Politecnico di Milano · Ai miei nonni “Nothing in life is to be feared, it is only to be...

Politecnico di Milano SCUOLA DI INGEGNERIA INDUSTRIALE E DELL’INFORMAZIONE

Corso di Laurea Magistrale in Ingegneria Chimica

TESI DI LAUREA MAGISTRALE

Reactor Network Model of biomass combustion in fluidized beds

Anno Accademico 2016-2017

Candidato:

Matteo MENSI

Matricola 858912

Relatore:

Prof. Tiziano FARAVELLI

Correlatori:

Dr. Giancarlo GENTILE

Prof. Alberto CUOCI

Ai miei nonni

“Nothing in life is to be feared,

it is only to be understood.

Now is the time to understand more,

so that we may fear less.”

Marie Curie

Ringraziamenti

Vorrei ringraziare il prof. Tiziano Faravelli per l’opportunità fornitami di

misurarmi con un problema stimolante ed in grado di mettermi alla prova,

accompagnandomi con discussioni e consigli. Ringrazio il prof. Alberto

Cuoci per avere avuto la pazienza tra i suoi mille impegni di aiutarmi a fare

i primi passi nel mondo della programmazione in C++. E ringrazio

Giancarlo per avermi seguito e supportato con attenzione e interesse,

spingendomi a mettere in dubbio tutto e a prendermi in giro per essere più

vecchio di mio padre. Con loro, ringrazio tutti il resto del gruppo CRECK

Modeling, i prof. Alessio Frassoldati ed Eliseo Ranzi per l’immenso aiuto

che hanno saputo fornire sui vari aspetti di questa tesi, partecipando a

discussioni e chiarendo i miei dubbi, nonché Alessandro e Matteo.

Ringrazio tutte le persone che ho incontrato durante i miei studi al

politecnico, partendo da Nicolò, Luca, Claudia, Chiara e Davide. Grazie per

le belle serate e per i momenti di studio. Ringrazio Francesco, Daniele,

Andrea, Claudia e Paola per le infinite pause caffè, sempre troppo lunghe.

Un grazie va sicuramente a tutti i ragazzi dell’aula tesisti: Paolo, Erica,

Fabio e Francesco, “compagni” di gruppo, nonché Giulio, Matteo, Luca e

tutti gli altri.

Menzione d’onore per Ayoub e Francesca. Tempo di qualità passato

assieme. Gne.

Ringrazio i ladri in cucina Francesco, Luca, Manu e tutti gli altri che sono

sparsi nei ringraziamenti e citarli due volte fa tanto favoritismo e non mi

piace. Un giorno capirò cosa era quella salsa strana giapponese.

Ringrazio il resto della sagra della salsiccia, Niccolò, Marco e Alberto. Si

spera che Niccolò sia ancora vivo. Assieme a Niccolò ringrazio Mattia, per

il brillante fatturato ottenuto l’anno scorso.

Uscendo dall’ambito universitario, mi sarebbe impossibile non ringraziare

i miei amici di sempre: Matteo, Nicolò P & S, Marco, Federico, Paolo A & A

& I, Daniele, Fabio V e anche Gabriele, Fabio F, Simone e Marco. Senza

dimenticarsi di Ilaria, Giulia e Paola. Una grossa fetta di chi sono ora (e

considerando la mole una grossa fetta è grossa grossa) è grazie a voi. Grazie

di tutto, vi voglio un mondo di bene ragazzi.

Potrei non aver ringraziato tutti. Non sono propriamente bravo e ordinato

in queste cose. Se meriteresti un grazie e non l’ho messo, allora questo è

per te: grazie.

Ora è forse il caso di passare alla famiglia di sangue. Ringrazio tutti, dal

primo all’ultimo, specialmente i miei nonni Virginio e Franca. Sono

estremamente fortunato ad avere due nonni come voi. Ringrazio mia

madre per la (poca) pazienza e per avermi insegnato la disciplina. Di mio

padre ammiro la curiosità infinita e la determinazione. Ad entrambi devo

una quantità di affetto che forse non sono in grado di dimostrare, ma ci

provo. D’altro canto, è solo con il loro supporto che ho potuto fare questo

percorso.

Per ultima, ringrazio Federica, la mia ragazza. Grazie per questi anni

assieme, per il supporto, il conforto, l’incoraggiamento e tutti i momenti

speciali. Soprattutto, grazie per sopportarmi.

Table of contents Table of contents .......................................................................................... 7

List of Figures ............................................................................................. 11

List of Tables .............................................................................................. 15

Sommario ................................................................................................... 17

Abstract ...................................................................................................... 19

1. Introduction ................................................................................... 21

1.1 Lignocellulosic biomasses ................................................................ 23

1.1.1 Conversion routes for lignocellulosic biomass to liquid fuels 24

1.1.2 Thermochemical mechanism of biomass conversion ............. 26

1.2 Biomass gasification ......................................................................... 28

1.2.1 Fluidized bed gasifier ................................................................ 29

1.2.2 Challenges in predictive modeling............................................ 30

1.3 Aim and structure of this thesis ...................................................... 31

2. State of the Art ............................................................................... 33

2.1 Biomass kinetics ............................................................................... 34

2.1.1 Biomass composition ................................................................. 35

2.1.2 Biomass devolatilization and pyrolysis kinetic scheme .......... 41

2.2 Gasification units ............................................................................. 43

2.2.1 Types of gasifier ......................................................................... 43

2.2.2 Energy and cost management ................................................... 46

2.3 FBBG Modeling ................................................................................. 47

2.3.1 Conversion process .................................................................... 48

2.3.2 Classification of models ............................................................. 49

3. NetSMOKE ....................................................................................... 57

3.1 Mathematical formulation ............................................................... 58

3.1.1 A reactor network according to NetSMOKE ............................. 59

3.1.2 Mathematical formulation of the units .................................... 59

3.1.3 Formulation of the reactor network problem .......................... 62

3.1.4 Dealing with multi-phase systems ............................................ 65

3.2 Numerical methods .......................................................................... 66

3.3 Premise to NetSMOKE ...................................................................... 70

3.3.1 Why C++ ..................................................................................... 70

3.3.2 Libraries used and compatibility .............................................. 71

3.3.3 NetSMOKE vs KppSMOKE ......................................................... 72

3.3.4 NetSMOKE and CFD simulations ............................................... 72

3.4 NetSMOKE structure ........................................................................ 73

3.4.1 Solver .......................................................................................... 74

3.4.2 Data structures ........................................................................... 75

3.4.3 Input reader system ................................................................... 77

3.4.4 Units ............................................................................................ 78

3.4.5 Reactor network class ................................................................ 81

3.5 NetSMOKE algorithm ....................................................................... 82

3.5.1 Reading input and creating a reactor network ........................ 82

3.5.2 Preliminary operations performed by the network class ....... 85

3.5.3 Solving the fully-coupled system .............................................. 88

3.6 Validation ......................................................................................... 89

3.6.1 Tests ............................................................................................ 90

3.6.2 Conclusions ................................................................................ 99

4. Applications and Case Studies ..................................................... 101

4.1 RTD analysis ................................................................................... 102

4.1.1 Residence time distribution of ideal reactors ........................ 103

4.1.2 RTD analysis in NetSMOKE ..................................................... 105

4.1.3 Tests and result ........................................................................ 105

4.2 Van Paasen reactor ........................................................................ 109

4.2.1 Simulation setup ...................................................................... 111

4.2.2 Results ...................................................................................... 114

4.3 NREL 4FBR Reactor ........................................................................ 116

4.3.1 Simulation setup ...................................................................... 118

4.3.2 Results ...................................................................................... 121

4.4 Higher-complexity ERN.................................................................. 123

4.4.1 Results ...................................................................................... 126

5. Conclusions ................................................................................... 131

5.1 Future work .................................................................................... 133

5.1.1 Future work on biomass kinetics ............................................ 133

5.1.2 Expanding NetSMOKE ............................................................. 133

6. 135

7. 135

8. 135

9. References .................................................................................... 135

Nomenclature ..................................................................................... 143

Appendix A .......................................................................................... 145

Appendix B .......................................................................................... 149

Appendix C .......................................................................................... 153

List of Figures Figure 1.1: Data from EIA on energy consumption from different renewable

sources in the US, up to the last months of 2017. It is possible to see how biomasses

grew in recent years. ................................................................................................ 22

Figure 1.2: Different perspectives of biomass modeling. ....................................... 24

Figure 1.3. Biomass conversion pathways .............................................................. 25

Figure 1.4. Thermochemical conversion of biomass in a gasifier. Adapted from

Stark et al. [9] ............................................................................................................ 27

Figure 1.5. Overall process scheme for gasification to drop-in liquid fuels ......... 28

Figure 1.6. a) Bubbling bed reactor, b) Circulating bed reactor ............................ 29

Figure 2.1: Multiscale problem. Adapted from Ranzi et al. [12] ............................ 34

Figure 2.2: Comparison of atomic ratios of different classes of biomasses and coals.

