Pyrolysis of Pine Bark, Wheat Straw and Rice Husk:

Thermogravimetric Analysis and Kinetic Study

Ana Isabel Marques Ferreiro

Thesis to obtain the Master of Science Degree in

Mechanical Engineering

Supervisors: Prof. Mário Manuel Gonçalves da Costa,

MSc Miriam Estefânia Rodrigues Fernandes Rabaçal

Examination Committee

Chairperson: Prof. Viriato Sérgio de Almeida Semião

Members of the Committee: Prof. Luís António da Cruz Tarelho

MSc Miriam Estefânia Rodrigues Fernandes Rabaçal

July 2015

ii

ACKNOWLEDGEMENT

Agradeço ao Professor Catedrático Mário Costa e à Mestre Miriam Rabaçal por todo o apoio,

motivação e amizade prestados ao longo de todo o meu percurso sob as suas orientações.

Em modo particular, quero dizer ao Professor Doutor Mário Costa que tudo o que me ensinou

a todos os níveis me ajudou a crescer como pessoa e que lhe estou muito grata por isso. Agradeço

também todas as críticas, porque sempre me ajudaram a melhorar e a fazer um trabalho melhor.

Quero agradecer ainda, em modo particular, à Mestre Miriam Rabaçal por me ter ajudado a

alcançar muitas metas, que melhor que ninguém sabe quais foram. Agradeço tudo o que aprendi

contigo, que foi muito, e espero ter oportunidade de continuar a aprender. A admiração que tenho por

ti é mesmo muito grande.

Agradeço aos meus colegas e amigos David Nascimento, Duarte Magalhães, Nuno Barbas,

José Branco, Vera Branco, Gonçalo Guedes, Filipa Ferreira e João Ribau pelo apoio, convívio e bons

momentos. Vocês tornaram tudo mais fácil.

Agradeço também a amizade da Rita Maia, Mª do Céu Miranda e Manuel Pratas. São pessoas

fantásticas e a quem tenho um enorme carinho.

Por último quero agradecer à minha família, em especial à minha mãe, ao meu pai e ao meu

irmão pelo incansável apoio, ajuda e conselhos. Agradeço-vos por estarem sempre comigo e nunca

me deixarem sentir sozinha.

Por último quero agradecer ao meu namorado Filipe Pinto por todo o amor, carinho, apoio e

por ter escolhido partilhar a sua vida comigo. Fazes-me muito feliz.

iii

RESUMO

Este estudo foca-se na pirólise da casca de pinheiro, palha de trigo e casca de arroz. Foram

obtidas curvas termogravimétricas para os três tipos de biomassas, usando taxas de aquecimento de

5, 10 e 15 K/min numa atmosfera de Árgon, para investigar o efeito do tipo de combustível no processo

de pirólise sob diferentes taxas de aquecimento. Foram obtidos diferentes perfis termogravimétricos e

taxas de perda de massa, dada a composição distinta das biomassas consideradas, mas a influência

das taxas de aquecimento foi marginal. Para melhor perceber qual o impacto da composição da

biomassa no processo de pirólise, os componentes principais (hemicelulose, celulose e lenhina) de

cada tipo de biomassa foram estimados. Adicionalmente, foi utilizado um algoritmo de dois passos para

estimar os parâmetros cinéticos de um modelo de reação única (SFOM) e de um modelo de três

reações paralelas (3PM). As energias de ativação obtidas através do ajuste de cada modelo aos

resultados experimentais estão de acordo com os valores da literatura. As energias de ativação que se

obtiveram utilizando o SFOM foram 55.5, 79.6 e 87 kJ/mol para a casca de pinheiro, palha de trigo e

casca de arroz, respetivamente. Para a celulose, hemicelulose e lenhina, as energias de ativação que

se obtiveram utilizando o 3PM foram, respetivamente, 152.5, 95.7 e 44.3 kJ/mol para casca de pinheiro,

143.3, 83.6 e 37 kJ/mol para palha de trigo, e 163.8, 107.3 e 37.2 kJ/mol para casca de arroz. Para

cada biomassa, todas as taxas de aquecimento foram ajustadas com erros da ordem de 5 – 7% para o

SFOM e de ~ 2% para o 3PM. Os resultados provaram que a ferramenta cinética implementada neste

estudo é capaz de reproduzir o processo de pirólise com boa precisão e mostraram que o nível de

complexidade do 3PM é suficiente.

Keywords:

Biomassa, pirólise, termogravimetria, estudo cinético, otimização

iv

ABSTRACT

The present work focuses on the pyrolysis of pine bark, wheat straw and rice husk.

Thermogravimetric curves were obtained for the three biomass fuels for heating rates of 5, 10 and 15

K/min in an inert atmosphere of Argon to investigate the impact of the type of biomass in the pyrolysis

behavior under different heating conditions. Distinctive thermogravimetric and differential

thermogravimetric curves were obtained owing to the different composition of the biomass fuels, but the

impact of the heating rates was marginal. In order to better understand the impact of the biomass

composition in the pyrolysis, their main components were estimated (hemicellulose, cellulose and

lignin). Additionally, a two-step optimization algorithm was used to estimate the global kinetic parameters

of a single reaction model (SFOM) and of a three parallel reaction model (3PM). The activation energies

obtained by fitting each model to the experimental data are within the values reported in the literature.

The activation energies obtained using the SFOM were 55.5, 79.6 and 87 kJ/mol for pine bark, wheat

straw and rice husk, respectively. For cellulose, hemicellulose and lignin the activation energies

obtained, using 3PM, were respectively, 152.5, 95.7 and 44.3 kJ/mol for pine bark, 143.3, 83.6 and 37

kJ/mol for wheat straw, and 163.8, 107.3 and 37.2 kJ/mol for rice husk. For each biomass fuel, all heating

rates were globally fitted, with errors of the order of 5%-7% for the SFOM and of ~ 2% for the 3PM. The

results proved that the kinetic tool implemented in this work was capable of reproducing the pyrolysis

behavior with good accuracy and showed that the degree of complexity of 3PM suffices.

Keywords:

Biomass, pyrolysis, thermogravimetry, kinetic study, optimization

v

TABLE OF CONTENTS

1. Introduction ...................................................................................................................................... 1

1.1. Motivation ................................................................................................................................ 1

1.2. Previous studies ...................................................................................................................... 3

1.3. Objectives ................................................................................................................................ 9

2. Theoretical foundations ................................................................................................................. 10

2.1. Biomass: definition and characterization ............................................................................... 10

2.2. Thermogravimetry ................................................................................................................. 12

2.3. Pyrolysis kinetics ................................................................................................................... 13

2.4. Optimization methods ............................................................................................................ 17

3. Materials and methods .................................................................................................................. 19

3.1. Biomass characterization ...................................................................................................... 19

3.2. Thermogravimetric study ....................................................................................................... 21

4. Kinetic analysis .............................................................................................................................. 22

4.1. Arrhenius plot method ........................................................................................................... 22

4.2. Optimization methods ............................................................................................................ 24

4.3. Comparison between methods .............................................................................................. 28

5. Results ........................................................................................................................................... 29

5.1. Experimental results .......................................................................................................... 29

5.2. Kinetic analysis results ...................................................................................................... 31

6. Conclusions ................................................................................................................................... 37

7. Future perspectives ....................................................................................................................... 38

8. References .................................................................................................................................... 39

vi

LIST OF FIGURES

Figure 1.1: Portuguese energetic balance related to the final energy consumption in 2013 (source:

http://www.apren.pt (APREN – Associação de Energias Renováveis). .................................................. 1

Figure 1.2. Biomass conversion technologies and the correspondent primary energy products (source:

www.e-education.psu.edu). ..................................................................................................................... 2

Figure 1.3. Typical TG and DTG curves [13]........................................................................................... 4

Figure 2.1. Scheme of the TGA furnace [33]......................................................................................... 13

Figure 2.2. General scheme of the decomposition of a component. .................................................... 14

Figure 2.3. General scheme for secondary reactions of the vapor phase of tars. ................................ 15

Figure 2.4. Chemical structure of all the components (source: Cellulose [12, 39]; hemicellulose [12,40];

lignin [38]). ............................................................................................................................................. 15

Figure 2.5. Multistep mechanism of cellulose pyrolysis [38]. ................................................................ 16

Figure 2.6. Multistep mechanism of hemicellulose pyrolysis [38]. ........................................................ 16

Figure 2.7. Multistep mechanism of the pyrolysis of LIG-H (top), LIG-O (mid) and LIG-C (bottom) [38].

