Trommer in a first PhD

Thesis [100], whereas tasks (4)–(6) were treated in the present thesis; main foci were

the Synpet5 experimental results and the modeling of both, Synpet5 and Synpet300.

The engineering design, fabrication, and testing of both reactors was the result of a

joint effort at the Professorship of Renewable Energy Carriers in cooperation with

the project partners.

9.1 Experimental Work

A solar chemical reactor for performing solar steam-gasification of carbonaceous

materials to syngas was successfully designed and tested. Solar experiments in a

116 9. Conclusions

high-flux solar furnace were performed for three distinct feedstock types. Coke

particles where either fed as dry powder (mean particle diameter 2.2 µm) or mixed

with water to form a slurry (mean particle diameter in the range 8.5–200 µm).

Vacuum residue was liquefied prior to injection. The reactor was operated smoothly

for approximately 100 experiments, each one lasting around one hour, yielding 64

valid experimental runs presented in Part I of this thesis. Best reactor performances

were measured for dry coke feeding, with carbon conversions of up to 87%, solar-

to-chemical energy conversion efficiencies up to 9%, typical reactor temperatures

above 1600 K and mean residence times around 1 second. The somewhat lower

values obtained for slurry and vacuum residue were caused by non-optimal feeding

systems, because the original reactor was designed for dry coke powder, not by

intrinsic drawbacks of these feedstock types. In fact, the design of the scaled-up

reactor foresees the use of coke-water slurries, which will facilitate feedstock handling

and control at large scales. The quality of the syngas produced was comparable to

the one typically obtained when heat is supplied by internal combustion. In general,

results indicate the technical feasibility of pyrolysing and steam-gasifying various

carbonaceous materials with concentrated solar energy.

9.2 Heat and Mass Transfer Modeling

A simple lumped-parameters model was formulated for solving the steady-state en-

ergy conservation equation that links the radiative power input with the power

consumed by the endothermic chemical reaction. The simulation was used, together

with the thermodynamic and kinetic analyses, to support the engineering design of

the Synpet5 reactor as well as to estimate the up-scaling potential. The retrospective

comparison the the experimental runs showed good agreement for the reactor’s tem-

peratures. Nevertheless the lumped-parameters model was not able to accurately

reproduce chemical conversion of carbon and steam at the reactor’s exit, especially

for Campaign 2.

A more sophisticated model was therefore developed which accounts for heat

and mass transfer in a polydisperse particle suspension subjected to concentrated

thermal radiation and undergoing an endothermic chemical transformation. The

exact reactor’s geometry was implemented while the incoming solar radiation was

characterized with preliminary MC calculations and the conduction in the reactor

walls was predicted by a FE simulation. Numerical results for average temperatures,

9.3. Outlook 117

coke and water conversion, and gas composition at the reactor outlet agreed well

with those obtained in the two experimental campaigns run with coke. The differ-

ent particle size distributions of the feedstock used in the two campaigns strongly

affected the absorption coefficient and chemical kinetics, and consequently mass and

heat transfer, justifying the detailed analysis of the polydisperse medium.

Finally the latter model was use to examine the lab-scale 5 kW prototype (Syn-

pet5), and its scaled-up version of 300 kW (Synpet300). The feedstock’s particle

size, the feeding flow rates, the solar power input, and the geometry of the reactors

were varied to identify the optimal operational conditions for maximum solar-to-

chemical energy conversion efficiency. Optimal initial feedstock particle sizes were

in the range 2–7 and 11–35 µm for the 5 and 300 kW reactors, respectively. The ad-

vantageous volume-to-surface ratio of the 300 kW scaled-up reactor and its enhanced

insulation lead to an optimal coke feeding rate of 44 kg/h for an solar-to-chemical

energy conversion efficiency of 24%, resulting in a syngas yield of mH2=8.8 kg/h

and mCO=15.6 kg/h. Moreover, the increased cavity dimensions permit the use of a

coarser feedstock to be reacted efficiently. Last, but not least, the tube-shaped cav-

ity was found to outperform the commonly employed barrel-shaped cavity, mainly

because most of the particles are directly exposed to the incoming high-flux solar

radiation.

9.3 Outlook

The scaled-up reactor designed by the ETH-team has been fabricated by CIEMAT

in 2007. The complete pilot plant is currently being assembled at the Plataforma

Solar de Almerıa in Spain, where the Synpet300 reactor will be installed on top of

the SSPS central receiver system. The new reactor is scheduled for operation in

2008 and will mark a further milestone in the development of a new gasification

technology for the solar upgrading of low-valued carbonaceous materials. The ulti-

mate objective is the commercial implementation of a solar gasification plant with

multiple receiver/mirror fields for the operation on an industrial MW scale.

The detailed model used to simulate radiative heat transfer in polydisperse re-

acting particulate media presented in this thesis was coupled to a rather simple fluid

flow solver, as the specific operational parameters allowed several simplifications. In

a more general context it is desirable to couple the radiation solver (here developed)

to a common 3D CFD full-featured solver, as these solvers typically lack of accu-

118 9. Conclusions

rate models for the treatment of participating media. In addition, model validation

would be enhanced by improving the accuracy of the experimental input parameters

such as Qsolar and mfeedstock, and by measuring also local flow properties such as gas

temperature and chemical composition in the volume. A measurement device for the

determination of gas temperatures in highly radiating environments was developed

for this purpose (Appendix A).

Part III

Appendices

Appendix A

Gas Temperature Measurement in

Radiating Environments1

An experimental methodology is developed for gas temperature measurements in

highly radiating environments. It consists of a suction thermocouple apparatus and

associated heat transfer model for determining the gas temperature from shielded

thermocouple readings by radiation, convection, and conduction dimensionless cor-

relations. The apparatus and methodology are calibrated and applied to measure

gas flow temperatures in a tubular furnace with wall temperatures up to 1223 K.

Results are compared with predictions by CFD simulations.

A.1 Additional Nomenclature

Latin Characters

ci parameter for empirical correlations

D diameter, m

L length, m

Greek Characters

δ relative difference

Subscripts

bare unshielded thermocouple

1This chapter has been published as: A. Z’Graggen, H. Friess, and A. Steinfeld. Gas temper-ature measurement in thermal radiating environments using a suction thermocouple apparatus.Measurement Science and Technology, 18:3329–3334, 2007.

122 A. Gas Temperature Measurement in Radiating Environments

furnace furnace

gas gas

Mo molybdenum, material of the thermocouple mantle

film gas film

sh shield

suc suction

tc shielded thermocouple

Dimensionless Numbers

Bi Biot number, h · L/k

Gz Graetz number, Re · Pr ·D/L

Nu Nusselt number, h ·D/kgas

Pr Prandtl number, µ · cp/k

Re Reynolds number, ρ · u ·D/µ

A.2 Introduction

Heat and mass transfer in radiating environments is of importance in numerous

high-temperature applications. Examples are combustion flames [108], fluidized

beds [118, 55], coal-fired furnaces [32, 64, 61], and, more recently, solar thermo-

chemical processes [60, 40, 71]. Especially the analysis of radiation heat transfer in

absorbing-emitting-scattering media is often based on model approximations. The

experimental validation of the resulting temperatures in radiative equilibrium plays

therefore a fundamental role in assessing the quality of such models. Optical tem-

perature measurements have been widely applied [17], but these require optical ac-

cess through protected transparent windows, which introduce design complications.

Fluid temperatures are commonly measured by means of bare thermocouples, but

measurement errors are caused by conduction along the thermocouple and by ra-

diative exchange with the surroundings [66]. Above 1000 K, the error induced by

radiation can be of several hundred degrees when the temperature difference between

fluid and surroundings is relative large and/or when the thermocouple is not prop-

erly shielded from incoming radiation [9]. The error can be significantly reduced by

making use of shielded suction thermocouples, where the shield attenuates the detri-

mental effect of radiative heat transfer and the suction flow amplifies the beneficial

effect of convective heat transfer. Suction - also referred to as aspiration - thermocou-

A.3. Experimental Setup 123

ples have been analyzed as a function of number of shields, aspiration velocity, and

surroundings temperature [9]. Correct temperature readings can then be achieved

by increasing the suction speed until no further change is observed [50], by correction

models based on intermittent measurements [39], and by correction models based on

the estimation of the radiative and convective heat transfer [63, 116]. The aforemen-

tioned studies are limited to their specific application because they suffer from one

or more of the following constraints: (1) information about the surrounding temper-

atures is required but difficult to estimate; (2) the coefficients for convective heat

transfer are affected by high uncertainties; and, (3) relatively large suction volumes

are needed. The present study shows how to alleviate these problems and provides

a more general methodology for the estimation of the measurement accuracy which

accounts for convection, radiation, and conduction heat transfer. The apparatus

is calibrated and applied to measure gas flow temperatures in a tubular furnace

with wall temperatures in the range 623-1223 K. A corroboration is accomplished

by comparison with predictions by CFD simulations.