The heating value increases with increasing H:C ratio and decreasing O:C. Adapted

from Jenkins et al. [13] ............................................................................................. 35

Figure 2.3: Cellulose unit. ......................................................................................... 36

Figure 2.4: Cellulose polymeric structure. .............................................................. 37

Figure 2.5: Monomers of hemicellulosic polysaccharides. .................................... 37

Figure 2.6: Hydroxycinnamyl alcohol polymers which lignins are composed of. 38

Figure 2.7: Diagram used for biomass characterization. ....................................... 39

Figure 2.8: Reference species used in the CRECK mechanism. .............................. 40

Figure 2.9: Types of gasifier ..................................................................................... 43

Figure 2.10: Updraft gasifier. The bed temperature profile is valid for a downdraft

gasifier as well. ......................................................................................................... 44

Figure 2.11: a) FBBG b) FCBG ................................................................................... 46

Figure 2.12: a) Three ways for indirect gasification, by external heat or internal

heat (gas and char). b) Indirect internal gasifier with char burning. ................... 47

Figure 2.13: Processes occurring in a FBBG. Adapted from Stark et al. [9] .......... 49

Figure 2.14: Multiscale modeling paradigms. Adapted from Bates [12] ............... 50

Figure 2.15. Kinetic post-processing method .......................................................... 53

Figure 3.1: General, top-level structure of NetSMOKE ........................................... 73

Figure 3.2: NetSMOKE splash screen in the terminal window. ............................. 74

Figure 3.3: Struct usage in C++. ................................................................................ 75

Figure 3.4: Data structures used in NetSMOKE. ..................................................... 76

Figure 3.5: Structure of the input reader system.................................................... 77

Figure 3.6: Structure of the unit class and all its ramifications. ............................ 78

Figure 3.7: Example of usage of polymorphism. .................................................... 80

Figure 3.8: Overview of the ReactorNetwork class. ............................................... 81

Figure 3.9: Preparing the ReactorNetwork class at runtime. ................................ 83

Figure 3.10: From raw unit data to a vector of objects. ......................................... 85

Figure 3.11: Preliminary operations of the network class in NetSMOKE. ............ 87

Figure 3.12: NetSMOKE approach for fully-coupled system solution. .................. 88

Figure 4.1: Typical CRTD and RTD function profiles for a step injection experiment.

................................................................................................................................. 104

Figure 4.2: Results from series of reactor RTD. Blue figure is one 1.0 s CSTR. Red

figure are two 1.0 s CSTRs. ..................................................................................... 106

Figure 4.3: Recycle cases configuration. ............................................................... 107

Figure 4.4: Residence time distribution for recycle cases. Blue lines for single

reactor, red lines for two CSTR in series with a recycle end-to-start. ................. 107

Figure 4.5: Configuration for a recycle with intermediate reactor and relative

CRTD curve. ............................................................................................................. 108

Figure 4.6: Configurations adopted for the bypass test case. .............................. 108

Figure 4.7: CRTD for a CSTR with bypass and two CSTR in parallel. ................... 109

Figure 4.8: Experimental setup for FBBG, taken from Van Paasen et al. [46] .... 109

Figure 4.9: Dimensions of the FBBG reactor, provided by ECN and taken from Stark

et Al. [9] ................................................................................................................... 110

Figure 4.10: From reactor to ERN model, taken from Gomez and Leckner [27] 112

Figure 4.11: Network model in NetSMOKE. .......................................................... 113

Figure 4.12: Plots of the main outlet species as a function of temperature. Dots are

experimental data from Van Paasen, red lines are with CRECK WGS and blue lines

are with Macak and Malecha [51] WGS. ............................................................... 115

Figure 4.13: Experimental set up used. Taken from by Bates et al. [39] ............. 117

Figure 4.14: Network morphology for the first modelling of the 4FBR NREL

reactor. .................................................................................................................... 120

Figure 4.15: Comparison between NetSMOKE results and Bates et al. [39]

measurements, red column is experiment and blue column prediction. ........... 122

Figure 4.16: ERN for the 4FBR reactor with a PFR representing bubbles. .......... 123

Figure 4.17: Three bubble expansion of the scheme in Figure 4.16. ................... 126

Figure 4.18: Outlet gas composition according to the three different ERN setups.

Red, blue and black bars are respectively zero, 1 and 3 bubbles morphologies.127

Figure 4.19: Total tars variation with the ERN complexity. Red, blue and black are

respectively 0, 1 and 3 bubble PFRs. ..................................................................... 128

Figure 4.20: Distribution of tars with temperature for the three ERN morphologies.

Red, blue and black stand for zero, 1 and 3 bubble reactors. .............................. 130

Figure C.1: Scheme of the FBBG model. ................................................................ 154

Figure C.2: Processes occuring on each element of volume of the PFR. ............. 155

List of Tables Table 2.1: CRECK Biomass kinetic scheme .............................................................. 42

Table 2.2: Different modelling approaches for fluidized beds. Adapted from

Gomez et al. [27] ....................................................................................................... 55

Table 3.1: System dimensions according to case in exam. ..................................... 64

Table 3.2: Comparison between OpenSMOKE and NetSMOKE, Isothermal CSTR.90

Table 3.3: Comparison between OpenSMOKE and NetSMOKE, Isothermal PFR. . 91

Table 3.4: Comparison between OpenSMOKE and NetSMOKE, Adiabatic CSTR. . 92

Table 3.5: Comparison between OpenSMOKE and NetSMOKE, Heat-exchange

CSTR. .......................................................................................................................... 93

Table 3.6: Comparison between OpenSMOKE and NetSMOKE, Adiabatic PFR. ... 94

Table 3.7: Comparison between OpenSMOKE and NetSMOKE, Heat-exchange PFR.

................................................................................................................................... 95

Table 3.8: Comparison OpenSMOKE++ vs NetSMOKE, series of reactors. ............ 96

Table 3.9: CSTR with a recycle. ................................................................................ 97

Table 3.10: PFR with recycle. ................................................................................... 98

Table 3.11: Trend of the norm of residuals of a PFR vs an approximating series of

CSTRs. ........................................................................................................................ 99

Table 4.1: Char gasification reactions from Ranzi et al. [12]. .............................. 102

Table 4.2: Dry ash free elemental composition of beech wood used in Van Paasen

et al. [46] .................................................................................................................. 110

Table 4.3: Inlet biomass composition used in NetSMOKE for reproducing Van

Paasen data. ............................................................................................................ 111

Table 4.4: General parameters for the network model used for Van Paseen et al.

[46] ........................................................................................................................... 113

Table 4.5: Reactor parameters for the network model used for Van Paseen et al.

[46] ........................................................................................................................... 113

Table 4.6: Values for WGS reaction. ...................................................................... 114

Table 4.7: dry composition of the oak used in Bates et al. [39] ............................ 116

Table 4.8: 4FBR reactor geometries from Gaston et al. [47] ................................. 117

Table 4.9: Biomass composition of oak chips used in the experiments. ............. 118

Table 4.10: General parameters for the network model used for Bates et al. [39]

................................................................................................................................. 121

Table 4.11: Reactor parameters for the network model used of Bates et al. [39]121

Table 4.12: Fluidization parameters calculated for the 4FBR reactor ................. 124

Table 4.13: Parameters calculated for the ERN with a single PFR to model the

presence of bubbles. ............................................................................................... 125

Table 4.14: Parameters calculated for the ERN with a series of PFR to model bubble

growth. .................................................................................................................... 125

▪17▪

Sommario Con il crescente interesse verso la gassificazione e la combustione di

biomasse, predire gli inquinanti nei differenti processi è diventato

fondamentale. La sfida in questo viene dalla forte interazione tra

fluidodinamica e chimica in reattori a letto fluidizzato, principali reattori

usati per la valorizzazione delle biomasse. Meccanismi cinetici dettagliati

sono obbligatori per investigare la formazione di tar, PAH e soot, ma se

accoppiati ad una simulazione CFD presentano tempi computazionali

proibitivi. Questo problema può essere risolto usando modelli di rete di

reattori equivalente, dove il campo di moto è fissato usando reattori ideali

interconnessi in vari modi sui quali è possibile applicare meccanismi

cinetici dettagliati.