............................................................................................................................................................... 16

Figure 2.8. Genetic Algorithm: scheme of the three types of children [42]. .......................................... 17

Figure 3.1. Triangulation method test for the biomass samples studied. .............................................. 20

Figure 3.2. Contents of cellulose, hemicellulose and lignin. ................................................................. 20

Figure 3.3. Temperature profiles for the three heating rates tested. ..................................................... 21

Figure 4.1. Representative example of the FWO method [49]. ............................................................. 23

Figure 4.2. Pyrolysis algorithm. ............................................................................................................. 25

Figure 4.3. Kinetic optimization procedure. ........................................................................................... 26

Figure 5.1. Pyrolysis yields for pine bark (PB), wheat straw (WS) and rice husk (RH) for 5, 10 and 15

K/min. ..................................................................................................................................................... 29

Figure 5.2. DTG curves as a function of temperature for pine bark (PB), wheat straw (WS) and rice husk

(RH) for 5, 10 and 15 K/min................................................................................................................... 30

Figure 5.3. Arrhenius plot method applied to experimental TG curves of pine bark (PB), wheat straw

(WS) and rice husk (RH), respectively. ................................................................................................. 31

Figure 5.4. TG (top), DTG (bottom) and predicted curves for pine bark (PB). ...................................... 34

Figure 5.5. TG (top), DTG (bottom) and predicted curves for wheat straw (WS). ................................ 34

Figure 5.6. TG (top), DTG (bottom) and predicted curves for rice husk (RH). ...................................... 35

Figure 5.7. DTG and predicted curves using the 3PM for pine bark (PB), wheat straw (WS) and rice

husk (RH) at 5 K/min. ............................................................................................................................ 35

vii

LIST OF TABLES

Table 1.1. Summary of the most relevant studies (continues). ............................................................... 6

Table 2.1. Chemical characteristics and structural components composition of various biomass groups

and sub-groups [24]. .............................................................................................................................. 11

Table 3.1. Proximate and ultimate analysis of the biomass fuels studied. ............................................ 19

Table 3.2. Composition range of the correlation method [45]. .............................................................. 20

Table 3.3. Comparison between the estimated composition of the three main components and literature

values [12,21,46]. .................................................................................................................................. 21

Table 4.1. Typical range of activation energies for the SFOM [13, 22, 47, 53]. .................................... 27

Table 4.2. Typical kinetic parameters for the 3PM and fraction of char produced by each component [12,

14, 20]. ................................................................................................................................................... 27

Table 4.3. Average simulation times and specific standard deviations ................................................. 27

Table 4.4. Work computer specifications. ............................................................................................. 27

Table 5.1. Characteristics of the first region of decomposition ............................................................. 30

Table 5.2. Kinetic parameters estimated with the Arrhenius plot method. ............................................ 31

Table 5.3. Kinetic parameters obtained by fitting the SFOM to the experimental data for the three

biomass fuels. ........................................................................................................................................ 32

Table 5.4. Activation energies and char fractions obtained by fitting the 3PM to the experimental data

for the three biomass fuels. ................................................................................................................... 32

Table 5.5. Pre-exponential factors and model constants obtained by fitting the 3PM to the experimental

data for the three biomass fuels. ........................................................................................................... 33

Table 5.6. Comparison between the Arrhenius plot method and the fitting procedure. ........................ 36

viii

NOMENCLATURE

Symbols

A Pre-exponential factor

C Carbon

C𝑖 Fraction of char produced by the ith component

E Activation energy

f(∝) Conversion function

f(E) Funtion of the distribution of the activation energy

g(∝) Integral function of conversion

H Hydrogen

k(T) Reaction rate

𝑁 Number of iterations

O Oxygen

R Ideal gas constant

T Temperature

𝑈𝑒𝑥𝑝 Experimental results

𝑈𝑝𝑟𝑒𝑑 Predictive results

VM Volatile matter

Vg Total amount of volatile gases released from particle

Greek Letters

α Conversion degree

β Heating rate

𝛿 Fitting error

ε Activation energy threshold

σDEV Standard deviation

γ Model constant

λ Shape parameter

η Width parameter

Acronyms

3PM Three parallel model

CELL Cellulose

daf dry, ash free

db dry basis

DSC Differential scanning calorimetry

DTG Differential thermogravimetric

GA Genetic algorithm

ix

HAB Herbaceous and agricultural biomass

HAR Herbaceous and agricultural straws

HCE Hemicellulose

LIG Lignin

LSQ Least squares

PB Pine bark

RH Rice husk

SFOM Single first order reaction model

TG Thermogravimetric

TGA Thermogravimetric analysis

WS Wheat straw

WWB Wood and woody biomass

1

1. INTRODUCTION

1.1. MOTIVATION

For the past few decades, fossil fuels have been chosen to fulfill world energy demands,

but since these resources are limited and generally have a negative impact on the environment,

recent dramatic developments in renewable energy production occurred. These are supported by

policies aiming to change the energy mix, especially for electricity production [1]. Biomass has a

major advantage over other renewable energy sources, as it can be stored and used on demand

to give controllable energy. It is therefore free from the weather conditions intermittency, a

problem for all other forms of renewable energy [2-4]. Bioenergy source is any fuel derived from

biomass – recently living organisms or their metabolic byproducts. Being highly available and

diverse, biomass is becoming a promising renewable source due to its capability to be linked with

many economic sectors like agriculture, forestry, food processing, paper and pulp and, of course,

the energy sector [5]. In order to support the growth of bioenergy, biomass supply has to grow as

well, but not all the available biomass from forests and fields can be removed. Agricultural and

forest residues and energy crops planted on idle or released cropland are an attractive alternative

[4, 6]. Additionally, biomass can be upgraded through conversion processes, like pyrolysis, to

increase its value as a fuel.

Figure 1.1 shows the Portuguese energetic balance related to the final energy

consumption in 2013 and biomass only contributes with 6.6%. The energy policies in Portugal

have been targeting mainly the development of solar and wind power [5], which may explain the

low share of biomass in the energy mix. Much work needs to be done in the Portuguese bioenergy

sector, however, an important limitation can be the investment costs and in the specific case of

Portugal, there is still lack of governmental support and incentives to ensure the interest from

private investors in bioenergy technologies.

Figure 1.1: Portuguese energetic balance related to the final energy consumption in 2013

(source: http://www.apren.pt (APREN – Associação de Energias Renováveis).

2

Figure 1.2 shows the biomass conversion technologies. The final products of the

conversion technologies are used mainly for the production of heat, power, fuels and chemicals,

where the unconverted residues can be used for soil amendment. Pyrolysis can be found in the

group of thermochemical conversion, along with gasification.

Figure 1.2. Biomass conversion technologies

(Source: http://www.extension.org/pages/26517/woody-biomass-properties#.VaDi9PlVhBc).

Pyrolysis is a form of thermochemical treatment that decomposes organic materials into

liquid, solid and gaseous forms in the absence of oxygen. Due to its versatility, pyrolysis is

becoming a more relevant process since all the three-output fractions have potential as fuels for

transports, power generation and combined heat and power [7]. Fast pyrolysis, in particular, is a

relatively mature technology in the verge of commercialization [8], mostly located in the North of

Europe [9-11]. Portugal however, despite showing great potential to adopt this technology with

an estimated dry biomass production in the order of 5630 thousand tons, has yet to start

developing this industry [27]. The most common reactors used in industrial facilities for the

pyrolysis process are the fluidized bed reactors and the rotating cone reactors for the production

of bio-oil. In fluidized bed reactors the biomass feedstock needs a pre-treatment that involves

drying (< 10% moisture), milling and/or sieving (particle size between 1 and 2 mm). This feedstock

is then fed into the reactor, generally with the aid of a screw, heated to approximately 500 ºC in

the absence of oxygen and decomposed into gaseous vapors and char particles, which are

separated in the cyclone chamber. At this point, charcoal is collected while the gases move on

into the condensers turning them into bio-oil. The non-condensable gases are recirculated into

the main reactor chamber to be re-used for pre-heating. In rotating cone reactors, the pyrolysis

3

process occurs in the rotating cone while mixes biomass particles and hot sand, in the absence

of oxygen. The charcoal and the sand resultant from the previous stage are recycled into the

combustion chamber where charcoal is burned to reheat the sand. The gaseous vapors are led

to the condenser yielding the bio-oil. The non-condensable gases and the extra heat coming from

the chamber can be used to generate steam for power generation or for drying the biomass. Some

of the industrial facilities adopted these technologies to produce electricity, bio-oil and district

heating. For instance, Fortum, founded in Joensuu (Finland) in 2013 [9], is a combined heat and

power production plant (CHP) that produces electricity, district heating and also aims to produce,

in the future, 50,000 tons of bio-oil per year from the conversion of lignocellulosic biomass in a

fluidized bed reactor. Dynamotive Energy Systems Corporation [10] in Canada that is a leader in

the bio-oil production also via a fluidized bed reactor, while BTG – Biomass Technology Group

[11] in Netherlands is using rotating cone reactor to produce bio-oil.

As most biomass upgrading processes, pyrolysis plants are typically optimized to woody

biomass [12]; however, other types of biomass sources need to be considered as discussed

above. But the use of “difficult” biomass fuels can bring complications in the operating system of

these reactors, mostly operational problems caused by high ash content typically found in

agricultural biomass, and yield/composition of the obtained products [8]. In this context, there is

a need to better understand the pyrolysis process of non-woody biomass through fundamental

research. Thermogravimetric studies and kinetic analysis can provide a better understanding of

the governing processes of the pyrolysis of alternative biomass.

1.2. PREVIOUS STUDIES

Di Blasi [13] made an extensive review of a significant number of studies focused on

modeling of the biomass pyrolysis process based on thermogravimetric studies. In this review the

author describes one component and multiple component mechanisms. The assumption of a

single component behavior, inevitably introduces imprecisions in the decomposition rates (and

conversion time), since it considers only one zone of decomposition. Multi-component

mechanisms of biomass pyrolysis have been developed to describe different decomposition

zones, based on the pseudo-components hemicellulose, cellulose and lignin that compose

biomass. Thermogravimetry provides one single curve of mass loss (and rate) over the residence

time and the different components specific decomposition will overlap in this single curve. Figure

1.3 shows a typical thermogravimetric (TG) and differential thermogravimetric (DTG) curves.

Hemicellulose and cellulose are associated with the shoulder and the peak of the DTG curves,

respectively, meaning that the rate curves of these two pseudo components overlap each other

during their decomposition process. On the other hand, lignin decomposes slowly over a very

broad range of temperatures. The accuracy in the predictions of weight loss characteristics is

improved as the number of model parameters is increased. However, simplicity is always desired

for the global reaction mechanisms. Di Blasi [13] pointed out that the application of these kinetic

models to the study of pyrolysis/devolatilization processes has been mostly concentrated on wood

species and, in a small number of cases, on agricultural residues. Therefore, it is necessary to

4

pursue the development of general reaction schemes that can be applied for other biomass fuels

and broaden the current application. Given the high variety in chemical composition among the

different species, this matter is of particular practical importance.