A.3 Experimental Setup

The experimental setup is schematically shown in A.1. It encompasses two config-

urations for measuring the gas temperatures in a furnace and for calibrating the

measurement apparatus. A close-up view of the suction thermocouple apparatus

(lance) is depicted in A.2. It is composed of a 1 mm diameter (Dtc) thermocou-

ple type K, shielded by an Inconel 600 tube of 4 mm inner diameter (Dsh), 8 mm

outer diameter, and 30 mm length. Two additional thermocouples type K are built

into the shield tube at the same axial position. The shield is further attached to a

supporting Inconel 600 tube of 8 mm inner diameter, 10 mm outer diameter, and

200 mm length. Gas is aspired through the probe by means of a membrane vacuum

pump, while pulsations are attenuated by a 0.75 liters steel pressure vessel. Down-

stream, the gas flow is cooled to ambient temperature in a heat exchanger (HX)

and passes through an electronic flow controller (Vogtlin Instruments: red-y smart

series). The measurement apparatus is mounted inside a tubular furnace (Carbolite

MTF 10/25/130 with a max. nominal temperature of 1300 K, equipped with a 18

mm inner diameter, 300 mm length Al2O3 working tube). From the opposite side,

gas is injected into the furnace at a controlled mass flow rate. The flow pattern is

rectified by means of a ceramic honeycomb structure. At the exit of the furnace,


N2furnace

pulsationattenuator

membranepump

data aquisition

HX flowcontroller

suction unitmeasuring unit

calibration unit

lance

lance

heating coil

flowrectifier

air @

furnace

suc

Tfurnace Ttc

Tsh1Tsh2

Tambient

V

V

Figure A.1: Scheme of the experimental setup.

Lin Ltc

DtcDsh

Tgas

Ttc

suc

rad

inTsh1

Tsh2

A

A

section A-A

shield

thermocouples

Tbare

Tbare

V

V

Q

Figure A.2: Scheme of the suction thermocouple apparatus.

the gas is released to the environment via a cooling tower. For the purpose of cali-

bration, the measurement apparatus is removed from the furnace and an electrical

heating coil is applied to the first 100 mm of the lance to simulate a strong radiative

environment.

A.4. Heat Transfer Model 125

A.4 Heat Transfer Model

The numerical heat transfer model is based on the application of the energy conser-

vation equation to the tip of the shielded thermocouple. The domain is indicated by

the dashed box in A.2. Three heat transfer modes are considered: 1) convection from

the shield to the gas and from the thermocouple to the gas, 2) radiation between the

thermocouple and the shield, and 3) conduction at the base of the thermocouple.

The Bi number in the radial direction at the tip of the thermocouple,Bi = h·Dtc/kMo,

has values around 10−3, justifying the lumped parameter model for the thermocou-

ple temperature Ttc. For flow regimes with Gzsh > Gzlim, where Gzsh = Dsh

Lin·Resh ·Pr

and Gzlim ≈ 20 , the tip of the probe is in the thermal entrance region and con-

vection from the shield can be neglected [44]. The convective heat transfer at the

thermocouple tip can therefore be expressed by:

Qconv = Stc · h · (Tgas − Ttc) (A.1)

where Stc is the thermocouple’s surface exposed to the gas flow and is the av-

erage heat transfer coefficient calculated from h = Nu · kgas (Tfilm) /Dtc, evalu-

ated at the temperature of the gas film surrounding the thermocouple tip, Tfilm =

0.5·(Tgas + Ttc). The mean Nusselt number is derived from the empirical correlation:

Nu = c1 · Rec2tc · Pr1/3 (A.2)

where the Reynolds number at the tip of the thermocouple is given by:

Retc =u · ρ (Tfilm) ·Dtc

µ (Tfilm)=

4 ·Dtc

π ·(D2

sh−D2tc

) · msuc

µ (Tfilm)(A.3)

The view factor from the thermocouple to the gas inlet is less than 2%. Thus,

the influence of radiation from the surroundings is neglected. The radiative heat

transfer is then estimated using the expression for diffuse radiative exchange between

concentric cylinders:

Qrad =Stc · σ · (T 4

sh − T 4tc)

1εtc

+ Dtc

Dsh·(

1εsh

)− 1

(A.4)


where Tsh = 0.5 ·(Tsh1 + Tsh2) . Since the base of the thermocouple is attached to the

shield, the shield temperature is included in the conductive heat transfer calculation:

Qcond = keff · Scond ·Tgas − Ttc

Ltc

(A.5)

where Ltc is the characteristic length, Scond is the thermocouple cross section, and

keff is the effective thermal conductivity given by the empirical correlation,

keff = kMo (Ttc) · c3 ·(

Tsh − Ttc

Tsh − Tgas

)c4

(A.6)

The parameter c3 accounts for temperature-independent effects, in particular for

the fact that the whole thermocouple cross-section is not relevant for conduction

since its core is made of insulating SiO2. The parameter c4 accounts for nonlinear

effects, such as convective heat transfer from the gas to the shield at the clamping

of the thermocouple and nonlinear temperature profiles. The parameters c1 and c2

of eq. (A.2), and c3 and c4 of eq. (A.6) are identified by a set of calibration measure-

ments performed with air (see Section A.5). Heat balance at the thermocouple tip

yields:

Qcond + Qconv + Qrad = 0 (A.7)

Finally, the gas temperature is calculated by the implicit equation:

Tgas = Ttc −1

Stc · h· Qrad + Qcond (Tgas) (A.8)

using the Nelder-Mead Simplex method [69].

A.5 Calibration Measurements

The experimental operating conditions and results for a set of calibration measure-

ments performed with air are listed in Table A.2. Geometrical dimensions and

material properties are listed in Table A.3.

Three sets of calibration runs were performed with nominal shield temperatures

of 473, 573, and 673 K. The suction flow was varied between 0.5 and 2.9 ln/min, re-

sulting in Re at the thermocouple tip in the range 35-257 (eq. (A.3)). The reciprocal

Graetz number was always smaller the 0.05, namely in the range 0.0036-0.0214. For

A.5. Calibration Measurements 127

Table A.2: Experimental operating conditions and measurement results for the cal-ibration of the suction thermocouple apparatus.

Vsuc msuc · 106 Ttc Tsh Tgas Retc 1/Gzsh

ln min−1 kg s−1 K K K – –

0.49 9.56 524 674 293 35 0.02140.99 19.13 435 673 293 76 0.01071.48 28.71 403 674 293 119 0.00711.98 38.29 382 673 293 162 0.00532.47 47.86 365 653 293 207 0.00432.81 54.45 354 636 293 238 0.00380.49 9.56 453 575 293 37 0.02140.99 19.13 388 572 293 80 0.01071.48 28.71 362 572 293 124 0.00711.98 38.29 350 574 293 168 0.00532.47 47.86 344 574 293 212 0.00432.81 54.45 340 575 293 242 0.00380.49 9.58 387 473 293 40 0.02140.99 19.15 348 474 293 84 0.01071.48 28.73 331 471 293 129 0.00711.98 38.31 324 476 293 174 0.00532.47 47.86 320 472 293 218 0.00432.91 56.31 317 471 293 257 0.0036

example, a suction flow of 1.48 ln/min at an imposed shield temperature of 471 K

resulted in a thermocouple reading of 331 K and values of 179.5 W/(m2K) and 15.3

W/(m K) for the heat transfer coefficient and the effective thermal conductivity,

respectively. Equations (A.2) and (A.6) were fitted to the experimental data by

nonlinear regression using the least squares method. The results of the parametric

identification for c1, c2, c3, and c4, and the accuracy of the regression R2 are listed in

Table A.4. Figure A.3 shows the comparison of the measured Nusselt numbers with

the empirical correlation. The bars show the propagated error for a measurement

accuracy of ±2 K for temperatures and ±1.5% for suction mass flow rates. The

ratio Gzsh/Gzlim is found to be larger than 1 for every measurement, implying that

the tip of the thermocouple belongs to the entrance region. The calibrated model

can be applied to gas species and temperatures beyond the calibration conditions

thanks to the dimensionless nature of the heat transfer model.


Table A.3: Geometrical dimensions of the suction thermocouple apparatus and ma-terial properties for air, N2, and Mo.

Lin m 0.01Ltc m 0.01Dsh m 0.004Dtc m 0.001

εsh2 0.8

εtc2 0.8

kMo3 W/(m ·K) 138–104

kAir4 W/(m ·K) 0.024–0.084

µAir4 kg/(m · s) 1.8 · 10−5–4.8 · 10−5

kN24 W/(m ·K) 0.025–0.077

µN24 kg/(m · s) 1.7 · 10−5–4.5 · 10−5

2Oxidized surfaces assumed gray for Inconel [117] and Molybdenum [99].3Values presented for 293 K and 1223 K [98].4Values presented for 293 K and 1223 K [106].

Table A.4: Parameters for the empirical correlations of the Nusselt number (c1 andc2 in eq. (A.2)), and of the effective thermal conductivity (c3 and c4 in eq. (A.6),and regression accuracy R2.

c1 c2 c3 c4 R2

0.2867 0.6806 0.0779 –1.4973 0.993

A.6 Furnace Measurements

The suction thermocouple apparatus and associated methodology were applied for

determining gas temperatures of a N2 flow through a tubular furnace at different

furnace temperatures. The experimental operating conditions, measurement results,

and calculated gas temperatures are listed in Table A.5. Experiments were carried

out at three values of temperature, Tfurncae = 623, 923 and 1223 K. Nitrogen gas flow

rate to the furnace varied between 5 and 10 ln/min, whereas the suction flow was

in the range 0.5-2 ln/min. Gas temperatures were calculated using eq. (A.8) with

parameters of Table A.4. Figure A.4 shows the experimental data (solid lines) and

calculated gas temperatures (dashed lines) as a function of suction mass flow rate for

1223 K and a N2 flow to the furnace of 10 ln/min. The temperature of the shield Tsh,

A.7. CFD Simulations 129

30 50 100 200 300

2

3

5

10

15

Retc

Nu

0.2867 ·Re0.6806tc

·Pr1/3

NuGzsh/Gzlim

1

2

4

16

Gz s

h/G

z lim

Figure A.3: Left axis: experimentally determined and numerically computed Nus-selt number as a function of Reynolds number. Right axis: corresponding Graetznumbers, normalized by Gzlim.

as well as the temperature of the bare thermocouple Tbare, is almost independent of

the suction flow since the convective heat transfer plays a minor role. In contrast,

and as expected, the temperature of the central thermocouple Ttc strongly depends

on Vsuc, approaching the gas temperature with increasing suction flow. For example,

for Tfurnace= 1223 K, the average temperatures for Tsh, Tbare, Tgas and are 993, 889,

and 452 K, respectively. The shielded thermocouple reading Ttc was 829, 733, 668

and 634 K for suction flows Vsuc of 0.6, 1.0, 1.5 and 2.0 ln/min, respectively. Also

indicated in Figure A.4 is the gas temperature TgasCFDobtained by CFD simulations

(see Section A.7).