Lavori precedenti sui modelli a rete di reattori riguardavano

principalmente sistemi omogenei gassosi. Quando più fasi erano coinvolte,

si tendeva a segregare il sistema e studiare i fenomeni disaccoppiati. Il

problema della segregazione è che si perdono informazioni sulle

interrelazioni tra i processi, nonché i problemi numerici dovuti ad una

lenta convergenza e alta probabilità di divergere nonostante il minor costo

computazionale dei metodi segregati.

Questo lavoro ambisce a gettare le fondamenta per modelli a rete di

reattori eterogenei “fully-coupled”. Un approccio di questo tipo consente di

includere meglio la competizione e le relazioni tra i vari processi. A questo

scopo, è stato sviluppato un tool numerico chiamato NetSMOKE, un

framework modulare che estende le librerie OpenSMOKE++ con la capacità

di trattare reti di reattori. La formulazione matematica del problema e la

struttura del software sono discusse. NetSMOKE è stato poi validato con

casi test e applicato nella riproduzione di dati sperimentali da reattori a

letto fluido pilota. I risultati ottenuti sono stati positivi.

▪19▪

Abstract With the rise of biomass gasification and combustion, predicting pollutants

in the different processes became fundamental. The challenges in this

derive from the strong relationships between fluid dynamics and

chemistry in fluidized beds, which are the main type of reactors used for

biomass valorization. Detailed kinetic mechanisms are mandatory to

investigate tar, PAH and soot formation, but their application in CFD

simulations presents prohibitive computational times. This issue can be

solved using equivalent reactor network (ERN) models, which allow to fix

the motion field using ideal reactors interconnected in various ways where

detailed kinetics can be applied.

Past work on these models involves gas-phase homogeneous cases. When

more than one phase is involved, segregation has been applied to study all

the phenomena in a disjointed fashion. The issue here is that pursuing

mathematical segregation reduces the capability to study interrelations

between processes and, from a purely numerical standpoint, while it

involves a lower computational cost in terms of raw CPU power

convergence is slow and solution can easily diverge.

This thesis aims to set the foundation for fully-coupled heterogeneous

reactor network models. A fully-coupled approach can encompass better

how each phenomenon influences the others. To this end, a numerical tool

was developed called NetSMOKE. It is a modular framework that extends

the OpenSMOKE++ libraries with capabilities to interpret and solve reactor

networks. Mathematical formulation of the problem is discussed. This tool

was then validated with test cases, and then used to reproduce

experimental data of pilot-scale fluidized beds reactors. Good agreement

was found.

▪21▪

Chapter 1

1. Introduction

Since the 1970s oil crisis there has been great effort to find viable

alternatives to fossil fuels. But that was due to economic reasons.

Nowadays, it is a matter of the survival of humanity.

World population is ever-growing. In 2016, the annual growth has been of

1.09% (which is a slower growth than 2015 and 2014 where it was of 1.14

and 1.12% respectively) which translates in roughly 80 million more

people, considering that currently on planet earth there are about 7.6

billion individuals [1]. Just two hundred years ago, world population was

estimated to be 1 billion.

This swelling in population was facilitated by the industrial revolution,

during which technologies to rapidly extract, transport, and convert energy

from fossil fuels were invented and widely adopted. In the same timespan

Chapter 1 ♦ Introduction

▪22▪

that took humanity to grow this large, emissions from fuel combustion have

raised from nearly zero to 8.76 GTc/year, IEA1 reports [2].

More than 60% of anthropogenic greenhouse gas emission is due to

combustions, and it is widely known that its accumulation in the

atmosphere is extremely likely to cause and have caused global average

surface temperature to rise.

Our dependence on fossil energy is simply unsustainable, and if business

as usual continues, the risks of catastrophic climate change will be even

worse than previously believed.

In this scenario, renewable biofuels play a critical role. They remain the

only sustainable energy source which can integrate into existing road and

aviation transport fuel infrastructure, and are indeed the most consumed

renewable energy source in the US (Figure 1.1).

Figure 1.1: Data from EIA on energy consumption from different renewable sources in the US, up to the last months of 2017. It is possible to see how biomasses grew in recent years.

1 International Energy Agency, founded in 1974 by the Organisation for Economic Co-operation and

Development (OECD), following the oil crisis.


▪23▪

In fact, IEA forecasts utilization of biofuels to quadruple by 2035 and to

supply a significant fraction (about 8%) of the global road transport fuel

demand. The issue with biofuels right now is that the current global

production of oil-equivalent products is almost entirely ethanol derived

from sugarcane or corn [3]: this raised issues on food competition and land

usage. As such, a technical limitation adds up, since most of the current

vehicle fleet and fuel distribution networks are unable to operate with

ethanol/gasoline fuel blends greater than 10% in ethanol volume.

All of this heads to a international regulatory direction towards the

development of second-generation biofuels. The 2009 Renewable Energy

Directive, issued by the EU commissions, requires biofuels to constitute

10% of transportation fuels by 2020. Recent proposals followed this

directive to focus on limiting the extent to which food-derived biofuels

count towards this mandate [4], [5].

Second generation biofuels differ from their predecessors because they

utilize lignocellulosic biomass – the most abundant and affordable form of

biomass – and include a variety of inedible feedstocks like

agricultural/forest residues, organic wastes and dedicated energy crops

(like willow or switchgrass). This reduces production costs and prevent

feedstock conflict between energy and food industries.

To understand how lignocellulosic biomasses can be converted into various

form of fuels or energy, an overview on the state of art biomass kinetic

modeling is following.

1.1 Lignocellulosic biomasses Kinetic modeling of lignocellulosic biomasses is out of the scope of this

thesis. Yet, as biomass application are a clear target of this work, as well as

providing the case study for this thesis, it is worth to spend some time

reviewing the available knowledge on biomass thermochemical

conversion modeling.


▪24▪

Figure 1.2: Different perspectives of biomass modeling.

Studying and understanding biomass kinetics is not an easy task. There are

several processes involved (Figure 1.2), happening in a multitude of scales,

and various possible thermochemical conversion pathways varying with

conditions. Being the matter solid, not only one has to face the pure

chemical part of biomass conversion, but even the fluid dynamics

consisting of mixing, intra and interparticle diffusion, thermal conductivity

and so on.

1.1.1 Conversion routes for lignocellulosic biomass to liquid fuels

Four main pathways for lignocellulose conversion to liquid fuels are shown

in Figure 1.3 and begin with either thermochemical or hydrolysis steps.

We can distinguish three different temperature levels, each favouring

specific conversion routes, plus an extra biotech processes category.

I. High temperature: When the temperature is higher than 700

°C, biomass is gasified into syngas from which it can be

catalytically synthesized into alkanes or alkenes via Fisher-

Tropsch.


▪25▪

II. Medium temperature: Around 500 °C, biomass is pyrolyzed

into bio-oil, which is a corrosive, unstable mixture of water, tars,

aromatics and char [1]. These bio-oils can be gasified to syngas or

upgraded to vehicle-ready fuels by deoxygenation under

hydrogen or other catalytic routes.

III. Mild temperatures: at under 200°C, acid hydrolysis occurs.

Dilute mineral acids at these conditions decompose the

lignocellulose to intermediate aqueous products which are then

separated and upgraded to value-added fuels and chemicals

through catalytic processes [6].

IV. Enzymatic promotion: Enzymatic hydrolysis of lignocellulosic

biomass results in sugar monomers and lignin. Sugar monomers

like glucose can be fermented by customized bacteria to produce

ethanol, while the by-product of lignin can be burned [7].

Figure 1.3. Biomass conversion pathways

Currently all four pathways constitute active fields of research. Biomass

gasification is currently being widely tested for commercial applications

due to its ability to utilize existing technologies (as coal gasifiers can be

revamped for a biomass application) and offering drop-in compatible fuels

for the current transportation infrastructure. Moreover, synthetic fuels are

SYNGASGASIFICATION

BIO-OILS

AQUEOUS PRODUCTS

SUGAR MONOMERS

PYROLISIS

ACID HYDROLYSIS

ENZYMATIC HYDROLYSIS

FERMENTATION

CATALYST

CATALYST

FISCHERTROPSCH

CELLULOSICBIOMASS

ALKANES &METHANOL

LIQUIDFUELS

FUELS &CHEMICALS

ETHANOL

TE

MP

ER

AT

UR

E


▪26▪

attractive because they demonstrate superior characteristics compared to

conventional diesels including a high cetane number (>70), low sulphur

and aromatics content [8], thus anything that aims to get these as products

is deemed interesting. The additional appeal of using biomasses is due to a

substantial carbon dioxide emission reduction from transportation

vehicles traffic, compared to traditional fossil fuels, thanks to biomass

being a “CO2-Neutral” feedstock.