Figure 1.3. Typical TG and DTG curves [13].

In the remainder of this section, a number of relevant studies will be reviewed. Focus was

given to studies that in their kinetic analysis implemented a single first order reaction (SFOM),

considering only one stage of decomposition, a three-parallel model (3PM) that describes the

global decomposition of cellulose, hemicellulose and lignin and a five parallel model (5PM) that

adds extractives. Table 1.1 summarizes previous relevant studies, where the type of biomass,

heating rates (β), mechanisms used, effects evaluated and main conclusions are listed. A

discussion of the estimation methods and mechanisms used, as well as the effect of the main

controlling parameters on the pyrolysis behavior is presented. Several studies [12, 14-16], among

others, estimated activation energies (Ea) using different methods and the estimated values are

discussed in section 4.2.

Estimation methods and mechanisms used

The Arrhenius plot method is a linear regression method able to estimate the kinetic

parameters (activation energy and pre-exponential factor) of one or multiple conversion stages,

of one single component [15-17]. The non-linear squares procedure is also an estimation method

that appeared as an alternative of the Arrhenius plot to estimate the kinetic parameters with better

accuracy [12, 14, 18-21], since it considers all the information obtained from the experimental

data.

SFOM mechanisms are usually applied to describe the decomposition of one single

component assuming one or multiple stage conversion [15-17]. Additionally, some authors have

included the prediction yields (gas, tar and char) when products were measured experimentally

5

[15, 16]. Slightly more complex mechanisms are used for a more detailed conversion description

by considering the thermal decomposition of multiple components such as 3PM and 5PM, used

by Grønli [14], Vamvuka et al. [20], Li et al. [19], Damartzis et al. [18] and Burhenne et al. [12].

These mechanisms are frequently referred to as multi-parallel reactions models, since it is

assumed that each reaction occurs independently of others. However, with the exception of Grønli

et al. [14], Vamvuka et al. [20], Li et al.[19], Damartzis et al. [18] and Burhenne et al. [12] measured

products yields, but this aspect was not considered in the mechanism.

Heating rate effects

Most of the referred authors studied the effects of the heating rates on the pyrolysis

behavior. Sharma et al. [22], Li et al. [19], Seo et al. [15], Mani et al. [21] and Damartzis et al. [18]

concluded that transition temperatures slightly increase with increasing heating rate, shifting the

DTG profiles. These constituents of biomass have characteristic individual decomposition peaks

in certain temperature ranges and it has been showed that an increase of the heating rate tends

to delay thermal decomposition processes towards higher temperatures [12, 14, 18, 19, 21].

Furthermore, at higher heating rates the distinct peaks associated with the different constituents

may not appear because some of them can be thermally decomposed simultaneously,

overlapping each other in DTG profiles. This behavior was observed by Grønli et al. [14],

Vamvuka et al. [20], Li et al. [19], Seo et al. [15], Mani et al. [21], Burhenne et al. [12] and Guerrero

et al. [16].

Seo et al. [15], Damartzis et al. [18] and Guerrero et al. [16] concluded that lower heating

rates lead to higher volatile matter production due to the fact that when heating rate decreases,

the required time to reach a certain temperature value increases, enabling other chemical

reactions to occur (for example, dehydration). As a consequence, the amount of devolatilized

matter is increased.

Sharma et al. [22] and Damartzis et al. [18] studied the effect of the heating rate in the

activation energy. The former authors concluded that the increasing of the heating rate leads

generally to the increase of the activation energy. The latter authors concluded that a higher

heating rate increased the activation energy values for hemicellulose and cellulose

decomposition, whereas decreased the activation energy values for lignin decomposition.

6

Table 1.1. Summary of the most relevant studies (continues).

Reference Sample β (K/min) Mechanism Studied effects Conclusions

Sharma et al. [22] Rice husk 5 – 100 Arrhenius plot method Multi stage conversion SFOM

Heating rates Size particle Reaction order

Ea varies with stage and heating rate Similar results independently of the particle size Variable reaction order gave the best results

Grønli et al. [14] Alder beech Birch oak Douglas fir Pine A Pine B Redwood Spruce

30 Least squares 5PM

Fuel type Overlapping of cellulose, hemicellulose and extractives

3PM suffices for hardwoods Softwoods need 5PM due to extractives Fixing Ea for all the components predicted good

results

Vamvuka et al. [20] Olive Kernel Straw

10 Least squares 3PM

Particle size Fuel type

Particle size had marginal influence on volatile and char yields (particle size between 250 µm and 1000 µm).

Overlapping of cellulose, hemicellulose

Li et al. [19] Corn straw 20 – 100 Least squares 3PM

Heating rate Reaction order

Different DTG profiles when heating rate changed Overlapping of cellulose and hemicellulose Maximum pyrolysis rate and temperature peak

increased with increasing heating rate Reaction order different of one predicted overall

better results Similar Ea for each pseudo-components regardless

the heating rates Ci and A parameters varied with the heating rate

variation.

7

Table 1.1. Summary of the most relevant studies (continued).

Reference Sample β (ºC/min) Mechanism Studied effects Conclusions

Seo et al. [15] Sawdust 5 – 30 Arrhenius plot method Multi stage conversion SFOM

Heating rates Overlapping of cellulose and hemicellulose Smaller heating rates allow higher volatile matter

production

Mani et al. [21] Wheat straw 5 – 20 Least squares 3PM

Heating rates Particle size

Maximum pyrolysis rate and temperature peak increased with increasing heating rate

Overlapping of cellulose and hemicellulose Lignin decomposed in a wide range of temperatures The char yield increased as the particle size and

heating rate of the pyrolysis process increased

Damartzis et al. [18] Cardoon 5 – 30 Arrhenius plot method Least squares 3PM

Heating rates Estimation methods

Higher heating rates delay thermal decomposition processes

Thermal decomposition rates increased with the heating rate

Hemicellulose and cellulose react at low temperatures

Lignin decomposed in a wide range of temperatures Higher heating rate increases Ea of hemicellulose

and cellulose Higher heating rate decreases Ea of lignin

Burhenne et al. [12] Wheat straw Rape straw Spruce + bark

20 Least squares 3PM

Fuel type Lignin is the main controlling factor of pyrolysis process

Woody biomass needs more energy to decompose than agricultural biomass

Guerrero et al. [16] Aple pomace 5 – 20 Arrhenius plot method SFOM

Heating rates Particle size

Overlapping of cellulose and hemicellulose Thermal decomposition rates increased with the

heating rate Smaller particles allow higher volatile matter

production

8

Particle size

Sharma et al. [22], Vamvuka et al. [20], Mani et al. [21] and Guerrero et al. [16] studied the

effects of the particle size. Sharma et al. [22] and Vamvuka et al. [20] observed that particle size,

between powder and grains and in the range of 250 μm to 1000 μm, had practically no influence on the

pyrolysis process. Mani et al. [21] studied particles within the range of 150 to 1350 μm; similar results

were obtained except for the smaller particles (150 and 250 μm) that showed a different pattern in the

TG profile at high temperatures. On the other hand, Guerrero et al. [16] used particles between 150 and

425 μm and concluded that the smaller particles are easy to degrade so they allow the generation of a

greater amount of volatile matter. However, the variation of the volatile matter with particle size was

marginal.

Fuel type

Studies that do not include the explicit study of the component composition [15-17, 21] can only

focus on differences in TG curves relatively to the product yields (gas, tars and char) and only one set

of kinetic parameters is estimated to describe the entire pyrolysis process. Overall, the activation energy

varies with the type of fuel, as well as the relative amounts of products.

Grønli et al. [9], Vamvuka et al. [20] and Burhenne et al. [12] studied the effetcs of the

composition of the biomass fuel and conclude that different amounts of each component lead to

distinctive pyrolysis peaks and width. Additionaly, Grønli et al. [9] and Burhenne et al. [12] showed that

the lignin content of any biomass feedstock is the main controlling factor in pyrolysis, since it shows a

slower decomposition. Furthermore, cellulose and hemicellulose are typically fully consumed and the

residual char comes from lignin. Grønli et al. [9] and Burhenne et al. [12] opted to fix the activation

energies of each pseudo-component for all the biomass fuel and obtained consistent predictions of the

DTG curves. Di Blasi [13] underlines that this is a valid procedure when heating rates do not vary

significantly. When moving to much higher heating rates a new set of activation energies needs to be

estimated. Grønli et al. [9] observed that when amount of extractives is high the transition of 3PM to

5PM is required.

Reaction order

Only Sharma et al. [22] and Li et al. [19] studied the effect of the reaction order. In both studies

it was concluded that variable reaction orders produce better results. For instances, the former authors

found better results below 350°C when considering a reaction order of 1.5, and above this temperature

a value of 2.0. The latter authors obtained variations between 1.3 and 3.7. However it is common

practice to assume a reaction order of one and typically this assumption leads to good predictions of the

pyrolysis behavior, as pointed out by Di Blasi [13] and seen in most of the previous studies [12, 14, 18,

20].

9

1.3. OBJECTIVES

The general objective of this work is to study experimentally and kinetically the pyrolysis of pine

bark, wheat straw and rice husk. The specific objectives are as follows:

Perform thermogravimetric experiments in a TGA to isolate the chemical kinetics from the

transport phenomenon;

Investigate the impact of the type of biomass in the pyrolysis behavior under different heating

conditions.

Estimate the activation energy of the pyrolysis reaction using the Arrhenius plot method.