A.7 CFD Simulations

The gas temperature Tgas determined using the suction thermocouple apparatus

was compared with that obtained by CFD simulations TgasCFD. Their difference

can be considered as a reasonable estimate of the accuracy of the measurement

methodology. The computational domain is limited to the gas volume inside the

working tube of the furnace. The simulated gas is N2 with properties taken from

[106] and it is assumed to be non-participating with respect to radiation. Reynolds


Table A.5: Experimental operating conditions, measurement results, and calculatedgas temperatures with N2 at different furnace temperatures.

Vfurnace mfurnace Vsuc msuc Tfurnace Tbare Ttc Tsh Tgas TgasCFDRetc 1/Gzsh

ln/min mg/s ln/min mg/s K K K K K K – –

5.00 93.67 0.60 11.24 623 463 451 517 406 412 41 0.02225.00 93.67 1.00 18.73 623 460 436 514 406 423 69 0.013310.00 187.34 0.50 9.37 923 609 567 708 387 352 32 0.025810.00 187.34 1.00 18.73 923 605 495 691 391 354 67 0.013010.00 187.34 1.50 28.10 923 602 471 682 395 357 102 0.008710.00 187.34 2.00 37.47 923 598 459 676 399 361 137 0.006610.00 187.34 0.50 9.37 1223 908 867 1021 435 398 26 0.027810.00 187.34 0.60 11.24 1223 906 829 1015 448 398 32 0.023710.00 187.34 1.00 18.73 1223 901 733 999 450 397 55 0.014310.00 187.34 1.50 28.10 1223 896 668 985 451 398 86 0.009510.00 187.34 2.00 37.47 1223 893 634 977 458 401 116 0.0072

numbers obtained are in the range 149-742, justifying the laminar flow assumption.

The following boundary conditions (BC) are set:

• working tube wall : imposed temperature determined experimentally (see Table

A.6);

• shield : imposed temperature equal to Tsh;

• thermocouple: imposed temperature equal to Ttc;

• furnace inlet : plug flow with imposed mass flow rate mfurnace at ambient tem-

perature 297 K;

• furnace outlet : imposed pressure equal to 1 bar;

• thermocouple outlet : imposed mass flow rate equal to msuc.

Since gravity causes the shape of flow field to differ significantly from rotational

symmetry, only one symmetry plane is employed, which divides the domain in two

halves. The simulations are based on tetrahedral discretization meshes for one of

these half-domains. The discretization error for the computed temperatures was

found to be less than 1 K by stepwise grid refinement. The tabulated values of

TgasCFDhave been derived from the raw simulation results by averaging over a circular

A.7. CFD Simulations 131

0.5 1 1.5 2

350

500

650

800

950

1100

Vsuc, ln/min

T,K

Tbare

Tsh

Ttc

Tgas

TgasCFD

Figure A.4: Experimental data (solid lines) and calculated gas temperatures (dashedlines) as a function of the suction mass flow rate at a nominal furnace temperatureof 1223 and N2 mass flow rate of 10 ln/min.

disk located at the axial position of the lance inlet (195 mm from the working tube

inlet). The diameter of the disk is set equal to 0.75 · Dsh, ensuring that TgasCFDis

not affected by the temperature boundary layer belonging to the radiation shield.

The following relative temperature differences are defined:

δbare =

∣∣∣∣Tbare − TgasCFD

TgasCFD

∣∣∣∣ (A.9)

δtc =

∣∣∣∣Ttc − TgasCFD

TgasCFD

∣∣∣∣ (A.10)

δmethod =

∣∣∣∣Tgas − TgasCFD

TgasCFD

∣∣∣∣ (A.11)

Figure A.5 shows δbare, δtc, and δmethod as a function of the temperature of the

surroundings (Tfurncae) and of Re at the thermocouple tip. As expected, two impor-

tant effects can be observed. Firstly, high temperatures have a detrimental effect

on the readings of the bare and the shielded thermocouples as a result of radia-

tion heat transfer. Secondly, high Re numbers have a beneficial effect on Ttc as a

result of convective heat transfer. In fact, for the limiting case of Retc → ∞, Ttc

approaches the gas temperature. The relative difference between the temperature


Table A.6: Measured temperatures of the working tube in the furnace as a functionof the axial position x, starting from the inlet. The center of the furnace as well asthe entrance of the suction thermocouple apparatus is at position x = 0.195 m.

x, m

Tfurnace, K 0.0 0.03 0.06 0.09 0.12 0.15 0.18 0.21 0.24 0.27 0.3

TBC, K1223 340 340 366 425 622 1082 1278 1301 1247 773 662923 311 325 350 400 515 846 966 981 929 653 503623 311 321 337 370 432 555 637 663 646 513 411

650 775 900 1025 115010

–2

10–1

100

Tfurnace, K

δ

δbare

δtc

δmethod

40 80120 Retc

Figure A.5: Relative differences between the gas temperature simulated by CFD(TgasCFD

) and the bare thermocouple reading (Tbare), the shielded thermocouple read-ing (Ttc), and calculated gas temperature (Tgas), at different Reynolds numbers. Seedefinitions of δbare, δtc , and δmethod in equations (A.9), (A.10), and (A.11), respec-tively.

measured with the bare thermocouple and that of the gas is 12, 70, and 126% for

furnace temperatures of 623, 923, and 1223 K, respectively. For increased suction

flows, the difference is somewhat lower. For example, for Tfurnace = 923 K, δtc = 55,

36, and 29% for Re = 40, 80, and 120, respectively. The relative difference between

Tgas and TgasCFDvaries from 2% to 10% for Tfurnace from 623 to 1223 K, as a result

of the approximations in the boundary conditions used for the CFD simulations.

A.8. Summary and Conclusions 133

A.8 Summary and Conclusions

An apparatus and associated methodology for gas temperature measurements in

highly radiating environments has been developed. The main advantages of the

presented approach are threefold: 1) knowledge of the surroundings’ temperature

is not required; 2) it is applicable to gas mixtures and temperatures beyond the

calibration conditions; 3) small suction mass flow rates (∼ 0.5 ln/min) are sufficient,

minimizing the impact on the experimental environment. The relative large differ-

ences between the true gas temperature and the readings of the bare and shielded

thermocouples demonstrate the importance of an accurate and validated heat trans-

fer model. Miniaturization of the measurement apparatus would further enhance the

accuracy at a given suction flow due to larger Reynolds numbers. In addition, the

resulting lower thermal inertia would lead to faster response, and it would reduce

the influence of the external flow pattern on the flow inside the apparatus.

Appendix B

Measurement Accuracy of Qsolar

The PSI solar concentrator does not permit the continuous online measurement of

the solar power entering the reactor trough its aperture, Qsolar. The position of

the Venetian-blind type shutter ηshutter, defined as the ratio between the radiation

outside and behind the shutter, and solar radiation incident on the heliostat Isolar are

recorded instead. Solar power input is then calculated with the following empirical

correlation:

Qsolar = Isolar · ηshutter · k (B.1)

where k is found from measurements taken before and after the experiments. There,

the image of the concentrated solar radiation on a diffusely reflecting target is ana-

lyzed by a calibrated CCD camera, resulting in an estimate for Qsolar. The location

of the maximal incoming power is then identified manually and the reactor setup is

moved until this location and the aperture coincide. There are two main sources of

error in this procedure: (1) the values of k are strongly influenced by soiling of the

heliostat, the parabolic concentrator and the target. Furthermore, k is also affected

by the care of the operator that estimates Qsolar. Typical average values for k of 8.5,

6.9 and 7.9 with a standard deviation of 12.9%, 7.9% and 7.6% were obtained for

Table B.1: Influence of the inaccurate positioning of the aperture with respect tothe focus of the parabolic concentrator (spot).

placement error, m 0 0.005 0.01 0.02 0.03 0.04 0.05

Qsolar, kW 5.0 4.97 4.79 4.14 3.24 2.31 1.51

relative error, % 0 0.52 4.28 17.26 35.11 53.78 69.76

136 B. Measurement Accuracy of Qsolar

Campaign 1, Campaign 2 and Campaign 3, respectively. (2) the reactor’s aperture is

not perfectly aligned to the location of the maximal power (spot). A slow dislocation

of the spot was observed during operation as a consequence of imperfect tracking

of the sun path by the heliostat. The impact of this misalignment on Qsolar was

analyzed with the same simulation setup used in Section 5.1.1 to characterize the

angular and radial distribution of the incoming solar radiation. Results are shown

in Table B.1. A placement error of up to 5 cm was investigated. During typically

operation the bias does not exceed 1 cm, producing an error of ±04.3%. This error

combined with the average standard deviation of k then results in an estimated

accuracy of ±913% for Qsolar. Note that additional error caused by occasional dust

deposition on the window of the reactor is not taken into account.