1.1.2 Thermochemical mechanism of biomass conversion

Thermochemical conversion of biomass (and solid fuels in general) is a

continuous process defined by the interplay of a number of complex

processes (heat and mass transfer, primary and secondary pyrolysis,

heterogeneous chemistry and oxidation chemistry of gas products) which

cannot be thought as discrete or subsequent.

Nevertheless, it is possible to distinguish four processes during the

thermochemical conversion [9]:

I. Drying: Characterized by processes occurring at temperatures

around 100 °C in which moisture is liberated by evaporation.

II. Devolatilization: Chemical conversion from raw biomass to

gases and char. This happens between 200 °C and 600 °C, and

product distribution is a function of temperature and biomass

composition.

III. Secondary pyrolysis: The intermediate devolatilization

products undergo further pyrolytic reactions, heterogeneous

reaction with the char, tar cracking and oxidation and PAH

growth reactions in the gas phase.

IV. Char consumption: After the devolatilization process is

complete, solid residue undergoes slower oxidation reaction. In

the case of fluidized bed reactors, it is possible to have loss from

elutriation or due to ashes disposal.


▪27▪

All of these processes, resumed in Figure 1.4, occur simultaneously under

fluidized bed gasification conditions, which, as aforementioned, is the most

attractive application for commercial use at the moment.

Many works in the field of chemical conversion of biomass have been

aimed at understanding the reaction pathways of pure components of

biomass (lignin, hemicellulose and cellulose), while limited rigorous

particle models have been worked on, which capture anisotropic

characteristics of biomass considering fluid dynamics and competing

reaction pathways. These computational-cost demanding models come do

not allow for sufficiently extensive chemistry model to represent variations

in tar components.

Most recent publications have been focused on improvement of sub-

models and secondary gas phase mechanism.

Figure 1.4. Thermochemical conversion of biomass in a gasifier. Adapted from Stark et al. [9]


▪28▪

Debiagi et al. [10], [11] have been working heavily on biomass mechanisms.

The latest mechanism includes 32 reactions and 28 species which, coupled

with the detailed gas kinetic mechanism of more than 20000 reactions and

500 species, makes up for the full mechanism.

1.2 Biomass gasification During the gasification pathway, elevated operating temperatures (800-

1400 °C) combined with an oxidant (usually steam or oxygen)

thermochemically convert carbonaceous feedstocks into a mixture of

carbon monoxide and hydrogen known as syngas, which can be

subsequently be converted into various products (Figure 1.5).

Ideally, syngas mixture produced should be close to a 2:1 ratio in H2/CO

terms, making it suitable for synthetic fuels production through F-T

process.

Figure 1.5. Overall process scheme for gasification to drop-in liquid fuels

Depending on the feedstock used, the process is referred as coal-to-liquids

(CTL), biomass-to-liquids (BTL), or gas-to-liquids (GTL). Generally, the

process that stems from carbonaceous feedstocks to provide liquid fuels

can be called XTL. Compared to reforming solutions adopted for natural

gas, where the biggest engineering challenge is about controlling the

reactions that leads from methane to the desired products (which can be

Biomass& Waste

Coal

Natural Gas

SyngasCO + H2

GasificationReforming

T = 800 – °C

FTSynthesis

T = 200 – °C

AlkenesCnH2n


▪29▪

achieved in various ways, even autothermic processes), coal and biomass

gasification introduce the complication of solid materials and diffusional

resistances in the system.

There are different gasifier configurations for solid particles: fixed bed,

fluidized bed and entrained flow.

Fixed bed designs are unattractive for CTL/BTL applications due to their

scale limitations. Entrained flow gasifiers have scale advantages, but raw

biomass is not an optimal feedstock: it would require pre-treatment

processes like torrefaction and fine grinding.

Fluidized bed gasifiers overcome both the scaling issues associated with

fixed beds and the dry fine particle requirement of entrained flow

configurations.

1.2.1 Fluidized bed gasifier

The base concept of this configuration is to take advantage of fluidization.

By using the oxidizer gas in combination with other inerts like nitrogen,

and pellets of inert substances like silica or alumina, the solid bed can be

fluidized in the sense of making it behave like a fluid in motion: smooth

fluidization utilises chaotic mixing in the bed (induced by the fluidization

agent) to ensure excellent gas-solid heat and mass transfer characteristics.

Figure 1.6. a) Bubbling bed reactor, b) Circulating bed reactor


▪30▪

Two possible configurations are used: bubbling beds, where lower gas

velocities limit the bed expansion in the lower portion of the reactor, and

circulating beds, where gas velocity is higher for uniform expansion and

solid residues are circulated back into the bed after separation from gas

products using a cyclone. These two configurations are shown in Figure 1.6.

Currently, fluidized bed are the ones of most interest. An in-depth

comparison with the alternatives is carried on in the next chapter.

1.2.2 Challenges in predictive modeling

A fluidized bed gasifier is a system largely affected by fluid dynamics: it is

a chaotic system composed of emulsion (pseudo-monophasic system),

bubbles and discrete particles. According to the velocities involved, the

dimension of the pellets and other conditions, there can be different flow

regimes that affect thermochemical conversion due to different heat and

mass transfer involved.

Due to the duality in dependence from both chemical kinetics and fluid

dynamics, studying fluidized bed gasifier can be challenging: CFD

simulations are already computationally intensive, making a coupled

system of CFD and detailed kinetic mechanisms prohibitive due to the

computational times required.

Nevertheless, large detailed kinetic schemes are a necessity when pollutant

formation has to be investigated, as the reactions that lead to PAH, tars and

soot formation are in extreme non-equilibrium conditions and can include

thousands of pathways and species.

Since biomass is now looked at as the most possible alternative to fossil

fuels for the foreseeable future, its application are going to rise in number

and scale, making pollutant predictions mandatory.

Recently, renovated interested has been given to equivalent reactor

network approaches to tackle this issue.


▪31▪

1.3 Aim and structure of this thesis As interests towards biomass continuously grows, a numerical tool to allow

researcher to work easily is needed. Equivalent reactor network models

are an appropriate approach to these issues: besides allowing to use

detailed kinetic mechanisms, they can also help in the jump from lab-scale

reactors to real-scale reactors.

The problem with most reactor network models and associated tools is that

they tend to either work only on homogeneous cases or utilize segregation

when used for heterogeneous systems. The problem with segregation of

the processes involved is that, while computationally convenient in terms

of raw power, convergence is slow and often never reached as methods like

sequential substitution easily diverge.

As such, the aim of this thesis work was to develop a tool to solve

heterogeneous reactor network models in a fully-coupled fashion.

This thesis is structured as following:

I. Chapter 1: Introduction, with historical background and

presentation of both the problem dealt with and the aim of the

work.

II. Chapter 2: State of art of the issue investigated. Here, biomass

kinetics, fluidized bed gasifiers and different modeling

approaches are discussed.

III. Chapter 3: The numerical tool developed, NetSMOKE, is

presented in his mathematical formulation, its features and its

structure. The program was validated with the methods

illustrated.

IV. Chapter 4: Various experimental fluidized bed gasifiers data

were reproduced using NetSMOKE. The results obtained are

presented and discussed, as well as the approach taken to setup

the simulations.


▪32▪

V. Chapter 5: Summarizes the work done with a recap of the

obtained information to give an outlook of potential future work.

Finally, this thesis is correlated by three appendixes, referenced

throughout the manuscript, that contain mathematical sub-models

developed and correlations adopted.

In Appendix A, a simplified model for an heterogeneous perfectly stirred

reactor is developed and presented.

Appendix B collects the fluidization correlations used for this work, with

the reference to the publications where they were presented.

Appendix C illustrates a in-development model for fluidized bubbling bed

reactors.

▪33▪

Chapter 2

2. State of the Art

In the introduction chapter a quick overview of all the topics part of this

thesis was given. Before undergoing the central part of this work, it is

appropriate to review the “state of the art” – the currently best practices –

of all the topics touched or studied.

This allows to get a solid foundation on the problem and how it is currently

tackled. It also enables the reader to fully understand how and why

something was implemented in the code or done in a simulation.

Thus, three main themes will be discussed: the current state of chemical

kinetic modeling for biomasses (with a rapid look at gas phase mechanisms

as well), different models at various level of detail and scale for gasification

of solids, and the CRECK Modeling in-house KppSMOKE, the current

numerical tool to perform automatic ERN-based kinetic post-processing.

Chapter 2 ♦ State of the Art

▪34▪

2.1 Biomass kinetics As previously mentioned in the first introduction chapter, mathematical

modeling of thermal degradation of biomass is a challenging problem

because its complexity occurs at several levels:

I. Multicomponent problem: Biomass is a complex feed of many

different substances, with high variability based on its origin, and

it requires a proper characterization.