Develop an optimization tool capable of fitting pyrolysis model predictions to experimental

curves in order to better estimate the kinetic parameters.

Test how complex should the kinetic model be to describe the pyrolysis process of a biomass

sample, by comparing SFOM and 3PM predictions to experimental curves.

This study contributes to filling a gap still existing in the kinetic modeling research of non-woody

biomass. Furthermore, the tool has the capability to be applied to a variety of biomass types, making it

a useful pre-processing approach that can be used to estimate the kinetic parameters of sub-models

used in complex numerical simulations of pyrolysis reactors operating under similar heating conditions.

10

2. THEORETICAL FOUNDATIONS

2.1. BIOMASS: DEFINITION AND CHARACTERIZATION

Biomass is biological material derived from living, or recently living organisms. It includes plants,

leftovers from agricultural materials and forestry processes, as well as organic industrial, animal and

human wastes [23]. Biomass has highly variable properties, especially with respect to moisture,

structural components and inorganic constituents. This is directly related to the conditions of biomass

growth, i.e., the properties of the soil that are associated with the location site, weather conditions and,

especially, with use of chemicals (pesticides) and fertilizers [4]. It is composed mostly by Carbon (C),

Hydrogen (H), Oxygen (O), and other minor elements like Nitrogen (N), Sulphur (S), Calcium (Ca),

Potassium (K), Silicon (Si), Magnesium (Mg), Aluminum (Al), Iron (Fe), Phosphorus (P), Chlorine (Cl),

Sodium (Na), Manganese (Mn), Titanium (Ti) that are generally present in the form of oxides [24]. The

structural organic components are the hemicellulose, cellulose and lignin, from where the designation

of lignocellulosic biomass was originated, and additionally extractives that may comprise organic and

inorganic matter in their composition.

Hemicellulose is described as a complex mixture of various polysaccharides (xylose, mannose,

glucose, galactose, some acids, etc.), or a macromolecular substance of different sugars. Hemicellulose

appears to have an irregular and amorphous structure, rich in branches that are very easy to degrade

to volatiles (CO, CO2, some hydrocarbons, etc.) at low temperatures [24-26]. Cellulose forms long

glucose polymeric chains bounded to each other by hydrogen bounds, with no branches, making its

structure very strong with high thermal stability. Lignin is a non-sugar polymer mainly composed of

aromatic rings, alcohols and some acids. Lignin has a highly branched and irregular structure and its

very chemically active, which turns its decomposition process very difficult within a wide range of

temperatures (100-900 ºC) [14, 24, 26]. Extractives are defined as those compounds that are not an

integral part of the biomass structure and their composition includes various saccharides and other

carbohydrates, proteins, hydrocarbons, oils, aromatics, lipids, fats, starches, phenols, waxes and

inorganic materials, which can be extracted using different solvents (water, ethanol, benzene, etc.) [25].

There are different groups of biomass which can be classified in (1) wood and woody biomass

(WWB), (2) herbaceous and agricultural biomass (HAB) including straws (HAS) and residues (HAR), (3)

aquatic biomass, (4) animal and human biomass wastes, (5) contaminated biomass and industrial

biomass wastes (semi-biomass) and (6) biomass mixtures (blends from the previous varieties). Table

2.1 shows the chemical characteristics and the structural components composition of various biomass

groups and sub-groups of 47 species.

11

Table 2.1. Chemical characteristics and structural components composition of various biomass groups

and sub-groups [24].

Biomass WWB mean HAB mean HAS mean

Proximate analysis (wt.%, ar)

Moisture 4.7 – 62.9 19.3 4.7 – 62.9 19.3 7.4 – 16.8 10.2

Volatile Matter 30.4 – 79.7 62.9 41.5 – 76.6 66.0 58.0 – 73.9 66.7

Fixed Carbon 6.5 – 24.1 15.1 9.1 – 35.3 16.9 12.5 – 17.8 15.3

Ash 0.1 – 8.4 2.7 0.1 – 8.4 2.7 4.3 – 18.6 7.8

Ultimate analysis (wt.%, daf)

C 48.7 – 57.0 52.1 42.2 – 58.4 49.9 48.5 – 50.6 49.4

O 32.0 – 45.3 41.2 34.2 – 49.0 42.6 40.1 – 44.6 43.2

H 5.4 – 10.2 6.2 3.2 – 9.2 6.2 5.6 – 6.4 6.1

N 0.1 – 0.7 0.4 0.1 – 3.4 1.2 0.5 – 2.8 1.2

S 0.01 – 0.42 0.08 0.01 – 0.60 0.15 0.08 – 0.28 0.15

Structural components (wt.%, daf)

Cellulose 12.4 – 65.5 39.5 23.7 – 87.5 46.1 18 – 54.8 45.4

Hemicellulose 6.7 – 65.6 34.5 12.3 – 54.5 30.2 18 – 39 31.5

Lignin 10.2 – 44.5 26.0 0.0 – 54.3 23.7 14.9 – 35.3 23.1

Extractives (wt.%, daf) 1.0 – 9.9 3.1 1.2 – 86.8 13.7 3.8 – 21.7 13.6

Ash analysis (wt.%, db)

Cl 0.01 – 0.05 0.02 0.04 – 0.83 0.21 0.03 – 0.64 0.41

SiO2 1.86 – 68.18 22.22 8.73 – 84.92 46.18 7.87 – 77.2 43.94

CaO 5.79 – 83.46 43.03 2.98 – 44.32 11.23 2.46 – 30.68 14.13

K2O 2.19 – 31.99 10.75 2.93 – 53.38 24.59 12.59 – 38.14 24.49

P2O5 0.66 – 13.01 3.48 3.14 – 20.33 6.62 0.98 – 10.38 4.13

Al2O3 0.12 – 15.12 5.09 0.67 – 2.59 1.39 0.1 – 5.57 2.71

MgO 1.1 – 14.57 6.07 1.42 – 8.64 4.02 1.67 – 14.1 4.66

Fe2O3 0.37 – 9.54 3.44 0.58 – 1.73 0.98 0.41 – 2.82 1.42

SO3 0.36 – 11.66 2.78 0.83 – 9.89 3.66 1.18 – 4.93 3.01

Na2O 0.22 – 29.82 2.85 0.09 – 6.2 1.25 0.16 – 3.52 1.35

TiO2 0.06 – 1.2 0.29 0.01 – 0.28 0.08 0.02 – 0.33 0.16

Mn (ppm) 775 – 35740 13160 – 3100 155 – 2790 865

* am – as measured, daf – dry ash free basis

A detailed analysis of Table 2.1 reveals that, relatively to the weight percentage within the

referred groups, the decreasing order is as follows:

Moisture and volatile matter is WWB > HAB > HAS

Carbon is WWB > HAB > HAS

Oxygen is HAS > HAB > WWB

Hydrogen is (WWB, HAB) > HAS

12

Cl and SiO2 is HAS > HAB > WWB

CaO, Al2O3, MgO, Fe2O3, Na2O, TiO2 and Mn is WWB > HAB > HAS

K2O, P2O5 and SO3 is HAB > HAS > WWB

As for the structural components and extractives all the groups mentioned show, relatively to

the weight percentage within the referred groups, the following decreasing order:

Cellulose is HAB > HAS > WWB

Hemicellulose is WWB > HAS > HAB

Lignin is WWB > HAB > HAS

Extractives is HAS > HAB > WWB.

HAS presents the higher amount of ash, which can reach to 18.6%. This brings up the matter

of the disposability of biomass ash. The complex character of this parameter is the reason for such a

problem because ash originates simultaneously from inorganic, organic and fluid matter during biomass

conversion [24]. Table 2.1 shows that the ash composition strongly varies with the group or sub-group

where biomass is inserted. Nitrogen, Sulphur and Chlorine, despite contributing with small amounts for

the biomass composition, are the main responsible for the pollutants emissions [27]. For instance, when

the ratio between Chlorine and Sulfur contents is superior to one indicates a highly corrosive potential,

by active oxidation mechanism [28]. This indicator presents higher values in the HAB group. Nitrogen,

which is an essential element for the biomass growth, is responsible for the emission of NO and NO2,

also a characteristic more evident in the HAB group [29]. The K2O content is also higher in the HAR

sub-group, when compared with other biomass, which is attributed to the use of fertilizers [30]. The

presence of high ash content usually gives rise to higher particle emissions and fouling and deposits on

the surfaces where heat transfer occurs [31].

2.2. THERMOGRAVIMETRY

Thermogravimetry is a technique that can measure both heat flows and weight changes that

occur in a material as a function of temperature and time in a controlled atmosphere [15, 16, 32, 33].

This technique is usually referred as differential scanning calorimetry – thermogravimetry (DSC-TGA).

This combination allows identifying endothermic and exothermic events that can be associated to weight

losses, like melting or decomposition. In particular, thermogravimetric analysis is a very useful technique

for determining (1) composition of multicomponent systems, (2) atmosphere effects on materials, (3)

reaction kinetics and (4) ash, moisture and volatile contents of materials, being for this reason a very

powerful tool in the study of biomass thermal conversion processes, like pyrolysis.

Figure 2.1 shows a schematic diagram of a TGA furnace. The procedure is simple: A small

amount of the sample (~ 5 mg) is placed on a crucible supported by a precision balance inside a high

temperature furnace. Specifically, for the pyrolysis process, the atmosphere in the furnace must be inert,

where generally it is used Nitrogen or Argon. The temperature is measured with a thermocouple placed

near the crucible (see Figure 2.1). The information regarding to temperature and mass variations is send

to a computer unit to process the data in the form of thermogravimetric (TG) and differential

thermogravimetric (DTG) curves. Due to the considered low amount of the sample, this process should

be repeated several times to obtain a representative pyrolysis behavior. These curves represent the

13

mass variation and the rate of mass loss, respectively, with temperature and allow to identify the main

reactions involved in the pyrolysis process and to estimate its correspondent kinetic parameters. This

process is limited since the number of reactions occurring simultaneously during a simple pyrolysis

process is greater than the ones that can be identified, thus, pyrolysis is typically studied using models

in which the overall pyrolysis behavior is considered as the combination of each individual component

[16].