The issue of online flux measurements on solar furnaces has been addressed by

several research groups. Interesting approaches, including the ‘moving bar’ principle,

are found in [5] and [104].

Appendix C

Radiation Absorption in the Gas

Phase1

The absorption coefficient of a single spectral line for a gas at temperature T , partial

pressure p and at wavenumber ν is described by [83]:

aij (ν, T, p) = Sij (T ) f (ν, νij, T, p) (C.1)

where Sij is the temperature-dependent line intensity, f is the normalized line shape

function, which takes into account temperature and pressure correction of the line

halfwidth as well as pressure shift of the line position, and νij is the transition

wavenumber of the line considered. The spectral absorption coefficient of a gas is

then calculated as the sum of all contributions aij at wavelength λ for each line:

κλ =pNA

RT

∑lines

aij (1/λ, T, p) (C.2)

Figure C.1 shows results for high resolution calculations in the range 0.1-40 µm

performed for a gas mixture of 50% CO and 50% H2O and at 1800 K and 1 bar for

the 918 (CO in HITEMP) and 744061 (H2O in HITRAN) lines in the database [83],

respectively. Also shown is the distribution of Planck’s emitted power Eλb at 5780

and 1800 K. The horizontal bars describe wavelength intervals that carry 99% of the

emitted power. Mean absorption coefficients of the gas mixture were calculated as

1FORTRAN code used in this chapter has been programmed in the framework of: A. Camen-zind, Spectral Radiation Properties of CO2, CO and H2O: Model Development and Implementa-tion, Semester Thesis, ETH Zurich, 2004.

138 C. Radiation Absorption in the Gas Phase

0.2 0.5 1 2 5 10 20

10–10

10–5

100

κλ,1/

m

λ, µm

10–5

100

105

e λb,W/(

m2·µm

)

H2O

CO

5780 K

1800 K

Figure C.1: Right axis: Absorption coefficients for CO and H2O at 0.5 bar as afunction of the wavelength. Left axis: emissive power of a blackbody at 5780 Kand 1800 K as a function of the wavelength. 99 % of the total power is carried byradiation in the intervals described by the horizontal bars. Solid and dotted linesare for temperatures of 5780 K and 1800 K, respectively.

a sum of each specie’s contribution, while κλ was weighted by Eλb:

κP =

∫∞0

κλEλbdλ∫∞0

Eλbdλ(C.3)

Values of approx. 0.06 and 0.006 m−1 were obtained for radiation emitted at 1800

and 5780 K, respectively. Absorption by the gas is therefore neglected vis-a-vis

absorption by the particles.

List of Figures

2.1 Schematic representation of a refinery. Only units involved in the

conversion of crude oil to vacuum residue and delayed coke are shown. 11

2.2 Particle size distribution functions of the petcoke feedstock types

listed in Table 2.1. (a) shows the number density (also called popula-

tion density); (b) shows the volume density. Type 1: jet milled, type

2: ball milled, type 3: sieved with a 80 µm screen, type 4: sieved with

a 200 µm screen, type 5: as provided by PDVSA. . . . . . . . . . . . 14

2.3 SEM micrographs of petcoke samples used in Campaign 1 (feedstock

type 1). (a) and (b) show unreacted particles, (c) and (d) show par-

ticles after pyrolysis above 1400 K, (e) and (f) show particles after a

typical experimental run with combined pyrolysis and gasification and

carbon conversion XC=0.75. Magnification: (a),(c) and (e) 4’000x;

(b),(d) and (f) 10’000x . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.4 Viscosity (a) and shear stress (b) as a function of the shear rate of a

water-coke slurry, measured for two types of feedstock [57]. The ar-

rows indicate the upward and downward curves, obtained by increas-

ing and subsequently decreasing the shear rate, respectively. Nicely

visible is the hysteresis loop, typical for thixotropic fluids. . . . . . . . 17

3.1 Scheme of the solar chemical reactor configuration used for the steam

gasification of carbonaceous materials at a power level of 5kW. The

location of the thermocouples (tc) of types K and S are indicated by

stars. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3.2 Feedstock injection setup used in Campaign 1. Petcoke particles are

fed by means of a brush conveyer and further entrained by the steam

injected from four tangential nozzles. . . . . . . . . . . . . . . . . . . 22

140 List of Figures

3.3 Feedstock injection setup used in Campaign 2. The coke-water slurry

is preheated in the tubes cast into the reactor’s insulation, the water

evaporates and the particles are injected into the cavity. . . . . . . . 23

3.4 Feedstock injection setup used in Campaign 3. The liquefied vacuum

residue is injected through a cold nozzle, the droplets are subsequently

heated, pyrolyzed and gasified. . . . . . . . . . . . . . . . . . . . . . . 24

3.5 Photograph of PSI’s high flux solar furnace. Main visible features are

the sun tracking heliostat, the Venetian-blind type shutter used to

control the incoming radiation and, inside the building, the parabolic

concentrator and the reactor. . . . . . . . . . . . . . . . . . . . . . . 25

3.6 Main components of the experimental setup at PSI’s high flux so-

lar furnace. Three different feeding systems are shown for the three

distinct experimental campaigns performed from 2004 to 2006. . . . . 26

4.1 Temperatures, solar power input and product gas composition for a

typical run of Campaign 1 (run # 7 in Table 4.1). The dashed vertical

line indicates the start of the petcoke feeding, the two vertical dotted

lines delimit the interval considered for the steady state calculations. 34

4.2 Breakdown of the total input power into heating of the reactants

Qheating, chemical reaction enthalpy Qchem, reradiation losses Qrerad

and conduction losses Qcond for the 23 runs of Campaign 1. Also

plotted is the measured solar power input, Qsolar with its respective

accuracy bounds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35


typical run of Campaign 1 (run # 10 in Table 4.2). The dashed verti-

cal line indicates the start of the slurry feeding, the two vertical dotted

lines delimit the interval considered for the steady state calculations. 38





accuracy bounds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

List of Figures 141

4.5 Carbon conversion and efficiencies for the 29 experimental runs of

Campaign 2 grouped by feedstock type and slurry molar ration H2O/C.

Feedstock types 2, 3 and 4 correspond to initial mean particle diam-

eters of 6.7, 17.6 and 30.8 µm, respectively. . . . . . . . . . . . . . . . 39


typical run of Campaign 1 (run # 12 in Table 4.3). The dashed

vertical line indicates the start of the petcoke feeding, the two ver-

tical dotted lines delimit the interval considered for the steady state

calculations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43





accuracy bounds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4.8 Fractions of hydrogen in the product gas derived by either pyrolysis

or gasification, and the feeding rates of reactants VR and steam for

experimental runs of Campaign 3. . . . . . . . . . . . . . . . . . . . . 44

5.1 Simulation setup used to determine the angular and radial distri-

butions of the incoming concentrated solar radiation at PSI’s solar

furnace [34]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

5.2 Solar radiative flux (left axis) and solar power (right axis) as a func-

tion of the aperture’s diameter for (a) PSI’s solar furnace, and (b)

CIEMAT’s solar tower. . . . . . . . . . . . . . . . . . . . . . . . . . . 53

5.3 Angular distribution of the concentrated solar radiation at the aper-

ture for PSI’s solar furnace and CIEMAT’s solar tower. . . . . . . . . 53

5.4 Lines of equal projected incident solar flux qsolar in kW/m2 on planes

parallel to the aperture. Results are given for a 50 mm aperture and

a nominal power trough the aperture Qsolar=5 kW. Also indicated is

the shape of the prototype reactor’s cavity (gray). . . . . . . . . . . . 54

5.5 Schematic of heat fluxes at the wall. Radiation emitted by particles

and other cavity wall elements, qic, is incident from the inside (right),

while the wall itself emits radiation toward the inside. On the outside

(left) the reactor is cooled by natural convection. Finally, conduction

occurs through the three material layers of the wall. . . . . . . . . . . 55

142 List of Figures

5.6 Overall heat transfer coefficient U for conductive losses trough the

reactor walls as a function of the location along the reactor axis. Also

indicated is the longitudinal cross-section of the reactor’s cavity. . . . 56

5.7 Complex refractive index of quartz n+ ik [72] and transmittance Trw

for a 3 mm thick window and for three selected incident directions θ

as a function of wavelength λ. . . . . . . . . . . . . . . . . . . . . . . 57

5.8 Schematic of heat fluxes at the window. Concentrated solar radiation

qsolar is incident from the outside of the reactor (left) whereas radiation

emitted by the cavity walls and particles, qic, is coming from the inside

(right). The window itself emits radiation on both sides. Finally, the

window is cooled by forced convection on the outside. . . . . . . . . . 58

5.9 Volume density of the polydisperse particles used in Campaigns 1

(type 1) and 2 (type 3). Solid lines are measured values for the unre-

acted samples, dashed and dashed-dotted lines are calculated results

at carbon conversions of 0.5 and 0.75, respectively. . . . . . . . . . . . 59

5.10 Equivalent monodisperse diameters calculated with eqs (5.10) and

(5.26) of the polydisperse particles used in Campaigns 1 (type 1) and

2 (type 3) as a function of the carbon conversion. . . . . . . . . . . . 60

5.11 Spectral distribution of the absorption and scattering coefficients, cal-

culated for a typical volume fraction of 5·10−5 and carbon conversions

of 0.0, 0.5 and 0.75, for the feedstock used in (a) Campaign 1, and

(b) Campaign 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

5.12 Scattering phase function for the equivalent diameter D32, for the

polydisperse medium (eq. (5.14)), and for the Henyey-Greenstein

approximation (eq. (5.15)), calculated for carbon conversions of (a)