II. Multiphase problem: Biomass reacts in a condensed phase

forming a solid, a liquid and a gas phase. Biochar is later involved

in heterogeneous gas-solid reactions, while in the gas phase

released gases and bio-oil go through an homogeneous

mechanism.

III. Multiscale problem: Coupling of kinetic and transport processes

is needed both at a particle and at a reactor scale.

Figure 2.1: Multiscale problem. Adapted from Ranzi et al. [12]

Multiscale nature of the problem, as seen in Figure 2.1, is evident both at a

dimension and time scale: we can compare the angstroms of the molecular

scale to reactor dimensions in meters, as well as fast propagating radicals

with a short lifespan and the several minutes required to heat and


▪35▪

devolatilize thick biomass particles or char gasification which occurs much

slower than the release of volatiles.

2.1.1 Biomass composition

Defining the chemical composition of complex biomass materials is a first

key and necessary capability for modeling of thermochemical processes of

biomass conversion to fuels and chemicals. A database of more than 600

samples is available for research purposes [11].

Compared to coal, biomass has a lower density and a lower heating value

(see Figure 2.2). This is because of the oxygen content, which is around 35-

40% on dry basis weight.

Figure 2.2: Comparison of atomic ratios of different classes of biomasses and coals. The heating value increases with increasing H:C ratio and decreasing O:C. Adapted from

Jenkins et al. [13]

Cellulose, hemicellulose and lignin are the major building blocks of woody

biomasses, with some extractives (respectively, 30-55, 13-35, 14-36 and


▪36▪

the overall set of samples. Proximate, ultimate and structural analyses are

used to obtain information on biomass composition: this is not a trivial task

since biomass samples are heterogeneous and demand reliable

preparation methods. Institutions such as the National Renewable Energy

Laboratory (NREL) and the American Society for Testing and Materials

(ASTM) are seeking solutions to standardize preparation procedures and

analysis methodology [14], [15].

Molecular structure

Lignocellulosic biomass materials are mainly constituted by a combination

of polysaccharides, which can be grouped into holocellulose (cellulose and

hemicellulose) and lignin species. Other components such as moisture,

minerals, extractives and acetyl groups are also present.

Structurally, it is a porous structure where cellulose microfibril represents

the important element surrounded by hemicellulose and pectin, which act

as ligand and embed lignin materials.

Cellulose

It is the most abundant structural polysaccharide in cell walls and accounts

for 15-50% of the dry weight of plant biomass.

Figure 2.3: Cellulose unit.

It is composed of β-D-glucopyranose units (Figure 2.3) linked by β-1,4

glycosidic bonds, which can be summarized as (C6H10O5)x, so that mass

elemental composition is C = 44.4%, H = 6.2%, and O = 49.4%. An example

polymeric structure is reported in Figure 2.4.


▪37▪

Figure 2.4: Cellulose polymeric structure.

The degree of polymerization (DP) is of approximately 10-15 thousand

chair configuration units: DP is a structural property with high impact on

mechanical properties, solubility and hydrolysis of the biomass. Due to

several strong H-bonds, it has resilience towards hydrolysis and enzymatic

activity.

Hemicellulose

Accounts for 25-50% of the dry weight. It is a heterogeneous polysaccharide

made of hexose and pentose monosaccharide units and classified in three

major classes (xylans, mannans and xyloglucans) according to the

branching and richness in the particular monomers shown in Figure 2.5.

Figure 2.5: Monomers of hemicellulosic polysaccharides.

OH

O

O H

OHO H

OH

OH

O

O H

OHO H

OH

O H

O H

O HOH

O

O

OH

O H

OH

O H

Mannose Galactose Xylose Arabinose


▪38▪

Polymerization degree is relatively small at 70-200, with larger molecules

found in hardwoods and smaller ones found in softwoods.

Lignin

Lignins are aromatic racemic polymers resulting from oxidative coupling

of 4-hydroxy-phenyl-propanoid units and, among the fundamentals

components, they are the ones most influenced by the nature of the

biomass. This component of biomasses can provide a high heating value

due to the high H/C ratio. They contribute to rigidity of walls of secondarily

thickened cells.

Figure 2.6: Hydroxycinnamyl alcohol polymers which lignins are composed of.

Lignin polymers are derived from the three hydroxycinnamyl alcohol

monomers in Figure 2.6 (p-Coumaryl, coniferyl and sinapyl alcohols) that

differ in their methoxylation degree and that produce the main monomers

of lignin.

Other components

Biomasses are usually characterized by varying degree of extractives,

moisture (water) and the ash content: ashes are mainly inorganic

compounds (like metals) residual after complete combustion.

Characterization procedure

Proximate analysis is performed as a thermogravimetric analysis (TGA) in

accordance with the ASTM procedure. It allows the determination of

moisture, volatile matter, fixed carbon, and ash.

O H

OH

O H

OH

OCH3

OC H3

O H

OH

OCH3

p-Coumaryl Alcohol Sinapyl Alcohol Coniferyl Alcohol


▪39▪

Ultimate or elemental analysis decomposes the sample via high-

temperature oxidation to determine the composition of the main atoms C,

H, N and S through measurements of the corresponding oxidized gases.

Oxygen is calculated by difference. Mass spectrometry allows to determine

the elemental composition from electron ionization [16].

Structural or biochemical analysis measures the main components of

cellulose, hemicellulose, lignin and sometimes extractives and proteins.

This analysis is very valuable if the interest is to analyze the successive

biomass decomposition. The problem is that current methods (wet

chemicals) are time-consuming, labor-intensive and can induce some

degree of decomposition rendering the results inaccurate. Promising

progress is shown by Fourier transform infrared spectroscopy, but in

general these detailed analyses are not available [15].

Figure 2.7: Diagram used for biomass characterization.

When direct information from a structural analysis is not available, this

composition can be derived from elemental analysis: the H/C/O atomic

balances allow to evaluate a suitable combination of a few reference

species in cellulose, hemicellulose, lignin and extractives as these constitute

the largest portion. In the CRECK group approach [17] lignin is accounted

as three species according to richness in oxygen, hydrogen or carbon.


▪40▪

Figure 2.8: Reference species used in the CRECK mechanism.

Extractives are included via two lumped species for hydrophobic (TGL) and

hydrophilic (TANN) extractives. This adds up to a total of seven reference

species, all reported in , from which each biomass can be represented.


▪41▪

Via a the van Krevelen [18] diagram reported in Figure 2.7, which relates

atomic H/C and O/C ratios, it is possible to produce a composition in terms

of reference species as a linear combination respecting the H/C/O balances.

To reduce the number of degrees of freedom, three reference mixtures are

defined as combinations of the seven reference species. RM-1 represents

holocellulose, while RM-2 and 3 are mixtures of lignins with some content

of extractives. Their compositions are defined from experimental findings

and can be easily modified.

From the input H/C/O composition, the linear system of balance equations

gives the mass composition in terms of the three reference mixtures. From

these values and the respective internal composition of each RM, the

composition in terms of the seven reference species is calculated.

Linear combinations of these reference species are able to describe most of

the biomasses contained in the databases [19]–[22].

2.1.2 Biomass devolatilization and pyrolysis kinetic scheme

Table 2.1 reports the most recent biomass kinetic scheme proposed by the

CRECK group. It comprises 29 reactions and 28 species. For secondary

reactions, the detailed gas phase mechanism for hydrocarbons and

oxygenated fuels with more than 500 species and 20000 reactions is

coupled. According to the predictive focus of the work in analysis, the

group is capable of providing skeletal and reduced kinetic mechanisms for

specific fuel cases, to ease the computational task with no significant loss

of quality in the prediction [23].

Kinetics and molecular scale modeling evolution

In 2008 Ranzi et al. [24] proposed the first iteration of the pyrolysis model,

which has been expanded in Debiagi et al. [11] with the addition of

extractive species. The most recent version of the CRECK biomass scheme

was presented in 2017 in Ranzi et al. [17].


▪42▪

Particle scale modeling

A detailed moving-mesh model for particle scale CFD-applications, embed

in OpenFOAM® and called BioSMOKE, was developed by Gentile et al. [25].