Figure 2.1. Scheme of the TGA furnace [33].

2.3. PYROLYSIS KINETICS

The pyrolysis kinetics is a mean of analyzing how the thermal conversion evolves through the

study of reaction rates, order of reaction and other parameters that can influence those rates [34]. The

heating rate can affect the distribution of the pyrolysis products and therefore the values of activation

energy may also differ. Consequently, the reaction rate can vary as well. However, many authors have

obtained good results with first order [12, 14-16]. Di Blasi [13] noticed that activation energies are higher

when it is assumed first-order reactions. The conversion can be described by one single process or

multiple, depending on the species taken into consideration. Empirical and predictive models have

already been developed to study the process of pyrolysis. Empirical models are based in apparent

kinetics, using estimation methods to find global kinetic parameters, while predictive models are

developed to be able to describe in detail the decomposition process using multiple species without any

fitting procedure.

There are several empirical pyrolysis kinetic models with distinctive levels of complexity already

discussed in the section 1.2, with distinctive levels of complexity available in the literature. The simplest

one is the single first order reaction model (SFOM) [13, 15, 16, 22, 35] that considers only one stage of

decomposition. Another one is the three parallel reactions model (3PM) [12-14, 18] that describes the

global decomposition of cellulose, hemicellulose and lignin. The decomposition of extractives can be

added, as in the work of Grønli et al. [9], who considered five parallel reactions (5PM). The reaction rate

constant of each component, is described by the Arrhenius law:

ki(Tp) = AiTpγ

exp (−Ei

RTp

) (2.1)

14

where ki(Tp) is the reaction rate constant of the ith component, Ai the pre-exponential factor (s-1), Ei the

activation energy (kJ/mol) and R is the ideal gas constant (J.K-1mol-1) [35].

A slightly more complex model is the distributed activation energy model (DAEM) [14,36,37].

This approach can be applied to the total amount of volatiles released or simply to the volatiles released

from a single component. Like the previous models, DAEM considers the Arrhenius law, but allows

expressing the distribution of the activation energy in a Gaussian form as:

𝑓𝑖(𝐸) =1

𝜎𝑖2𝜋1 2⁄exp (

−(𝐸 − 𝐸𝑖,0)2

2𝜎𝑖2 ) (2.2)

where 𝐸 is the activation energy (J/mol), 𝐸𝑖,0 the mean activation energy (J/mol) and σ the standard

deviation (J/mol) [37]. But this distribution is symmetric and, since the asymmetry of reactivity

distributions have to come into consideration, the Weibull distribution is used in the form:

𝑓(𝐸) =𝜆

𝜂(

𝐸 − 휀

𝜂)

𝜆−1

exp [(𝐸 − 휀

𝜂)

2

] (2.3)

where λ is the shape parameter, η is the width parameter, and ε is the activation energy threshold

(E ≥ ε) [36].

There are also some multi-component mechanisms that are able to predict products yields of

the three main components of biomass (hemicellulose, cellulose and lignin). Di Blasi [13] proposed this

model based on an extensive examination of literature data. Figure 2.2 shows a general scheme of the

decomposition of a component. νc is the volatile matter.

Figure 2.2. General scheme of the decomposition of a component.

The first step of the decomposition (depolymerization) does not lead to chemical composition

changes, but to modify the physical properties, as porosity. This scheme was originally developed for

cellulose, but also can be applied to hemicellulose and lignin. Extractives and ash contents are

integrated in the hemicellulose component. This mechanism can also consider secondary reactions of

the vapor phase of tars (see Figure 2.3) that account the complete decomposition of tars.

15

Figure 2.3. General scheme for secondary reactions of the vapor phase of tars.

The Bio-PoliMi mechanism is a predictive multistep devolatilization model that considers 43

species and 14 chemical reactions to describe, in detail, the devolatilization process of cellulose,

hemicellulose and lignin, including product speciation. Due to the complexity of lignin structure, were

differentiated three structures identified as LIG-C, LIG-O and LIG-H, being respectively rich in carbon,

oxygen and hydrogen [38]. Figure 2.4 shows the chemical structure of all the components.

Figure 2.4. Chemical structure of all the components (source: Cellulose [12, 39]; hemicellulose [12,40];

lignin [38]).

Figure 2.5 shows the multistep mechanism of cellulose pyrolysis. The devolatilization of

cellulose involves multiple reactions that lead to the formation of the levoglucosan, decomposition

products, char and water. The volatile products include CO, CO2, CH4, CH3CHO, C3H6O, among others.

16

Figure 2.5. Multistep mechanism of cellulose pyrolysis [38].

Figure 2.6 shows the multistep mechanism of hemicellulose pyrolysis. The devolatilization of

hemicellulose involves multiple reactions that lead to the formation of xylose, char and other

decomposition products. The volatile products include CO, CO2, CH4, CH2O, C2H5OH, CH2OH, among

others.

Figure 2.6. Multistep mechanism of hemicellulose pyrolysis [38].

Figure 2.7 shows the multistep mechanism of the pyrolysis of the three lignin reference

components. The devolatilization of hemicellulose involves multiple reactions that lead to the formation

of intermediate lignin species, char and other decomposition products. The volatile products include CO,

CO2, H2, CH4, CH2O, phenols and some acids, among others.

Figure 2.7. Multistep mechanism of the pyrolysis of LIG-H (top), LIG-O (mid) and LIG-C (bottom) [38].

17

2.4. OPTIMIZATION METHODS

The purpose of this work was not to implement new optimization techniques or improving

existing ones but to use and combine already existing techniques to perform kinetic analysis. For this

reason, the various optimization methodologies will not be extensively reviewed here. For more

information, the reader is referred to [41, 42]. In this work, the genetic algorithm (GA) and the non-linear

curve fitting (LSQ) MATLAB solvers were used.

A genetic algorithm (GA) is a method for solving optimization problems based on a randomly

selection process that mimics the process of natural selection. A typical genetic algorithm requires: (1)

a genetic representation of the solution domain and (2) a fitness function to evaluate the solution domain.

The GA has the objective to find global minima for nonlinear problems through an iterative process. In

each iteration, individuals (in this case kinetic parameters) are selected from the current population and

uses them as parents to produce the children for the next generation [40]. Elite children are the

individuals in the current generation with the best fitness values. So they go automatically to the next

generation. The other children’s genome are modified (through crossover and randomly mutation, see

Figure 2.8) to form a new generation. The algorithm terminates when it reaches the stopping criteria

defined by the user (maximum number of generations, stall generation number, etc.) or if it reaches the

optimal solution, i.e., convergence of the error to the minimum value [41, 42].

Figure 2.8. Genetic Algorithm: scheme of the three types of children [42].

GA methods have a broad range of applicability. Authier et al. [43] applied a genetic algorithm

to model kinetically the devolatilization of coal and described it as an efficient optimization procedure.

Cai et al. [44] applied a genetic algorithm to estimate the optimal kinetic parameters associated with

peanut shells decomposition and compared with the obtained from thermogravimetric curves.

The least square fitting is an optimization method that solves nonlinear problems by minimizing

the error (𝛿) between a fitting function (𝑈𝑝𝑟𝑒𝑑 (𝑖,𝑗)) and reference data (𝑈𝑒𝑥𝑝(𝑖,𝑗)

) according to:

𝛿 = √1

𝑁𝛽∑ ∑ (𝑈𝑒𝑥𝑝(𝑖,𝑗)

− 𝑈𝑝𝑟𝑒𝑑(𝑖,𝑗))

2𝑁

𝑗=1

𝛽

𝑖=1 (2.5)

18

where β is the number of heating rates considered for the optimization, N is the number of iterations.

This method is based on an initial guess, and the function coefficients (in this case kinetic parameters)

can be evaluated in a deterministic way or, recurring to optimization solvers, through a specific number

of function evaluations. Matlab has a number of algorithms available for the function evaluation. During

the course of the tool development, it was observed that the default set up produced fast, stable and

very satisfactory results.

The combination of these two optimization methods has been used by Authier et al. [43],

although for coal devolatilization studies. The author proved that this two-step optimization procedure

leads to more accurate and reliable results. This is due to the fact that the genetic algorithm does not

need an initial guess and rapidly converges to an optimal solution. The GA output is then used has an

initial guess for the non-linear least squares fitting procedure to provide an adequate start. The addition

of this last procedure gives more precise results since it does not consider random effects.

19

3. MATERIALS AND METHODS

3.1. BIOMASS CHARACTERIZATION

Table 3.1 lists the proximate and ultimate analysis for the three biomass fuels (pine bark, wheat

straw and rice husk) used in this work.

Table 3.1. Proximate and ultimate analysis of the biomass fuels studied.

Parameter Pine bark Wheat straw Rice husk

Proximate analysis (wt.%, ar)

Volatile matter (VM) 63.7 64.9 65.5

Fixed carbon (FC) 21.2 11.5 14.6

Ash 2.6 14.7 10.5

Moisture 12.5 8.9 9.4

Ultimate analysis (wt.%, daf)

C 52.6 51.6 52.1

H 7.4 6.8 7.2

N 1.0 0.6 0.6

S < 0.02 < 0.02 < 0.02

Oa 39.0 41.0 40.1

a Calculated by difference, ar – as received, daf – dry-ash-free

Comparing the chemical analysis of the three samples, the main characteristic that stands out

is the ash content. Pine bark has significantly lower content of ash than wheat straw and rice husk. On

the other hand pine bark has the higher quantity of moisture, while wheat straw and rice husk have lower

but similar content. Other than that, the remaining parameters are quite similar and the results discussed

here are in agreement with the survey data listed in Table 2.1. It is not possible to use this method due

to the samples containing a higher quantity of hydrogen than the triangulation method reference

biomass. Figure 3.1 shows that the samples fall outside the PoliMi triangle.