0.0; and (b) 0.75, for the feedstock used in Campaign 1. Directions

θ = π/2 – π not shown in plot, because no backward scattering peak

was observed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

6.1 Diagram of the considered process and it’s energy and mass flows used

in the simulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

6.2 Carbon conversion (a), cavity (b) and shell (d) temperatures and

chemical efficiency (c) as a function of the carbon feeding rate and

the reactor volume for the 5 kW prototype reactor and the baseline

parameter listed in Table 6.1. . . . . . . . . . . . . . . . . . . . . . . 72

List of Figures 143

6.3 Predicted carbon conversion (a), cavity temperature (b), chemical

efficiency (c) and process efficiency (d) as a function of the carbon

feeding rate and the solar power for a scaled-up reactor. The baseline

parameter are listed in Table 6.1. . . . . . . . . . . . . . . . . . . . . 73

6.4 Numerically calculated (lumped parameters model) and experimental

measured data for the (a) 23 solar experimental runs of Campaign

1; and (b) 20 solar experimental runs of Campaign 2, ordered by

increasing Qsolar. Shown are the average cavity and wall temperatures,

the carbon conversion and, the chemical and process efficiencies. Full

circles indicate numerically calculated values; open circles indicate the

experimentally measured data; the error bars indicate the propagated

inaccuracy of the input parameters. . . . . . . . . . . . . . . . . . . . 74

7.1 Axis-symmetric model domain, featuring five concentric cylindrical

compartments. Indicated are also the boundary and inlet conditions.

The grid spacing in x-direction is adapted to the expected tempera-

ture gradients, shown is an example for the Monte-Carlo solver. . . . 82

7.2 Typical finite volume considered in the simulation with faces ‘w’ and

‘e’, centerpoint ‘P’, finite length ∆x, and cross-section Sj. Also shown

are the respective face-centered and cell-centered variables. Advection

is solved in x-direction for each compartment j separately, whereas

radiative exchange is considered along both, the x and the r directions. 83

7.3 Flowchart of the Monte-Carlo raytracer used the calculate the radia-

tive source term in a domain composed of finite volumes. Note that

the volumes are delimited by either interfacial faces or/and domain

boundaries. Computation steps that involve one or several random

numbers are show by the bold boxes. . . . . . . . . . . . . . . . . . . 85

7.4 Typical volume element used in the MC solver. A ray emitted from a

boundary face (gray) with initial power qray and wavelength λ under-

going absorption and scattering is shown (dotted arrow). The par-

ticipating medium is described by κλ,g, κλ,s, and σsλ,g the absorption

coefficients for the gas and solid phase, and the scattering coefficient

of the solid phase. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

144 List of Figures

7.5 Schematic of a ray passing trough several elements of width Ln and ex-

tinction coefficient βn. Also shown is the residual penetration length

ln in the last element. . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

7.6 Flowchart of the coupled solver. The bold box describes the MC

raytracer described in Fig. 7.3. . . . . . . . . . . . . . . . . . . . . . . 92

7.7 Relative RMS error for the temperature and underrelaxation parame-

ter γ (left axis) and total emitted power, scaled by its maximal value,

used as an indicator for system convergence (right axis) as function

of the iteration number i. Values for run #12 of Campaign 1 with

ξ = 32. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

7.8 Relative grid convergence errors in radial direction, εr, and in axial

direction, εx, for carbon and steam conversions at the outlet as a

function of the grid refinement factor. The grid refinement factor 1,

2, 4 and 8 corresponds to 10, 20, 40, 80 elements in x-direction and

to 2, 4, 8 and 16 elements in r-direction. . . . . . . . . . . . . . . . . 95

7.9 Numerically calculated and experimental measured data for the (a)

23 solar experimental runs of Campaign 1; and (b) 20 solar experi-

mental runs of Campaign 2, ordered by increasing Qsolar. Shown are

the average cavity and wall temperatures, and the carbon and steam

conversions at the outlet. Full circles indicate numerically calculated

values; open circles indicate the experimentally measured data; the

error bars indicate the propagated inaccuracy of the input parameters. 96

7.10 Numerically calculated temperatures profiles for the gaseous and solid

phases and the radiative source term along the reactor at two radial

positions: center (r = 0.0 m) and close to the wall (r = 0.025 m). The

baseline parameters listed in 7.1 have been employed for (a) Campaign

1; and (b) Campaign 2. . . . . . . . . . . . . . . . . . . . . . . . . . . 98

7.11 Variation of the chemical composition (chemical species’ molar frac-

tions) along the reactor at two radial positions: center (r = 0.0 m)

and close to the wall (r = 0.025 m). The baseline parameters listed

in 7.1 have been employed for (a) Campaign 1; and (b) Campaign 2. . 100

7.12 2D contour map of the carbon conversion XC and reaction rate dXC/dt

(upper plot) and corresponding temperatures (lower plot). The base-

line parameters listed in 7.1 have been employed for (a) Campaign 1;

and (b) Campaign 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

List of Figures 145

8.1 Scheme of the up-scaled chemical reactor configuration for the solar

steam-gasification of coke, featuring a continuous gas-particle vortex

flow confined to a cavity receiver and directly exposed to concentrated

solar radiation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

8.2 (a) Temperature profiles (average values in the radial direction), and

(b) reaction extents along the reactor axis, for the baseline parame-

ters of Synpet5 (Table 8.1) and with monodisperse particles of initial

diameters 0.57, 0.96, 4.6, 62, and 176 µm, corresponding to curves 1,

2, 3, 4, and 5 respectively. . . . . . . . . . . . . . . . . . . . . . . . . 108

8.3 Reaction extent at the exit of the reactor as a function of the feed-

stock’s initial particle diameter for (a) Synpet5, and (b) Synpet300.

Curves are plotted for coke and water feeding rates corresponding to

0.5, 1, 2, and 4 times the baseline values of Table 8.1. Indicated are

the equivalent diameters for the polydisperse feedstock types 1, 3, and

5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

8.4 Reaction extent (left axis) and solar-to-chemical energy conversion

efficiency (right axis) as a function of the coke mass flow rate (nor-

malized to the baseline rate) for (a) Synpet5; and (b) Synpet300. The

feedstock types are 1, 3, and 5. The solid line is for optimal initial

particle diameters in the range 2-7 and 11-35 µm for Synpet5 and

Synpet300, respectively. . . . . . . . . . . . . . . . . . . . . . . . . . 110

8.5 Solar-to-chemical energy conversion efficiency (a), and syngas yield

(b), as a function of input solar power for Synpet300 using feed-

stock type 5. The dotted, dash-dotted, and dashed lines represent

the performance for selected reaction extents of 0.8, 0.95 and 0.99,

respectively. The solid curve in both figures represents the locus of

maximum ηchem. Indicated are selected corresponding values of XC

in (a) and mcoke in (b). . . . . . . . . . . . . . . . . . . . . . . . . . . 111

8.6 Reactor cross-sections of constant volume where geometry ‘0’ corre-

sponds to the baseline dimensions (Table 8.1). . . . . . . . . . . . . . 112

8.7 Contour maps of reaction extent as a function of geometry (see Fig. 8.6)

and particle diameter for the baseline parameters listed in Table 8.1

and for: (a) Synpet5, and (b) Synpet300. . . . . . . . . . . . . . . . . 113

A.1 Scheme of the experimental setup. . . . . . . . . . . . . . . . . . . . . 124