Table 2.1: CRECK Biomass kinetic scheme

Pyrolysis Reactions Kinetic Parameters A (s−1), Eact

Cellulose

1 CELL → CELLA 1.5 × 1014 × exp(−47000/RT)

2 CELLA →

0.4 HAA + 0.05 GLYOX + 0.15 CH3CHO + 0.25 HMFU + 0.35

ALD3 + 0.15 CH3OH + 0.3 CH2O + 0.61 CO + 0.36 CO2 + 0.05 H2

+ 0.93 H2O + 0.02 HCOOH + 0.05 C3H6O2 + 0.05 G{CH4}+

2.5 × 106× exp(−19100/RT)

3 CELLA → LVG 3.3 × T × exp(−10000/RT)

4 CELL → 5 H2O + 6 CHAR 6 × 107 × exp(−31000/RT)

Hemicellulose

5 GMSW → 0.70 HCE1 + 0.30 HCE2 1 × 1010 × exp(−31000/RT)

6 XYHW → 0.35 HCE1 + 0.65 HCE2 1 × 1010 × exp(−28500/RT)

7 HCE1 → 0.6 XYLAN + 0.2 C3H6O2 + 0.12 GLYOX + 0.2 FURF + 0.4 H2O +

0.08 G{H2} + 0.16 CO

3 × T × exp(−11000/RT)

8 HCE1 →

0.4 H2O + 0.79 CO2 + 0.05 HCOOH + 0.69 CO + 0.01 G{CO} +

0.01 G{CO2} + 0.35 G{H2} + 0.3 CH2O + 0.9 G{COH2} + 0.625

G{CH4} + 0.375 G{C2H4} + 0.875 CHAR

1.8 × 10−3 × T × exp(−3000/RT)

9 HCE2 → 0.2 H2O + 0.275 CO + 0.275 CO2 + 0.4 CH2O + 0.1 C2H5OH +

0.05 HAA + 0.35ACAC + 0.025 HCOOH + 0.25 G{CH4} + 0.3

G{CH3OH} + 0.225 G{C2H4} + 0.4 G{CO2} + 0.725 G{COH2}+

5 × 109 × exp(−31500/RT)

Lignins

10 LIGC → 0.35 LIGCC + 0.1 COUMARYL + 0.08 PHENOL + 0.41 C2H4 + 1.0

H2O + 0.7 G{COH2} + 0.3 CH2O + 0.32 CO + 0.495 G{CH4}+

1 × 1011 × exp(−37200/RT)

11 LIGH → LIGOH + 0.5 ALD3 + 0.5 C2H4 + 0.2 HAA + 0.1 CO + 0.1 G{H2} 6.7 × 1012 × exp(−37500/RT)

12 LIGO → LIGOH + CO2 3.3 × 108 × exp(−25500/RT)

13 LIGCC → 0.3 COUMARYL + 0.2 PHENOL + 0.35 HAA + 0.7 H2O + 0.65 CH4 +

0.6 C2H4 + H2 + 1.4 CO + 0.4 G{CO} + 6.75 CHAR

1 × 104 × exp(−24800/RT)

14 LIGOH → 0.9 LIG + H2O + 0.1 CH4 + 0.6 CH3OH + 0.05 G{H2} + 0.3

G{CH3OH} + 0.05 CO2 + 0.65 CO + 0.6 G{CO} + 0.05 HCOOH +

0.85 G{COH2} + 0.35 G{CH4} + 0.2 G{C2H4} + 4.25 CHAR+

1 × 108 × exp(−30000/RT)

15 LIG → 0.7 FE2MACR + 0.3 ANISOLE + 0.3 CO + 0.3 G{CO} + 0.3

CH3CHO 4 × T × exp(−12000/RT)

16 LIG → 0.6 H2O + 0.4 CO + 0.2 CH4 + 0.4 CH2O + 0.2 G{CO} + 0.4 G{CH4} +

0.5 G{C2H4} + 0.4 G{CH3OH} + 2 G{COH2} + 6 CHAR

8.3 × 10−2 × T × exp(−8000/RT)

17 LIG → 0.6 H2O + 2.6 CO + 1.1 CH4 + 0.4 CH2O + C2H4 + 0.4 CH3OH+ 1 × 107 × exp(−24300/RT

Extractives

18 TGL → ACROL + 3 FFA 7 × 1012 × exp(−45700/RT)

19 TANN → 0.85 FENOL + 0.15 G{PHENOL} + G{CO} + H2O + ITANN 2 × 101 × exp(−10000/RT)

20 ITANN → 5 CHAR + 2 CO + H2O + G{COH2} 1 × 103 × exp(−25000/RT)

Metaplastic

21 G{CO2} → CO2 1 × 106 × exp(−24000/RT)

22 G{CO} → CO 5 × 1012 × exp(−50000/RT)

23 G{COH2} → CO + H2 1.5 × 1012 × exp(−71000/RT)

24 G{H2} → H2 5 × 1011 × exp(−75000/RT)

25 G{CH4} → CH4 5 × 1012 × exp(−71500/RT)

26 G{CH3OH} → CH3OH 2 × 1012 × exp(−50000/RT)

27 G{C2H4} → C2H4 5 × 1012 × exp(−71500/RT)

28 G{PHENOL} → PHENOL 1.5 × 1012 × exp(−71000/RT)

H2O Evap.

29 ACQUA → H2O 1 × T × exp(−8000/RT


▪43▪

2.2 Gasification units Following, an overview of the main type of gasifiers and approaches to coal

and biomass gasification is presented. While it might stem from the focus

of this work, it is important to have a clear idea of the alternatives, since it

allows to understand why fluidized bed applications are currently the most

attractive. Later on, this will be useful to understand how NetSMOKE can

be used to represent particular flow conditions of each of these gasification

approaches.

2.2.1 Types of gasifier

As mentioned previously, there are three types of gasifier: fixed or moving

bed, fluidized bed and entrained flow and they are quickly reported in

Figure 2.9.

Figure 2.9: Types of gasifier

Among these designs there are variations such as spouted bed, draught

tube, internally circulating fluidized beds, etc.

Fixed and moving bed designs

There are two possible flow configurations for fixed bed gasifiers: updraft

(counter-current) or downdraft (co-current). In updraft configuration the

gas flows from bottom to top. In this situation the gas leaves the near the

pyrolysis zone thus having a high content of tars, as it picks up moisture

and organic compounds from the cooler zones of the bed (see Figure 2.10).


▪44▪

The solid fuel is almost completely converted into gas and tars. Advantages

are relative insensitivity to pellet size and they can use wet fuels.

Figure 2.10: Updraft gasifier. The bed temperature profile is valid for a downdraft gasifier as well.

With a downdraft the gas enters top and exit from bottom and thus leaves

from the hottest zone, granting lower organic compounds concentration

but some particulate.

If the bed is moving, these configurations are still valid granted that fuel

bed moves from the top of the reactor to the bottom.

Fixed and moving bed applications are limited to few MW2 fuel power due

to difficulties in maintaining a regular conversion front – radially wise –

when the reactors is wide.

Entrained flow gasifiers

These systems are attractive for large-scale implementations (>400 MW):

fine coal or biomass are co-fed with the oxidizing agent to the main

chamber, and entrained ash can be removed with a liquid system. Thanks

to extremely turbulent flow and high temperatures, the produced syngas

or tail gas is clean and free of tars.

Given the high temperatures, coal and biomass ashes are melted into an

inert slag. This slag can be processed and utilize in cements blend to

valorize it as a by-product. With coal, this can be prohibitive since coal ash

2 Megawatts


▪45▪

has a high melting point, thus an higher amount of oxygen is needed for

slug operation. In the case of biomasses, ashes melt easily, and the oxygen

demand is kept low. However, molten slag from biomasses is corrosive and

aggressive to the reactor lining.

The other main drawback is the difficulty in pre-treating biomass for

drying and particle size reduction by grinding, which is expensive

compared to coal.

These reasons, together with the fact that currently the quantity of biomass

that can be delivered to a plant is limited, prevented introduction of

entrained-flow gasification in biomass applications. In the short terms, all

points out to fluidized bed being the status quo.

Fluidized bed gasifier

Fluidized bed gasifiers have a number of advantages over fixed beds,

especially with regards to mixing, reaction rates, and scalability. It is

possible to distinguish two different types of fluidized bed gasifiers:

bubbling (FBBG3) and circulating (FCBG4). Both are represented in Figure

2.11.

An FBBG operates at low gas superficial velocities, in the range of 0.5 to 2

m/s, in a way that most of the fuel remains in the lower part of the reactor,

and fluidization makes it bubble much like in a cauldron or pot. We have

then a bottom bed and the rest of the structure acts as a freeboard for the

gas.

FCBGs work with velocities from 2 to 5 m/s. Thus, bed expansion takes place

in the whole reactor with partial entraining: tail gas passes through a

cyclone or other solid-gas separation device to recover and recycle the fuel.

This favors solid contact time.