The correlation method [45] was used to estimate the contents of cellulose (CELL),

hemicellulose (HCE) and lignin (LIG). The mass fractions of cellulose and hemicellulose were calculated

as follows:

CELL = −1019.07 + 293.810(O C⁄ ) − 187.639(O C⁄ )2 + 65.1426(H C⁄ ) − 19.3025(H C⁄ )2

+ 21.7448(VM) − 0.132123(VM)2 (3.1)

HCE = 612.099 + 195.366(O C⁄ ) − 156.535(O C⁄ )2 + 511.357 (H

C) − 177.025(H C⁄ )2

− 24.3224(VM) + 0.1453063(VM)2

(3.2)

20

Figure 3.1. Triangulation method test for the biomass samples studied.

where O/C and H/C are the molar fractions and VM is the volatile matter in wt.%, dry-ash-free (daf). The

mass fraction of lignin was calculated by difference. Table 3.2 lists the composition range of the

correlation method. The precision of the correlation for cellulose is 90% and for hemicellulose is 81%

[45].

Table 3.2. Composition range of the correlation method [45].

O/C (molar ratio) H/C (molar ratio) VM (wt. %)

Range 0.56 - 0.83 1.26 - 1.69 73 - 90

Figure 3.2 shows the contents of cellulose, hemicellulose and lignin of the three biomass fuels

estimated through Eqs. (3.1) and (3.2). The figure reveals that the pine bark is richer in hemicellulose,

while both the wheat straw and rice husk are richer in cellulose.

Figure 3.2. Contents of cellulose, hemicellulose and lignin.

Table 3.3 shows the comparison between the estimated composition of the three main

components and values reported in the literature. The estimated composition of pine bark and rice husk

is consistent with the literature data, but the lignin fraction of the wheat straw and rice husk are slightly

higher than that reported by other authors.

21

Table 3.3. Comparison between the estimated composition of the three main

components and literature values [12,21,46].

Biomass Correlation Literature

CELL HCE LIG CELL HCE LIG

Pine bark (wt.%, db) 28.2 24.2 45.1 20 - 45 19 - 50 22 - 46

Wheat straw (wt.%, db) 34.1 15.1 41.4 29 - 43 18 - 39 19 - 30

Rice husk (wt.%, db) 33.0 18.5 35.7 18 - 44 18 - 35 18 - 30

3.2. THERMOGRAVIMETRIC STUDY

The thermogravimetric tests were performed in a SDT 2960 simultaneous DSC-TGA (TA

Instruments), under an inert atmosphere of Argon, with a constant flow of 100 mL/min. The samples

were grinded to less than 1 mm and placed on a measuring crucible with an initial weight of 5 (± 1) mg.

Due to the simplicity of the used method to detect and record mass loss variations, the precision of the

measurements is only influenced by the TGA balance sensitivity, which is 0.1 µg.

Figure 3.3 shows the temperature profiles for the three heating rates tested. The drying stage

consisted in heating each biomass from 300 K to 383 K over 10 min, followed by a plateau of 30 min.

When this stage was completed, the sample was heated to 1173 K followed by a plateau of 30 min,

using heating rates of 5, 10 and 15 K/min. Since the objective of this work does not consider the study

of the drying process, the results that will be presented in the following sections will be normalized to

dry basis. Also, it will only be considered the interval of the pyrolysis process until 1073 K. At this

temperature, the pyrolysis process has approach completion as typified by the DTG curves that will be

presented in the results section.

Figure 3.3. Temperature profiles for the three heating rates tested.

22

4. KINETIC ANALYSIS

In this chapter it is discussed in detail the different methods (Arrhenius plot and optimization

methods) considered to estimate the kinetic parameters for the pyrolysis process, using the kinetic

models SFOM and 3PM. At first kinetic analysis the FWO Arrhenius plot method was used to estimate

the kinetic parameters to compute the mass loss curves. But this method showed to be time consuming

and limited regarding the quantity of experimental information that can be considered. To fight this

fragility a combination of existing optimization tools in MATLAB (genetic algorithm and least squares

procedure) was implemented to guarantee that the kinetic parameters will always be determine faster,

systematically and above all, with precision.

4.1. ARRHENIUS PLOT METHOD

The Arrhenius plot method is typically used to estimate 𝐸𝑎 and 𝐴 kinetic parameters of non-

isothermal reactions where the heating rate is constant in time, as in thermogravimetric analysis. Within

this context, the most commonly used methods are the integral iso-conversion methods Flynne-Walle-

Ozawa (FWO) [16, 47, 48] and Kissinger-Akahira-Sunose (KAS) [15, 16, 47, 48]. Both methods were

tested in this work, but only FWO will be further described since it gave better results. The FWO method

relies on the following steps [16]:

1. Establishing the equation of the decomposition rate

𝑑𝛼

𝑑𝑡= 𝐴 exp [

−𝐸𝑎

𝑅𝑇] 𝑓(𝛼) (4.1)

where t is time (s), α the fraction of reacted sample or, in other words, the conversion degree and 𝑓(𝛼)

is the conversion function that represents the reaction model. The unknown parameters are Ea, A and

𝑓(𝛼).

2. Definition of the conversion degree (α)

𝛼 =𝑚𝑜 − 𝑚

𝑚𝑜 − 𝑚𝑓

(4.2)

where mo is the initial mass, mf the final mass and m the mass at a specific time and temperature.

3. Adaptation of Equation 4.1 for non-isothermal reactions

𝑑𝛼

𝑑𝑡= 𝛽

𝑑𝛼

𝑑𝑇= 𝐴. exp [

−𝐸𝑎

𝑅𝑇] 𝑓(𝛼) (4.3)

which can be integrated with respect to α and T resulting in

23

ln 𝛽 ≅ log [𝐴𝐸𝑎

𝑅𝑔(𝛼)] − 2.315 − 0.4567 [

𝐸𝑎

𝑅𝑇] (4.4)

where β is the heating rate (K/s) and g(α) is the integral function of conversion. For first order reactions,

typically used to describe the devolatilization process, the integral function reads as [32]

𝑔(𝛼) = −ln (1 − 𝛼) (4.5)

4. Assuming the conversion degree (∝) to be a constant value [32], the kinetic parameters can be

determined from the slope of a series of lines, each correspondent to a specific heating rate, when

plotting ln(β) vs 1/T. Figure 4.1 shows a representative example of the FWO method [16], where it

were tested five heating rates.

Figure 4.1. Representative example of the FWO method [49].

.

It is very important to notice that if the apparent estimated activation energy increases with the

conversion degree, we are dealing with a complex reaction mechanism, otherwise, if there is not a

significant change it is assumed that we are dealing with a single step reaction [16]. There are a few

examples of studies in the literature that apply the Arrhenius plot to describe the devolatilization using

[50,51], as reviewed in section 1.2. However, the Arrhenius plot method is time consuming and prone

to the introduction of errors during the process when dealing with multiple data sets.

24

4.2. OPTIMIZATION METHODS

Given the impossibility of measuring and quantifying the amount of extractives present in the

biomass samples, this work focuses only on 3PM and SFOM to test the degree of complexity of the

model required to describe the pyrolysis process. Species were not considered in the model because

they were not measured.

The total weight loss of each component i (in the case of the SFOM it is just one component) is

governed by a single reaction as function of time, temperature and mass loss history as follows [35]:

�̇�𝑝𝑖𝑟,𝑖 =dmp,i

dt= ki(T)(VMi − Vg,i) (4.6)

where dmp,i/dt is the mass loss in weight percent of the ith component, VMi is the maximum volatile

matter that can be loss and Vg,i is the total amount of volatile gases that have left the particle, both in

wt.%. For the 3PM, the amount of VM is corrected using the fraction of char produced by the i th

component, Ci, as follows [14]:

VMi = VM(1 − Ci) (4.7)

The rate constant ki (T) is expressed by the Arrhenius equation as follows

ki = AiTpγ

exp (−Ei

RTp

) (2.1)

where R is the ideal gas constant (J.K-1mol-1), Ai is the pre-exponential factor (s-1), Ei is the activation

energy (J.mol-1) and γ is the temperature power coefficient [5]. The mass balance is integrated over the

time considering the experimental duration using an explicit Euler method with a small enough time step

to ensure sufficient accuracy. This can be easily confirmed graphically by observing if the resultant mass

loss curves present smooth gradients (small enough time step) or sharp corners (too large time step).

The mass balance discretized equation reads as follows

𝑚𝑝,𝑖𝑡+∆𝑡= 𝑚𝑝,𝑖𝑡

+ �̇�𝑝𝑖𝑟,𝑖∆𝑡 (4.8)

The experimental heating rates are imposed to the mass balance. Figure 4.2 shows the process flow of

the pyrolysis algorithm.

25

Figure 4.2. Pyrolysis algorithm.