146 List of Figures

A.2 Scheme of the suction thermocouple apparatus. . . . . . . . . . . . . 124

A.3 Left axis: experimentally determined and numerically computed Nus-

selt number as a function of Reynolds number. Right axis: corre-

sponding Graetz numbers, normalized by Gzlim. . . . . . . . . . . . . 129

A.4 Experimental data (solid lines) and calculated gas temperatures (dashed

lines) as a function of the suction mass flow rate at a nominal furnace

temperature of 1223 and N2 mass flow rate of 10 ln/min. . . . . . . . 131

A.5 Relative differences between the gas temperature simulated by CFD

(TgasCFD) and the bare thermocouple reading (Tbare), the shielded ther-

mocouple reading (Ttc), and calculated gas temperature (Tgas), at dif-

ferent Reynolds numbers. See definitions of δbare, δtc , and δmethod in

equations (A.9), (A.10), and (A.11), respectively. . . . . . . . . . . . 132

C.1 Right axis: Absorption coefficients for CO and H2O at 0.5 bar as a

function of the wavelength. Left axis: emissive power of a blackbody

at 5780 K and 1800 K as a function of the wavelength. 99 % of the

total power is carried by radiation in the intervals described by the

horizontal bars. Solid and dotted lines are for temperatures of 5780

K and 1800 K, respectively. . . . . . . . . . . . . . . . . . . . . . . . 138

List of Tables

2.1 Coke feedstock types used in the experiments. . . . . . . . . . . . . . 13

2.2 Approximate main elemental chemical composition (ultimate analy-

sis), low heating value and molar ratios H/C and O/C for PD coke. . 16

2.3 Approximate main elemental chemical composition (ultimate analy-

sis), low heating value and molar ratios H/C and O/C for vacuum

residue. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

3.1 Theoretical maximal solar-to-chemical energy conversion efficiency

ηchem for coke and vacuum residue for the 5 kW prototype reactor

and its 300 kW scale-up as a function of the cavity temperature. . . . 30

4.1 Steam and petcoke feeding rates, composition of the product gas,

nominal temperatures and performance parameters for 23 valid runs

of experimental campaign 1 conducted with dry petcoke particles. . . 36

4.2 Water and petcoke feeding rates, composition of the product gas,

nominal temperatures and performance parameters for 29 valid runs

of experimental campaign 2 conducted with a coke-water slurry. . . . 41

4.3 Steam and VR feeding rates, composition of the product gas, nom-

inal temperatures and performance parameters for 12 valid runs of

experimental campaign 3 conducted with liquefied VR. . . . . . . . . 46

4.4 Average operational parameters and results of the solar experimental

campaigns conducted with dry coke powder, cokewater slurry, and VR. 47

5.1 Complex refractive index of coke [14, 29]. . . . . . . . . . . . . . . . . 61

5.2 Arrhenius parameters of the kinetic rate constants for the steam gasi-

fication of coke [100]. . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

148 List of Tables

6.1 Baseline model parameters for the studies conducted for the 5 kW

lab scale reactor and it’s scale-up. . . . . . . . . . . . . . . . . . . . . 71

7.1 Baseline model parameters for two representative solar experimental

runs for Campaigns 1 and 2. . . . . . . . . . . . . . . . . . . . . . . . 97

7.2 Relative sensitivity of input/output parameters for both experimental

campaigns. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

8.1 Baseline operational parameters for the two reactors analyzed. . . . . 106

A.2 Experimental operating conditions and measurement results for the

calibration of the suction thermocouple apparatus. . . . . . . . . . . . 127

A.3 Geometrical dimensions of the suction thermocouple apparatus and

material properties for air, N2, and Mo. . . . . . . . . . . . . . . . . . 128

A.4 Parameters for the empirical correlations of the Nusselt number (c1

and c2 in eq. (A.2)), and of the effective thermal conductivity (c3 and

c4 in eq. (A.6), and regression accuracy R2. . . . . . . . . . . . . . . . 128

A.5 Experimental operating conditions, measurement results, and calcu-

lated gas temperatures with N2 at different furnace temperatures. . . 130

A.6 Measured temperatures of the working tube in the furnace as a func-

tion of the axial position x, starting from the inlet. The center of the

furnace as well as the entrance of the suction thermocouple apparatus

is at position x = 0.195 m. . . . . . . . . . . . . . . . . . . . . . . . . 132

B.1 Influence of the inaccurate positioning of the aperture with respect

to the focus of the parabolic concentrator (spot). . . . . . . . . . . . 135

Bibliography

[1] S. Abanades and G. Flamant. Solar hydrogen production from the thermal

splitting of methane in a high temperature solar chemical reactor. Solar En-

ergy, 80:1321–1332, 2006. 1.3

[2] R. Alvarez. An experimental investigation of the steam gasification of petcoke.

Semester Project, ETH Zurich, 2003. 3.1

[3] R. Alvarez. Solar thermal steam gasification of petcoke in an opaque fluidized

bed reactor. Master’s thesis, ETH Zurich, 2004. 3.1

[4] M. J. Antal, L. Hofmann, and J. R. Moreira. Design and operation of a solar

fired biomass flash pyrolysis reactor. Solar Energy, 30:299–312, 1983. 1.3

[5] J. Ballestrın and R. Monterreal. Hybrid heat flux measurement system for

solar central receiver evaluation. Energy, 29:915–924, 2004. B

[6] I. K. Basily and K. Thamer. Pyrolysis products dependence on the nature of

heavy hydrocarbon feedstock. Applied Catalysis, 73:125–133, 1991. 4.3

[7] W. H. Beattie, R. Berjoan, and J.-P. Coutures. High temperature solar pyrol-

ysis of coal. Solar Energy, 31:137–143, 1983. 1.3

[8] R. Berber and E. A. Fletcher. Extracting oil from shale using solar energy.

Energy, 13:12–23, 1988. 1.3

[9] L. G. Blevins and W. M. Pitts. Modeling of bare and aspirated thermocouples

in compartment fires. Fire Safety Journal, 33:239–259, 1999. A.2

[10] F. Bogner and K. Wintrup. Handbook of Synfuels Technology, chapter The

Fluidized-Bed Coal Gasification Process (Winkler Type), pages 111–125.

McGraw-Hill, New York, 1984. 1.2

150 Bibliography

[11] C. F. Bohren and D.R.Huffman. Absorption and Scattering of Light by Small

particles. Wiley, New York, 1983. 5.2.1, 5.2.1

[12] A. W. Bowman and A. Azzalini. Applied Smoothing Techniques for Data

Analysis The Kernel Approach with S-Plus Illustrations. Clarendon Press,

1997. 7.3.3

[13] R. Buck, J. F. Muir, R. E. Hogan, and R. D. Skocypec. Carbon dioxide

reforming of methane in a solar volumetric receiver/reactor: the CAESAR

project. Solar Energy Materials, 24:449–463, 1991. 1.3

[14] L. D. Burak. Radiative properties of coke particles of coal-dust flame.

Inzhenerno-Fizicheskii Zhurnal, 45:297–302, 1983. 5.2.1, 5.1, C

[15] V. L. Cabal. Climate change and European cleaner coal technology. Energy

& Environment, 14:3–15, 2003. 1

[16] P. Charvin, S. Abanades, G. Flamant, and F. Lemort. Two-step water split-

ting thermochemical cycle based on iron oxide redox pair for solar hydrogen

production. Energy, 32:1124–1133, 2007. 1.3

[17] P. R. N. Childs, J. R. Greenwood, and C. A. Long. Review of temperature

measurement. Review of Scientific Instruments, 71:2959–2978, 2000. A.2

[18] J. K. Dahl, K. J. Buechler, R. Finley, T. Stanislaus, A. W. Weimer,

A. Lewandowski, C. Bingham, A. Smeets, and A. Schneider. Rapid solar-

thermal dissociation of natural gas in an aerosol flow reactor. Energy, 29:715–

725, 2004. 1.3

[19] J. K. Dahl, A. W. Weimer, A. Lewandowski, C. Bingham, F. Bruetsch, and

A. Steinfeld. Dry reforming of methane using a solar-thermal aerosol flow

reactor. Industrial & Engineering Chemistry Research, 43:5489–5495, 2004.

1.3

[20] H. Dincer, F. Boylu, A. Sirkeci, and G. Atesok. The effect of chemicals on the

viscosity and stability of coal water slurries. International Journal of Mineral

Processing, 70:41–51, 2003. 2.1.1

[21] M. Dry. The SASOL route to fuels. ChemTech, 12:744–750, 1982. 1.2

Bibliography 151

[22] R. Edinger and S. Kaul. Humankind’s detour toward sustainability: past,

present, and future of renewable energies and electric power generation. Re-

newable and Sustainable Energy Reviews, 4:295–313, 2000. 1

[23] F. O. Ernst, A. Tricoli, , S. E. Pratsinis, and A. Steinfeld. Co-synthesis of H2

and ZnO by in-situ Zn aerosol formation and hydrolysis. AIChE, 52:3297–3303,

2006. 1.3

[24] J. T. Farmer and J. R. Howell. Advances in heat transfer, chapter Comparison

of Monte Carlo Strategies for Radiative Transfer in Participating Media, pages

333–429. Academic Press, 1998. 7.3.1

[25] M. Fasciana and F. Noembrini. Solar pet-coke gasification, direct irradiation.

Master’s thesis, ETH Zurich, 2003. 3.1

[26] M. Flechsenhar and C. Sasse. Solar gasification of biomass using oil shale and

coal as candidate materials. Energy, 20:803–810, 1995. 1.3

[27] E. Fletcher and R. Moen. Hydrogen and oxygen from water. the use of solar

energy in a one-step effusional process is considered. Science, 197:1050–1056,

1977. 1.3

[28] E. A. Fletcher. Solarthermal processing: A review. Journal of Solar Energy

Engineering, 123:63–74, 2001. 1.3

[29] F. Goodarzi. Optical properties of vitrinite carbonized at different pressures.

Fuel, 64:156–162, 1985. 5.2.1, 5.1, C

[30] M. R. Gray. Upgrading Petroleum Residues and Heavy Oils. Taylor & Francis,

CRC Press, 1994. 4.3

[31] D. W. Gregg, R. W. Taylor, J. H. Campbell, J. R. Taylor, and A. Cotton. Solar

gasification of coal, activated carbon, coke and coal and biomass mixtures.

Solar Energy, 25:353–364, 1980. 1.3

[32] Y. Guo, C. Chan, and K. Lau. Numerical studies of pulverized coal combustion

in a tubular coal combustor with slanted oxygen jet. Fuel, 82:893–907, 2003.

A.2

152 Bibliography

[33] R. P. Gupta, T. F. Wall, and J. S. Truelove. Radiative scatter by fly ash

in pulverized-coal-fired furnaces: Application of the Monte Carlo method to

anisotropic scatter. International Journal of Heat and Mass Transfer, 26:1649–

1660, 1983. 7.3.1

[34] P. Haueter, T. Seitz, and A. Steinfeld. A new high-flux solar furnace for high-

temperature thermochemical research. Journal of Solar Energy Engineering,

121:77–80, 1999. 3.3, 5.1, C

[35] M. Haug and C. Difiglio. Biofuels for Transport - An International Perspective.