3 Fluidized bubbling bed reactor. 4 Fluidized circulating bed reactor.


▪46▪

Figure 2.11: a) FBBG b) FCBG

2.2.2 Energy and cost management

Depending on the way the heat necessary for the process is provided to the

gasifier, it is possible to distinguish two approaches: autothermal and

allothermal gasification (Figure 2.12).

Direct gasification

In autothermal (or direct) gasification, the heat is released during the

partial oxidation of the fuel in the gasifier itself. According to the use of air

or oxygen (thus acting on the equivalence ratio), it is possible to obtain

different heating value gases.

Indirect gasification

Allothermal or indirect gasification uses steam as the fluidizing agent while

heat is provided not by oxidation. This yields medium heating value gases

at the outlet and allows to save on the oxygen. Indirect gasification can be

further investigated according to the heat sources being external or

internal.

An example of an external heat source can be solar gasification or heat

exchangers. An indirect internal gasifier can work with the internal gas or

internal char: in the char-indirect configuration, a secondary fluidized bed


▪47▪

combustor is coupled to the main FB to burn the residual char. Heat is

exchanged by circulation of the bed between the two reactors.

The alternative is to recycle hot gases leaving the reactor. This usually is

not enough for realizing an autothermal process, but it is often used to

maintain bed temperature when the pellets are rich in moisture.

Figure 2.12: a) Three ways for indirect gasification, by external heat or internal heat (gas and char). b) Indirect internal gasifier with char burning.

Commercially, internal indirect gasifiers are more interesting and

currently offered. External indirect gasifiers have not been successful due

to the costs involved and technical problems related to heat transfer tubes

erosion.

2.3 FBBG Modeling There is a great amount of published works and reviews on modeling of

biomass and coal conversion: most of the available literature focuses on

the particle scale. Only a fraction of the reviews have been devoted to

reactor scale modeling of fluidized bed gasification, and mostly of coal.

Despite the different chemical and physical properties, no conceptual

differences subsist between biomasses and coal with respect to the model

structures and mathematical description of the process. This allows to


▪48▪

conserve most of the existing modeling elements for fluidized bed coal

combustors, but caution is necessary as there are differences such as in the

mode of conversion of char particles and in the amount of heat transferred

to surfaces. This means that once the model is structurally defined, there

are a number of processes and parameters that need to be adjusted

according to the fuel,

The choice of a model – and this is valid in a broader sense – depends on

the objectives and the experimental information available. Advanced

models are far more useful due to a wider information obtained from them

but are bound to reasonable and realistic input data. Simple models are

appropriate for many situations, i.e. preliminary predictions, as long as the

user is aware of their limitations.

2.3.1 Conversion process

As aforementioned, a biomass particle undergoes a series of processes:

initial drying and devolatilization, oxidation of volatiles and char and char

gasification by carbon dioxide and steam. The particles are affected by

shrinkage, porosity increase and primary fragmentation (caused by

thermal and internal pressure stresses of volatile release). Secondary and

percolative fragmentation and attrition of char particles take place

together with char conversion. Fluidization establish a (ideally) pseud-

homogeneous medium where these processes take place.

It is possible to distinguish two main detail levels in the particle scale and

reactor scale (Figure 2.13).

At the particle scale, volatile release and surface mechanism of char

conversion happen. The fuel particle will lower in density and dimensions

due to devolatilization and surface reactions. Particles communicate with

the reactor by heat and mass transfer, which rates are determined by the

fuel reactivity but also by the fluid dynamics of the bed. Its description

should include particle size, density and thermal conductivity. Detailed

models tend to include elutriation into account by means of semi-empirical

rates.


▪49▪

Figure 2.13: Processes occurring in a FBBG. Adapted from Stark et al. [9]

Reactor level focuses on the macro-scale of the bed, and such on the gas-

phase reactivity as well as the hydrodynamics. Thus, it is important to

consider a multi-phase region capable of representing both the emulsion

and the bubbles formed by fast gas entering the reactor. These bubbles can

grow along the height of the bed and exchange mass and heat with it and

influence mixing. Here, factors like residence times, fuel feed points and

rates, velocities must be considered. A meso-level, sort of a layer to link the

different scales of study through appropriate boundary conditions for heat,

mass and species transport, was proven important [26].

2.3.2 Classification of models

Fluidized bed gasifiers are a challenge, as they present a high degree of

complexity in both the chemistry and the fluid dynamics, which are strictly

interrelated. This poses difficulties, as, technically speaking, no model


▪50▪

implemented is close to ideal: coupling both detailed CFDs and complex

kinetic schemes is prohibitive as of now. Simulations with simplified

kinetics can take weeks even with great resources at disposal.

Figure 2.14: Multiscale modeling paradigms. Adapted from Bates [12]

This naturally stems into multiple kinds of model available reported in

Figure 2.14. Roughly, they can be distinguished in three categories:

Computational fluid-dynamics models (CFDM), fluidization models (FM)

and black-box models (BBM) [27]. A summary of the following section can

be found in Table 2.2.

Black box models

These models do not resolve the processes occurring within the reactor and

apply generalized heat and mass balances around the reactor or to zones

within the reactor. They lack kinetics and make use of assumptions

regarding chemical equilibrium or pseudo-equilibrium, or empirical

relationships.

While their predictive capability is severely limited, they can be useful for

larger-scale studies at, for example, plant-scale or supply chain level, where

anything slightly detailed can become an issue [28].


▪51▪

Computational fluid dynamics models

CFDMs have the advantage to fully resolve hydrodynamics, but due to the

associated computational cost, they are currently limited in terms of

geometric sizes and associated kinetics. 2D and 3D reactive simulations are

limited to lab scale geometries and low times, except Eulerian-Eulerian

methods which can be applied to industrial scales as long as a reduced

kinetic scheme is used. Application of CFDMs to biomass gasification is by

any means promising.

Further classification can be applied in this field, distinguishing

beforehand two approaches to fluid dynamics: one can study motion by

tracking single particles (Lagrangian models) or by studying the phase as a

continuum with the Navier-Stoke equation (Eulerian models).

Lagrangian-Lagrangian models like the lattice-Boltzmann models (LBM)

treat both solid and gas phases as particles, and coupling is done via elastic

collisions. This is a direct numerical simulation approach (DNS) for

multiphase systems and has been proven useful for deriving drag laws. It

has the highest computational expense; thus, it is not suitable for reactive

simulations.

Lagrangian-Eulerian models (LEM) treat solids as particles and gas phase

as a continuum. LEMs can be further separated into discrete element

models (DEM) and discrete particle models (DPM), according to the

Eulerian grid size implemented. In DEM, the grid size is much larger than

the particles and requires empirical closures (drag laws) for gas-solids

momentum transfer. In DPM, the grid is smaller than the particles by an

order of magnitude, so the flow on the surface is directly computed. Here,

it is possible to apply simplified kinetics, but due to the scaling with the

number of particles (for a large scale bed it is around 1012 [29]) it is limited

to small reactors. LEM has been recently applied to biomass gasifiers in

Pepiot & Desjardins [30].

In Eulerian-Eulerian models (EEM), often referred to as two-fluid models,

all phases are described as inter-penetrating continuum. Closures for solid


▪52▪

phase viscosity and normal stress are required and the kinetic theory of

granular flows (KTGF) is a popular method for this [31]. A recent

application of the Eulerian-Eulerian approach to fluidized beds is

presented in Bakshi et al.[32].

Discrete particle models (DPM) treat bubbles as lagrangian elements to

track velocity and size at each time step but applying incompressible quasi-

steady state assumption to the governing equations, greatly simplifying the

solution for both phase motions. This enables commercial scale beds to be

modelled, provided care is taken in accounting for wall effects and bubble

interactions and to the empirical relations used to describe bubble

phenomena, as the method is highly sensitive in this regard. Yang et al. [33]

recently used DPM to validate a proposed two-fluid models.

Fluidization models

Fluidization models (FM) directly assume that the bed is composed of

various phases (usually two) or regions with a predefined topology,

allowing transport of mass and heat between them. They do not solve

momentum equations but use semi-empirical correlations for fluid-

dynamic pattern and bubble dynamics.

It is worth noting that here the term phase discerns from the merely

thermodynamic definitions, meaning instead a region with a predefined

configuration (for particle mixing, flow pattern etc.).

Usually FM are 1D models, but 3D is achievable [34]. Early models treated

gas and solids as completely mixed, but over time the need of a multiphase

approach process was recognized. The two-phase theory of fluidization [35]

was the start for making this possible, as it allowed to distinguish between

bubble and emulsion phases. Now, with Davidson model [36] and empirical

correlations to estimate the division of gas between the phases, bubble

fraction and such, it is possible to obtain transport rates and degrees of

mixing. Over the years, different models showed up with different

simplifications or added detail, often taking the names of the authors or

their fundamental differences among the rest, but they are all based on the


▪53▪

same concept of estimating essential fluid dynamics information via

similar semi-empirical relationships.