The kinetic optimization procedure was programmed using an object-oriented structure with two

steps: the genetic algorithm (GA) from MATLAB’s global optimization toolbox and the MATLAB’s least

squares fitness function (LSQ). The error function to minimize, in other words, the objective function

reads as [52]:

𝛿 = √1

𝛽𝑁∑ ∑ (𝑈𝑒𝑥𝑝(𝑖,𝑗)

− 𝑈𝑝𝑟𝑒𝑑(𝑖,𝑗))

2𝑁

𝑗=1

𝛽

𝑖=1 (2.5)

where β is the number of heating rates considered for the optimization, N is the number of time

integrations at each rate, and Uexp and Upred are the experimental and predicted yields at each time step

and rate, respectively. The objective function can have a number of local minima when multiple heating

rates are considered. The GA was used to provide an initial guess of the global minimum, which was

then fed to the LSQ to minimize further the error. Figure 4.3 shows the process flow of the kinetic

optimization procedure.

26

Figure 4.3. Kinetic optimization procedure.

For both kinetic models considered, initial bounds were imposed in the GA following typical

ranges from the literature (see below). The maximum number of generations was set to 100, with a limit

of stall generation number of 50. The population size for each generation was set to 30. Finally, it was

allowed 80% crossover between generations and 5% of mutation within an individual to ensure variety

and new chromosomes in the next generation. Regarding the least squares function, the same bounds

of GA were set for the 3PM, while for the SFOM no bounds were imposed. The maximum function

evaluations and the maximum iterations number were both set to 1×105. Both methods converge when

the average change in the fitness value is less than 1×10-12.

The kinetic parameters are E, A and γ in the case of the SFOM, and Ei, Ai, γi and Ci in the case

of the 3PM. Table 4.1 shows the typical range of activation energies reported in the literature for the

SFOM. In the case of woody biomass and wheat straw, the wide range of activation energies

encountered can be a consequence of the different heating conditions used in the experiments and/or

of the different biomass characteristics (particle size and composition) [13]. Table 4.2 shows the typical

kinetic parameters for the 3PM and the fraction of char produced by each component. These parameters

are typically used for woody biomass.

27

Table 4.1. Typical range of activation

energies for the SFOM [13, 22, 47, 53].

Biomass E (kJ/mol)

Woody biomass 89 - 175

Wheat straw 55 - 130

Rice husk 72 - 82

Table 4.2. Typical kinetic parameters for the 3PM and fraction of char produced by

each component [12, 14, 20].

ECELL EHCE ELIG ACELL AHCE ALIG CCELL CHCE CLIG

175 - 240 80 - 129 18 - 87 1013 - 1020 106 - 1011 1 - 105 0.14 - 0.38 0.15 - 0.43 0.1 - 0.47

To ensure consistency in the predicted results with the 3PM, it was estimated only one activation

energy and one fraction of char produced by each component (cellulose, hemicellulose and lignin) for

each biomass, regardless of the heating rate. These tests showed good agreement with the literature

data for the SFOM, but for the 3PM it was necessary a larger universe for the activation energy of the

cellulose, for the pre-exponential factor of the hemicellulose and for the char fraction produced by the

three main biomass components. Regarding the activation energy of the cellulose the lower bound was

set to 80 kJ/mol, as for the pre-exponential factor of hemicellulose the upper bound was set to 1011 and

for CCELL, CHCE and CLIG the new ranges were reset to 0 - 0.60, 0 – 0.60 and 0.1 - 0.60, respectively. It

is important to notice that since the Arrhenius plot method does not allow fitting the temperature

exponents, the limits were imposed empirically from −10 to +10. Table 4.3 shows the average

simulation times and specific standard deviations relative to the chosen model and Table 4.4 lists the

work computer specifications.

Table 4.3. Average simulation times and specific standard deviations

relative to the chosen model.

Model Average time (min) σDEV (min)

SFOM 9.6 1.8

3PM 29.3 7.4

Table 4.4. Work computer specifications.

CPU Intel® core™ i7-4770 CPU @ 3.40GHz

RAM 8.00 GB DDR3 1333 MHz

Hard Drive 1TB 7200 RPM 64MB Cache

28

4.3. COMPARISON BETWEEN METHODS

The most evident difference between the methods is that more constants of the Arrhenius

equation can be estimated using the optimization method, in particular the gamma coefficient. A

common characteristic shared is the fact of both can consider multiple reactions, but in the case of the

Arrhenius plot method the number of reactions is limited, up to five [13]. Beyond that, the application of

the two-step fitting procedure is more accurate and robust than the Arrhenius plot method. The fitting

procedure delivers results much faster, is less prone to errors introduced by the user and uses all the

experimental data available – there is no loss of information. In the Arrhenius plot method, due to the

relatively small finite number of points typically considered, information is lost. Nevertheless,

implementation errors may be introduced during the development stage of the fitting procedure. In order

to validate the implementation, the fitting procedure was tested using the results published by Burhenne

et al. [12] and Grønli et al. [14] and an excellent agreement was obtained.

29

5. RESULTS

5.1. EXPERIMENTAL RESULTS

Figure 5.1 shows the pyrolysis yields for pine bark, wheat straw and rice husk for the heating

rates of 5, 10 and 15 K/min. At the end of all stages of decomposition, each biomass reaches a similar

value typical of the mass solid residue, regardless of the heating rate, as seen in other studies [16].

Independently of the heating rate, pine bark finishes decomposition at higher temperatures after rice

husk and wheat straw that finish decomposition earlier, hence at lower temperatures. These results are

consistent with the estimated composition. Pine bark has the highest amount of lignin after rice husk

and wheat straw (cf. Figure 3.2), and it is the lignin content that controls the pyrolysis process - the

higher the amount of lignin, the slowest is the decomposition of the biomass.

Figure 5.2 shows the DTG curves as a function of the temperature for pine bark, wheat straw

and rice husk for the heating rates of 5, 10 and 15 K/min. The results show that pine bark behaves

differently than wheat straw and rice husk, which have similar weight loss curves. The first region of

decomposition, referred to as such due to the overlap of the first and second stages of decomposition

[12–14,18–20,54], corresponds to the hemicellulose and cellulose, respectively. The DTG curves also

show that the hemicellulose decomposition (first stage) usually appears as a more or less pronounced

“shoulder” instead of a distinct peak, as it happens for cellulose [12–14,18,19]. The second region (third

stage) corresponds to the decomposition of the lignin that occurs in a wide range of temperatures until

it reaches temperatures of 973 K [12–14,18,54]. In summary, Figure 5.2 reveals that the distinct peaks

associated with the different constituents are not so noticeable, indicating that the components thermal

decompose simultaneously, overlapping each other in DTG profiles, as seen in other studies [12,14–

16,19,20].

Figure 5.1. Pyrolysis yields for pine bark (PB), wheat straw (WS) and rice husk (RH) for 5, 10 and 15

K/min.

Table 5.1 shows the characteristics of the first region of decomposition for the three biomass

fuels. In the table Ti is the temperature where the decomposition of the first region begins, THCE is the

temperature in which hemicellulose maximum pyrolysis rate occurs, TCELL is the temperature in which

30

cellulose maximum pyrolysis rate occurs and Tf is the temperature where the decomposition of the first

region ends.

Figure 5.2. DTG curves as a function of temperature for pine bark (PB), wheat straw (WS) and rice

husk (RH) for 5, 10 and 15 K/min.

Table 5.1. Characteristics of the first region of decomposition

for the three biomass fuels.

Biomass β Ti (K) THCE (K) TCELL (K) Tf (K) Tf −Ti (K)

Pine bark

5 505 543 609 649 141

10 510 547 616 658 148

15 513 552 617 666 153

Wheat straw

5 520 555 587 620 100

10 524 565 599 630 106

15 530 573 605 638 108

Rice husk

5 525 568 592 627 102

10 528 569 603 634 106

15 536 573 614 648 112

The decomposition characteristics are consistent with other studies [13,14,18,19,55]. The

pyrolysis maxima for the hemicellulose occur in between 543 K and 573 K and for the cellulose occur in

between 587 K and 617 K. Also, the cellulose decomposition resulted in a much higher decomposition

maximum compared to the hemicellulose and lignin decomposition [18,19,54]. Figure 5.1 and Table 5.1

reveal that as the heating rate increases, the transition temperatures slightly increase and the DTG

curves tend to be wider. This is due to the fact each biomass constituents have individual decomposition

peaks in specific temperature ranges and the increase of the heating rate tends to delay thermal

decomposition processes towards higher temperatures [18,19]. In the case of pine bark Tf −Ti is larger

than the corresponding values for rice husk and wheat straw, underlining that the higher amount of lignin

leads to the slowest decomposition [12,13,18,54].

31

5.2. KINETIC ANALYSIS RESULTS

Figure 5.3 shows the Arrhenius plot method applied to experimental TG and DTG curves of pine

bark (PB), wheat straw (WS) and rice husk (RH), respectively, presented in the previous section (see

figures 5.1 and 5.2), and Table 5.2 shows the kinetic parameters, Ea and A, estimated with the Arrhenius

plot method.

Figure 5.3. Arrhenius plot method applied to experimental TG curves of pine bark (PB), wheat straw

(WS) and rice husk (RH), respectively.

Table 5.2. Kinetic parameters estimated with the Arrhenius plot method.

All the samples show similar activation energy, specifically 161, 152 and 166 kJ/mol for pine

bark, wheat straw and rice husk, respectively. This method was successfully applied from a conversion

factor of 0.2 to 0.6, correspondent to a range of temperatures within the interval between T i and Tf.

Conversion factors superior to 0.7 due to the impossibility of establishing a correlation, since the

correspondent temperature is approximately 800 K, near to the pyrolysis end. From α = 0.2 to α = 0.6

the values of the apparent Ea are similar, which is an indicator of the presence of a single step reaction

(reinforced by the correlation factor (R2) being close to the unit value), as Guerrero et al. concluded [16].

Table 5.3 shows the kinetic parameters obtained by fitting the SFOM to the experimental data

for the three biomass fuels. The obtained results are consistent with the literature data except for pine

bark [22,47,53,55]. This may be because most of the literature results are relative to various types of

pine wood, excluding generally the pine bark.