Internation Energy Agency, 2004. 1

[36] L. G. Henyey and J. L. Greenstein. Diffuse radiation in the galaxy. Astrophys-

ical Journal, 88:70–83, 1940. 5.2.1

[37] C. Higman and M. van der Burg. Gasification. Gulf Professional Publishing,

2003. 1.2

[38] S. Hiller, O. Deussen, and A. Keller. Tiled blue noise samples. In Vision,

Modeling, and Visualization 2001, 2001. 7.3.3

[39] A. W. D. Hills and A. Paulin. The construction and calibration of an inex-

pensive microsuction pyrometer. Journal of Physics E, 2:713–717, 1969. A.2

[40] D. Hirsch and A. Steinfeld. Radiative transfer in a solar chemical reactor

for the co-production of hydrogen and carbon by thermal decomposition of

methane. Chemical Engineering Science, 59:5771–5778, 2004. 7.3.1, A.2

[41] D. Hirsch and A. Steinfeld. Solar hydrogen production by thermal decom-

position of natural gas using a vortex-flow reactor. International Journal of

Hydrogen Energy, 29:47–55, 2004. 1.3

[42] D. Hirsch, P. von Zedtwitz, T. Osinga, J. Kinamore, and A. Steinfeld. A new

75 kW high-flux solar simulator for high-temperature thermal and thermo-

chemical research. Journal of Solar Energy Engineering, 125:117–131, 2003.

3.1

[43] J. R. Howell. Thermal radiation in participating media, the past, the present

and some possible futures. Journal of Heat Transfer, 1100:1220–1229, 1988.

7.3.1

Bibliography 153

[44] F. P. Incropera and D. P. DeWitt. Fundamentals of Heat and Mass Transfer.

John Wiley & Sons, 2002. 5.1.3, 7.1, 7.1, 7.2, 7.2, A.4

[45] E. Jochem, G. Andersson, D. Favrat, H. Gutscher, K. Hungerbhler, P. R. von

Rohr, D. Spreng, A. Wokaun, and M. Zimmermann. Steps towards a sus-

tainable development. a white book for R&D of energy-efficient-technologies.

Technical report, CEPE/ETH Zurich and Novatlantis, Zurich, 2004. 1

[46] T. B. Johansson and L. Burnham. Renewable Energy: Sources for Fuels and

Electricity. Island Press, 1993. 1

[47] R. Kandiyoti, A. Herod, and K. Bartle. Solid Fuels and Heavy Hydrocarbon

Liquids - Thermal Characterization and Analysis. Elsevier, 2006. 1

[48] T. Kappauf and E. A. Fletcher. Hydrogen and sulfur from hydrogen sulfide-VI.

solar thermolysis. Energy, 14:443–449, 1989. 1.3

[49] J. Katzer. The Future of Coal - an Interdisciplinary MIT Study. Massachusetts

Institute of Technology, 2007. 1

[50] M. B. Khalil, F. M. El-Mahallawy, and S. A. Fareg. Accuracy of temperature

measurements in furnaces. Letters in Heat and Mass Transfer, 3:421–432,

1976. A.2

[51] S. Kim and B. E. Dale. Global potential bioethanol production from wasted

crops and crop residues. Biomass and Bioenergy, 26:361–375, 2004. 1

[52] T. Kodama. High-temperature solar chemistry for converting solar heat to

chemical fuels. Progress in Energy and Combustion Science, 29:567–597, 2003.

1.3

[53] T. Kodama, A. Kiyama, T. Moriyama, and O. Mizuno. Solar methane reform-

ing using a new type of catalytically-activated metallic foam absorber. Journal

of Solar Energy Engineering, 126:808–811, 2004. 1.3

[54] A. Kogan. Direct solar thermal splitting of water and on-site separation of the

products. International Journal of Hydrogen Energy, 23:89–98, 1998. 1.3

[55] A. K. Kolar and R. Sundaresan. Heat transfer characteristics at an axial tube

in a circulating fluidized bed riser. International Journal of Thermal Sciences,

41:673–681, 2002. A.2

154 Bibliography

[56] T. Kollig and A. Keller. Efficient multidimensional sampling. In EURO-

GRAPHICS 2002, volume 21, 2002. 7.3.3

[57] F. Kritter. Injection system of coal water slurries for solar reactors. Master’s

thesis, ETH Zurich, 2005. 2.4, C

[58] S. Lee. Alternative Fuels. Taylor & Francis, 1996. 1, 1.2

[59] O. Levenspiel. Chemical Reaction Engineering. Wiley, New York, 1999. 3.4.4

[60] W. Lipinski, A. Z’Graggen, and A. Steinfeld. Transient radiation heat transfer

within a non-gray non-isothermal absorbing-emitting-scattering suspension of

reacting particles undergoing shrinking. Numerical Heat Transfer, Part B,

47:443–457, 2005. 7.3.1, A.2

[61] F. C. Lockwood, A. P. Salooja, and S. A. Syed. A prediction method for

coal-fired furnaces. Combustion and Flame, 38:1–15, 1980. A.2

[62] M. Lundberg. Model calculations on some feasible two-step water splitting

processes. International Journal of Hydrogen Energy, 18:369–376, 1993. 1.3

[63] M. Luo. Effects of radiation on temperature measurement in a fire environ-

ment. Fire Sciences, 15:443–461, 1997. A.2

[64] J. Marakis, C. Papapavlou, and E. Kakaras. A parametric study of radiative

heat transfer in pulverized coal furnaces. International Journal of Heat and

Mass Transfer, 43:2961–2971, 2000. A.2

[65] J. Matsunami, S. Yoshida, O. Yokota, M. Nezuka, M. Tsuji, and Y. Tamaura.

Gasification of waste tyre and plastic (pet) by solar thermochemical process

for solar energy utilization. Solar Energy, 65:21–23, 1999. 1.3

[66] L. Michalski, K. Eckersdorf, J. Kucharski, and J. McGhee. Temperature Mea-

surement, chapter 17. Temperature Measurement of Fluids, pages 361–380.

John Wiley & Sons Ltd, 2001. A.2

[67] D. P. Mitchell. Generating antialiased images at low sampling densities. Com-

puter Graphics, 21:65–72, 1987. 7.3.3

[68] M. F. Modest. Radiative heat transfer. Academic Press, 2003. 3.1, 5.1.3, 5.2.1,

5.2.1

Bibliography 155

[69] J. Nelder and R.Mead. A simplex method for function minimization. Computer

Journal, 7:308–313, 1965. 6.1, A.4

[70] T. Osinga, U. Frommherz, A. Steinfeld, and C. Wieckert. Experimental inves-

tigation of the solar carbothermic reduction of ZnO using a two-cavity solar

reactor. Journal of Solar Energy Engineering, 126:633–636, 2004. 1.3

[71] T. Osinga, G. Olalde, and A. Steinfeld. Solar carbothermal reduction of ZnO:

Shrinking packed-bed reactor modeling and experimental validation. Industrial

& Engineering Chemistry Research, 43:7981–7988, 2004. A.2

[72] E. D. Palik. Handbook of optical constants of solids. Academic Press, 1985.

5.1.3, 5.7, C

[73] H. Y. Park and T. H. Kim. Non-isothermal pyrolysis of vacuum residue (VR) in

a thermogravimetric analyzer. Energy Conversion and Management, 47:2118–

2127, 2006. 4.3

[74] S. V. Patankar. Numerical heat transfer and fluid flow. Hemisphere Publishing

Corporation; New York, 1980. 7.3.2

[75] International Energy Agency. Industry Attitudes to Combined Cycle Clean

Coal Technologies. 1994. 1

[76] Lurgi — Kohle- und Mineraloltechnik. Lurgi Handbuch. Lurgi

Gesellschaften,Frankfurt a.M., 1970. 1.2

[77] Plataforma Solar de Almerıa. Annual Report 2006. CIEMAT, Spain, 2006.

8.1

[78] World Energy Council. 2007 Survey of Energy Resources. Regency House,

London, 2007. 1

[79] J. Rezaiyan and N. P. Cheremisinoff. Gasification Technologies. Taylor &

Francis, 2005. 1.2, 2.1, 4.1

[80] P. J. Roache. Verification and Validation in Computational Science and En-

gineering. Hermosa Publishers, Albuquerque, 1998. 7.3.3

156 Bibliography

[81] M. Roeb, A. Noglik, N. Monnerie, M. Schmitz, C. Sattler, G. Cerri,

G. de Maria, A. Giovanelli, A. Orden, D. de Lorenzo, J. Cedillio, A. le Digou,

and J.-M. Borgard. Development and verification of process concepts for the

splitting of sulphuric acid by concentrated solar radiation. In M. Romero, edi-

tor, 13th International Symposium on Concentrated Solar Power and Chemical

Energy Technologies, 2006. 1.3

[82] H.-H. Rogner. An assessment of world hydrocarbon resources. Annual Review

of Energy and the Environment, 22:217–262, 1997. 1

[83] L. S. Rothman, C. P. Rinsland, A. Goldman, S. T. Massie, D. P. Edwards,

J.-M. Flaud, A. Perrin, C. Camy-Peyret, V. Dana, Y.-Y. Mandin, J. W.

Schroeder, R. R. Gamache, R. B. Wattson, K. Yoshino, K. V. Chance, K. W.

Jucks, L. R. Brown, V. Nemtchinovi, and P. Varanasi. The HITRAN molecu-

lar spectroscopic database and HAWKS (HITRAN atmospheric workstation):

1996 edition. Journal of Quantitative Spectroscopy & Radiative Transfer,

60:665–710, 1998. 7.2, C, C

[84] D. Ruppert and M. P. Wand. Multivariate locally weighted least squares

regression. The Annals of Statistics, 22:1346–1370, 1994. 7.3.3

[85] P. L. Russell. Oil shales of the world, their origin, occurrence and exploitation.