In Table 2.2 the most important modeling approaches are summarized.

Reactor Network Models – Kinetic post processing

A sub-category of FM is reactor network models (RNM). They employ an

arrangement of chemical reactors in which the gas phase mixing behavior

is idealized, using 0D PSRs5 or 1D PFR6s. Ehrhardt et Al. [37] were the first

to propose a kinetic post-processing (KPP) technique to predict NOx

reburning in a pilot scale furnace.

Figure 2.15. Kinetic post-processing method

First, volumes and fluxes between the reactors (which reproduce the

hydrodynamic of the system) are calculated separately or pre-specified,

allowing for a reproducible approach with versatile input data (CFD

simulations for kinetic post processing (Ehrhardt et al. [37], empirical

correlations) and use cases. Secondly, by decoupling solution of motion

field and chemical kinetics, much more comprehensive and larger gas

5 Perfectly stirred reactor 6 Plug flow reactor

CFD Simulation

Simplified kinetic

mechanism

Steady state network solution

Detailed kinetic

mechanism

ERN Generation

TemperaturesFlow fields

Pollutants prediction


▪54▪

phase mechanisms can be tractably solved. This is schematically

summarized in Figure 2.15.

Being able to apply detailed kinetic mechanisms to complex fluid-dynamic

systems is fundamental for pollutant prediction. Reduced kinetic

mechanisms are often sufficient to predict the major species, but the

pathways for growth of tar, soot and PAH components are extremely

complex, with thousands of different reactions and a theoretical infinite

number of species, to the point that even for experimental measurements

there is the need to classify them into “bins”, or categories, based on factors

like molecular weight or number of carbon atoms in the molecule. They are

also present in traces, thus to be able to model them great level of detail is

needed. Recent application of reactor network models to biomass

combustors is found in Stark et al. [9], [38] and Bates et al. [39].


▪55▪

Table 2.2: Different modelling approaches for fluidized beds. Adapted from Gomez et al. [27]

NA

ME

&

AB

BR

EV

IAT

ION

C

ON

CE

PT

R

ES

UL

TS

A

DV

AN

TA

GE

S

DIS

AD

VA

NT

AG

ES

M

OD

EL

S U

SIN

G T

HIS

A

PP

RO

AC

H

Co

mp

uta

tio

na

l fl

uid

d

yn

am

ics

mo

de

ls

(CF

DM

)

Mo

me

ntu

m e

qu

ati

on

is

ex

pli

citl

y s

olv

ed

De

tail

ed

in

form

ati

on

of

fie

lds

in t

he

re

act

or

Use

ful

for

ex

plo

rin

g

ha

rdw

are

de

tail

s

Tim

e c

on

sum

ing

so

luti

on

DN

S:

dir

ect

nu

me

rica

l si

mu

lati

on

LE

S:

larg

e e

dd

y

sim

ula

tio

ns

DP

M:

dis

cre

te p

art

icle

m

od

els

Co

nst

itu

tiv

e r

ela

tio

ns

an

d c

losu

re l

aw

s a

re

ad

op

ted

Un

cert

ain

ty o

f p

ara

me

ters

in

clo

sure

re

lati

on

s

TF

M:

two

flu

id m

od

els

EE

M:

eu

leri

an

-eu

leri

an

m

od

els

EL

M:

eu

leri

an

-la

gra

ng

ian

m

od

els

LL

M:

lag

ran

gia

n-

lag

ran

gia

n m

od

els

Flu

idiz

ati

on

mo

de

ls

(FM

)

Tw

o p

ha

ses

tre

ate

d:

em

uls

ion

an

d b

ub

ble

, to

po

log

y a

ssu

me

d

Pro

file

s o

f sp

eci

es

an

d

tem

pe

ratu

res

Oft

en

su

ffic

ien

t fo

r e

ng

ine

eri

ng

ap

pli

cati

on

s

Ass

um

ing

flo

w s

tru

ctu

re

lim

its

ran

ge

of

ap

pli

cab

ilit

y

DH

M:

Da

vid

son

-Ha

rris

on

m

od

el

KL

M:

Ku

nii

-Le

ve

nsp

iel

mo

de

l

Th

e m

om

en

tum

eq

ua

tio

n

is n

ot

solv

ed

De

tail

ed

in

form

ati

on

bu

t li

mit

ed

co

mp

are

d t

o

CF

DM

Co

mp

rom

ise

of

pre

cisi

on

a

nd

nu

me

rica

l co

mp

lica

tio

n

No

t a

pp

rop

ria

te f

or

ha

rdw

are

de

sig

n

BE

M:

bu

bb

le a

sse

mb

lag

e

mo

de

l

Se

mi

em

pir

ica

l co

rre

lati

on

s d

esc

rib

e g

as

an

d s

oli

d f

low

pa

tte

rns

Co

rre

lati

on

s u

sed

are

sp

eci

fic

to c

on

dit

ion

s

CC

BB

MM

: co

un

ter-

curr

en

t b

ack

mix

ing

m

od

el

ER

NM

: e

qu

iva

len

t re

act

or

ne

two

rk m

od

els

Bla

ck

-bo

x m

od

els

(B

BM

)

Ov

era

ll m

ass

an

d h

ea

t b

ala

nce

Am

ou

nt

of

ga

s,

com

po

siti

on

an

d h

ea

tin

g

va

lue

Ve

ry e

asy

to

use

D

esc

rip

tio

n i

s cr

ud

e

HM

BM

: h

ea

t a

nd

ma

ss

ba

lan

ce m

od

el

EM

: e

qu

ilib

riu

m m

od

el

On

ly g

lob

al

mo

de

ls

F

ew

in

pu

ts

No

de

scri

pti

on

of

pro

cess

es

insi

de

th

e

rea

cto

r

TM

: th

erm

od

yn

am

ic

mo

de

l

PE

M:

pse

ud

o-e

qu

ilib

riu

m

mo

de

l

MZ

M:

mu

lti

zo

ne

mo

de

l

▪57▪

Chapter 3

3. NetSMOKE

Up to now, almost all the attempts at fluidization modeling through reactor

network have been applied to heterogeneous gas phase cases. Only recent

works have been applied to heterogeneous combustions due to the rise in

interest towards biomasses.

In these cases, most tried to de-couple biomass devolatilization, gas phase

conversion and char gasification into three different processes. This

approach, while yielding solid results, is not computationally efficient due

to segregation, which is the term that describes the mathematical

discretization into different sequential process rather than a single

concurrent one. Sequential substitution, often utilized to solve the global

system made of discretized processes, while computationally less

expensive in terms of raw power compares to more refined solution

procedures, is usually slow to converge and, by lacking estimation

Chapter 3 ♦ NetSMOKE

▪58▪

correctors and gradient methods to decide a solving path, might not reach

convergence at all.

Approaches called fully-coupled are usually preferable due to a much

closer resemblance to reality and the ability to implement elaborate solving

methods. Computationally speaking, they are not as efficient in terms of

raw data processing due to the need to compute a numerical jacobian

matrix for the system. On the other hand, fully-coupled methods are much

more robust than sequential substitution and reach convergence faster.

Chemical engineering problems often present large systems of stiff

equations, thus the ability to scale into large problems while retaining

robustness is vital.

The goal of this work was to set foundation for a fully-coupled reactor

network approach to heterogeneous systems, without the need to solve

different subsystems in different moments but rather study them

concurrently.

The product of this thesis is NetSMOKE, a C++ framework based on the

OpenSMOKE++ libraries which is highly modulable which can be used to

solve both homogeneous and heterogeneous reactor network models in a

fully coupled manner.

3.1 Mathematical formulation From a mathematical standpoint, solving a reactor network is nothing

conceptually complicated. On the other hand, this kind of problem can be

a numerical challenge, due to the fact that non-linear system can have

convergence problems if the first guess solution is not sufficiently close to

the real one. Also, the problem can scale easily into large number of

reactors.

Chapter 3 ♦ NetSMOKE

▪59▪

3.1.1 A reactor network according to NetSMOKE

NetSMOKE implements a very simple view of a reactor network. The

network is composed of reactors and various auxiliary units.

All the reactors are blocks where reactions take place. They are

characterized by residence time and energy condition (isothermal,

adiabatic or

Politecnico di Milano · Ai miei nonni “Nothing in life is to be feared, it is only to be...

Documents

Transcript of Politecnico di Milano · Ai miei nonni “Nothing in life is to be feared, it is only to be...