Pergamon Press, 1990. 1

[86] A. Saltelli, S. Tarantola, F. Campolongo, and M. Ratto. Sensitivity analysis

in practice: a guide to assessing scientific models. John Wiley & Sons Ltd,

2004. 7.4

[87] W. G. Schlinger. Handbook of Synfuels Technology, chapter The Texaco Coal

Gasification Process. McGraw-Hill, New York, 1984. 1.2

[88] E. Schlunder, editor. VDI-Warmeatlas, Berechnungsblatter fur den

Warmeubergang. VDI-Verlag, Dusseldorf, 1984. 7.2

[89] L. Schunk, P. Haeberling, S. Wepf, D. Wuillemin, A. Meier, and A. Steinfeld.

A solar receiver-reactor for the thermal dissociation of zinc oxide. Journal of

Solar Energy Engineering, 2008. 1.3

Bibliography 157

[90] P. F. Sens and J. K. Wilkinson. Coal-water mixtures. Elsevier, London, 1989.

2.1.1

[91] R. Siegel and J. R. Howell. Thermal radiation heat transfer. Taylor and

Francis, 2002. 5.2.1, 7.3.1, 7.3.1

[92] R. Speck. Solar petcoke gasification, indirect irradiation. Master’s thesis, ETH

Zurich, 2003. 3.1

[93] A. Steinfeld. Solar thermochemical production of hydrogen - a review. Solar

Energy, 78:603–615, 2005. 1

[94] A. Steinfeld and R. Palumbo. Encyclopedia of Physical Science & Technology,

Vol. 15, chapter Solar Thermochemical Process Technology, pages 237–256.

Academic Press, 2001. 1.3

[95] R. W. Taylor, R. Berjoan, and J. P. Coutures. Solar gasification of carbona-

ceous materials. Solar Energy, 30:513–525, 1983. 1.3

[96] C. L. Tien. Thermal radiation in packed and fluidized beds. Journal of Heat

transfer, 110:1230–1242, 1988. 7.3.1

[97] M. Toda, M. Kuriyama, H. Konno, and T. Honma. The influence of parti-

cle size distribution of coal on the fluidity of coal-water mixtures. Powder

Technology, 55:241 – 245, 1988. 2.1.1

[98] Y. Touloukian, R. Powell, C. Ho, and P. Klemens. Thermophysical Properties

of Matter, Volume 1. Thermal Conductivity: Metallic Elements and Alloys.

Plenum Press, New York, 1970. A.3

[99] Y. S. Touloukian and C. Y. Ho. Thermophysical Properties of Matter, Volume

7. Thermal Radiative Properties: Metallic Elements and Alloys. Plenum Press,

New York, 1970. A.3

[100] D. Trommer. Thermodynamic and Kinetic Analyses of Solar Thermal Gasifi-

cation of Petroleum Coke. PhD thesis, ETH Zurich, 2006. 1.1, 5.2, 5.3, 5.3, 9,

C

[101] D. Trommer, F. Noembrini, M. Fasciana, D. Rodriguez, A. Morales,

M. Romero, and A. Steinfeld. Hydrogen production by steam-gasification of

158 Bibliography

petroleum coke using concentrated solar power - I. thermodynamic and kinetic

analyses. International Journal of Hydrogen Energy, 30:605–618, 2005. 1, 1.3,

4.1, 4.2, 4.3, 7.4

[102] D. Trommer and A. Steinfeld. Kinetic modeling for the combined pyrolysis

and steam-gasification of petroleum coke and experimental determination of

the rate constants by dynamic thermogravimetry in the 500-1520 K range.

Energy & Fuels, 20:1250–1258, 2006. 1.1, 4.2, 5, 5.3, 5.3

[103] H. R. Tschudi and G. Morian. Pyrometric temperature measurements in solar

furnaces. Journal of Solar Energy Engineering, 123:164–170, 2001. 3.1

[104] S. Ulmer, E. Lupfert, M. Pfander, and R. Buck. Calibration corrections of

solar tower flux density measurements. Energy, 29:925–933, 2004. B

[105] H. Usui, T. Saeki, K. Hayashi, and T. Tamura. Sedimentation stability and

rheology of coal water slurries. International Journal of Coal Preparation and

Utilization, 18:201–214, 1997. 2.1.1

[106] N. B. Vargaftik. Tables on the Thermophysical Properties of Liquids and

Gases. Hemisphere Publishing Corporation, 1975. A.3, A.7

[107] I. Vishnevetsky and M. Epstein. Production of hydrogen from solar zinc in

steam atmosphere. International Journal of Hydrogen Energy, 32:2791–2802,

2007. 1.3

[108] R. Viskanta and M. P. Menguc. Radiation heat transfer in combustion systems.

Progress in Energy and Combustion Science, 13:97–160, 1987. A.2

[109] E. V. Vogt, M. J. V. D. Burgt, J. H. Chesters, and J. C. Hoogendoorn. De-

velopment of the Shell-Koppers coal gasification process. Philosophical Trans-

actions of the Royal Society of London. Series A, Mathematical and Physical

Sciences, 300:111–120, 1981. 1.2

[110] P. von Zedtwitz, W. Lipinski, and A. Steinfeld. Numerical and experimental

study of gas-particle radiative heat exchange in a fluidized-bed reactor for

steam-gasification of coal. Chemical Engineering Science, 62:599–607, 2007.

7.3.1, 7.4

Bibliography 159

[111] P. von Zedtwitz, J. Petrasch, D. Trommer, and A. Steinfeld. Hydrogen pro-

duction via the solar thermal decarbonization of fossil fuels. Solar Energy,

80:1333–1337, 2006. 1.3

[112] P. von Zedtwitz and A. Steinfeld. The solar thermal gasification of coal - energy

conversion efficiency and CO2 mitigation potential. Energy, 28:441–456, 2003.

1, 1.3

[113] D. V. Walters and R. O. Buckius. Monte Carlo methods for radiative heat

transfer in scattering media. Annual Review of Heat Transfer, 5:131–176, 1992.

7.3.1, 7.3.1

[114] W. T. Welford and R. Winston. High collection nonimaging optics. Academic

Press, New York, 1989. 4.1

[115] C. Wieckert, U. Frommherz, S. Kraupl, E. Guillot, G. Olalde, M. Epstein,

S. Santen, T. Osinga, and A. Steinfeld. A 300 kW solar chemical pilot plant

for the carbothermic production of zinc. Journal of Solar Energy Engineering,

129:190–196, 2007. 1.3

[116] M. S. Wojtan and K. A. G. Jones. Improvements in the design and operation

of a suction pyrometer. Measurement and Control, 14:301–305, 1981. A.2

[117] W. D. Wood, H. W. Deem, and C. Lucks. Thermal Radiative Properties.

Plenum Press, New York, 1964. A.3

[118] D. Xie, B. Bowen, J. Grace, and C. Lim. Two-dimensional model of heat trans-

fer in circulating fluidized beds. Part II: Heat transfer in a high density CFB

and sensitivity analysis. International Journal of Heat and Mass Transfer,

46:2193–2205, 2003. A.2

[119] Y. S. Yang, J. R. Howell, and D. E. Klein. Radiative heat transfer through a

randomly packed bed of spheres by the Monte Carlo method. Journal of Heat

Transfer, 105:325–332, 1983. 7.3.1

[120] A. Z’Graggen, H. Friess, and A. Steinfeld. Gas temperature measurement

in thermal radiating environments using a suction thermocouple apparatus.

Measurement Science and Technology, 18:3329–3334, 2007. 1.1

160 Bibliography

[121] A. Z’Graggen, P. Haueter, G. Maag, M. Romero, and A. Steinfeld. Hydrogen

production by steam-gasification of carbonaceous materials using concentrated

solar energy - IV. reactor experimentation with vacuum residue. International

Journal of Hydrogen Energy, 33:679–684, 2008. 1.1, 1.3

[122] A. Z’Graggen, P. Haueter, G. Maag, A. Vidal, M. Romero, and A. Steinfeld.

Hydrogen production by steam-gasification of petroleum coke using concen-

trated solar power — III. reactor experimentation with slurry feeding. Inter-

national Journal of Hydrogen Energy, 32:992–996, 2007. 1.1, 1.3

[123] A. Z’Graggen, P. Haueter, D. Trommer, M. Romero, J. C. de Jesus, and

A. Steinfeld. Hydrogen production by steam-gasification of petroleum coke

using concentrated solar power — II. reactor design, testing, and modeling.

International Journal of Hydrogen Energy, 31:797–811, 2006. 1.1, 1.3, 6.1

[124] A. Z’Graggen and A. Steinfeld. Radiative exchange within a two-cavity con-

figuration with a spectrally selective window. Journal of Solar Energy Engi-

neering, 126:819–822, 2004. 4.2, 4.3

[125] A. Z’Graggen and A. Steinfeld. Heat and mass transfer analysis of a suspension

of reacting particles subjected to concentrated solar radiation — application

to the steam-gasification of carbonaceous materials. International Journal of

Heat and Mass Transfer, 2007, submitted. 1.1

[126] A. Z’Graggen and A. Steinfeld. Hydrogen production by steam-gasification

of carbonaceous materials using concentrated solar energy — V. reactor mod-

eling, optimization, and scale-up. International Journal of Hydrogen Energy,

2008, submitted. 1.1

Curriculum Vitae

Name: Andreas Z’Graggen

Nationality: Swiss

Citizen of: Schattdorf (UR)

Date of birth: July 25, 1978

2003-2008 Doctoral studies at the Professorship in Renewable En-

ergy Carriers, ETH Zurich; supervision: Prof. Dr. Aldo

Steinfeld

1997-2003 Diploma studies in Mechanical Engineering at ETH Zurich;

majors in Energy Technology and Product Development

1993-1997 Maturita tipo C (scientifico), Liceo Cantonale, Lugano

rights / license: research collection in copyright - …30596/... · diss. eth no. 17741 solar...

Documents