Pine Bark Wheat Straw Rice Husk

α Ea (kJ/mol) A (s-1) R2 Ea (kJ/mol) A (s-1) R2 Ea (kJ/mol) A (s-1) R2

0.2 167 1.14×1012 0.999 145 4.76×1012 0.999 165 2.16×1014 0.995

0.3 151 4.20×1012 0.996 142 2.03×1012 0.999 169 3.19×1014 0.997

0.4 161 1.87×1013 0.993 174 1.03×1015 0.994 166 1.18×1014 0.999

0.5 164 1.46×1013 0.992 151 7.67×1012 0.999 166 8.64×1013 0.998

0.6 169 4.31×1012 0.992 149 3.97×1012 0.998 166 6.70×1013 0.998

Average 161 9.64×1012 0.994 152 2.11×1014 0.998 166 1.61×1014 0.997

32

Table 5.3. Kinetic parameters obtained by fitting the SFOM to the experimental data for the three

biomass fuels.

Biomass E

(kJ/mol)

A (s-1) ×1011 γ δ (%)

5 10 15 5 10 15

Pine bark 55.3 27.5 47.3 71.5 -3.842 -0.982 2.450 7.1

Wheat straw 77.5 1.86 93.8 2.77 -2.532 3.801 8.992 5.2

Rice husk 84.6 97.3 23.7 38.9 -2.981 0.440 -1.692 4.8

Table 5.4 shows the activation energies and char fractions and Table 5.5 shows the pre-

exponential factors and model constants, all obtained by fitting the 3PM to the experimental data for the

three biomass fuels. Given the optimization procedure, through which the kinetic parameters were

optimized specifically for each biomass, some variations were obtained for the activation energies of the

cellulose, hemicellulose and lignin. Nevertheless, the values are very close when comparing all the

biomass fuels, as verified also by [56,57]. Also, activation energies of hemicellulose and cellulose are

higher than lignin, this is due to the reactivity of the components [12,18,20]. It is expected that the

cellulose and the hemicellulose decomposed almost entirely due to its weaker structure making lignin

the main contributor to the formation of char [12,58], as typified by CLIG in Table 5.4. Li et al. [19], studied

the variation of the parameters of A and Ci with the heating rate concluding that both parameters change

with a significant increase of heating rate. Due to the fact that heating rates did not varied significantly

in this study, the variation of Ci with the heating rate was not considered.

Table 5.4. Activation energies and char fractions obtained by fitting the 3PM to

the experimental data for the three biomass fuels.

Biomass ECELL

(kJ/mol)

EHCE

(kJ/mol)

ELIG

(kJ/mol)

CCELL

(wt.%)

CHCE

(wt.%)

CLIG

(wt.%)

δ

(%)

Pine bark 152.5 95.7 44.3 0.080 0.038 0.574 2.2

Wheat straw 143.3 83.6 37.0 ≈ 0 ≈ 0 0.441 1.9

Rice husk 163.8 107.3 37.2 ≈ 0 ≈ 0 0.567 2.2

* All the values ≈ 0 are smaller than 1×10-5.

33

Table 5.5. Pre-exponential factors and model constants obtained by fitting the 3PM to the

experimental data for the three biomass fuels.

Biomass β (K/min) ACELL (s-1) AHCE (s-1) ALIG (s-1) γCELL (-) γHCE (-) γLIG (-)

Pine

bark

5 9.14×1018 1.03×1017 337.8 -5.322 0.315 1.369

10 2.09×1018 7.57×1016 827.7 -5.773 -2.470 7.212

15 9.42×1018 9.98×1016 175.3 -7.945 -6.917 -2.815

Wheat

straw

5 4.68×1017 2.46×107 1000 -2.587 -0.784 -1.096

10 5.92×1018 6.24×107 101 -4.644 -0.703 -5.204

15 8.33×1017 4.35×107 692 -8.543 -1.197 6.324

Rice

husk

5 1.67×1018 7.47×107 3.8 -2.263 -0.089 -0.237

10 2.06×1016 2.15×107 360 0.350 4.926 -7.904

15 3.41×1018 3.41×107 412 -2.616 1.166 0.624

Figures 5.4 to 5.6 show the TG, DTG and predicted curves for pine bark, wheat straw and rice

husk, respectively. The SFOM captures the pyrolysis behavior in a satisfactory way but only the 3PM

reproduces it rather well, since it considers three stages of decomposition. In the case of wheat straw,

nearly in the end of the decomposition process (~920 K), DTG curves show a small pyrolysis peak that

indicates the occurrence of inorganic reactions [59].

When comparing the fitting errors obtained for the SFOM and for the 3PM (cf. Tables 5.4 and

5.5), it becomes obvious that the latter describes better the pyrolysis processes. Nevertheless, the fitting

error of the SFOM is very satisfactory, evidencing the appropriateness of the optimization method. Only

the 3PM is capable to predict correctly the maximum pyrolysis rate for all biomass fuels, but in the case

of pine bark there is a slight shift of +20 K.

34

Figure 5.4. TG (top), DTG (bottom) and predicted curves for pine bark (PB).

Figure 5.5. TG (top), DTG (bottom) and predicted curves for wheat straw (WS).

35

Figure 5.6. TG (top), DTG (bottom) and predicted curves for rice husk (RH).

Figure 5.7 shows DTG and predicted curves using the 3PM for pine bark, wheat straw and rice

husk, where the contributions of the cellulose, hemicellulose and lignin are plotted. Since the results are

similar for all heating rates, only the DTG curves for 5 K/min are included in the Figure 5.7. For the cases

of the wheat straw and rice husk the predicted pyrolysis maxima of the cellulose and hemicellulose

components show excellent agreement with the experimental curves. In the case of pine bark the

predicted pyrolysis maxima do not overlap. This may be due to an over estimation of the cellulose

component as discussed previously.

Figure 5.7. DTG and predicted curves using the 3PM for pine bark (PB), wheat straw (WS) and rice

husk (RH) at 5 K/min.

The results of the activation energies obtained through the Arrhenius plot method and the fitting

procedure cannot be directly compared because (1) in the latter, the temperature power coefficient, γ,

36

was considered and (2) the estimation procedure is different since the former estimates a single value

of Ea and A for all the heating rates tested and the latter can extend the estimation procedure to multiple

values of Ea, A and γ, according to the number of heating rates tested. Giving that the range of variation

of the heating rates in this study is not significant, it is reasonable to consider the activation energy

constant [14].

Nevertheless, TG curves were also predicted with the parameters estimated with the former and

compared to the experimental results. Table 5.6 shows the error and execution time for the Arrhenius

plot method and the fitting procedure for the results obtained with SFOM. The value of the error

associated with the overall process and the execution time of the Arrhenius plot method are significantly

higher than those of the fitting procedure.

Table 5.6. Comparison between the Arrhenius plot method and

the fitting procedure.

Method Arrhenius Fitting

Error (%) (PB; WS; RH) 22; 32; 21 2.2; 1.9; 2.2

Execution time Several hours ≈ 9.6 min

37

6. CONCLUSIONS

The proposed objectives were successfully achieved. The pyrolysis behavior of pine bark, wheat

straw and rice husk was studied by thermogravimetry using heating rates of 5, 10 and 15 K/min. A fitting

tool was developed to estimate kinetic parameter based in optimization methods. The kinetic parameters

were obtained by fitting the SFOM and 3PM to the experimental curves using a two-stage optimization

procedure. The main conclusions of this work are as follows:

The results reveal that for each biomass the variation of the heating rates had a small impact in

the pyrolysis process, particularly in the total mass loss; the biomass decomposition, however,

started earlier in time but at a slightly higher temperature for the highest heating rate.

The activation energies obtained in this work using the SFOM were 55.5, 79.6 and 87 kJ/mol

for pine bark, wheat straw and rice husk, respectively.

The activation energies for cellulose, hemicellulose and lignin obtained in this work using the

3PM were, respectively, 152.5, 95.7 and 44.3 kJ/mol for pine bark, 143.3, 83.6 and 37 kJ/mol

for wheat straw, and 163.8, 107.3 and 37.2 kJ/mol for rice husk.

Overall, the optimized parameters for the SFOM resulted in a very satisfactory fitting error of

7.1%, 5.2% and 4.8% for pine bark, wheat straw and rice husk, respectively. The optimized

parameters for the 3PM resulted in a good fitting error (≈ 2%) and were generally within typical

values.

The obtained results proved that the kinetic tool developed in this work was capable of

reproducing the pyrolysis behavior with good accuracy and showed that the degree of complexity of

3PM suffices. This tool showed the advantage of being fast, able to estimate multiple kinetic parameters

and to test any type of lignocellulosic biomass in multiple heating rates.

38

7. FUTURE PERSPECTIVES

A few aspects of the fitting procedure for 3PM can be improved. Namely, the activation energies

for each component should be fixed for all types of biomass. Furthermore, a sensitivity analysis varying

significantly the relative amount of each component would show how sensitive is the procedure to major

composition variations.

Finally, the tool developed is a base for future developments, specifically the inclusion of

transport effects (energy and momentum balance for other application such as drop tubes) and also to

include kinetic models for the thermal conversion of biomass under different atmospheres (gasification

and steam pyrolysis).

39

8. REFERENCES

[1] IEA Statistics -renewables information, “International Energy Agency.” 2014.

[2] The Royal Commission on Environmental Pollution, “Biomass as a renewable energy source.”

[3] H. Viana, W. B. Cohen, D. Lopes, and J. Aranha, “Assessment of forest biomass for use as

energy. GIS-based analysis of geographical availability and locations of wood-fired power plants

in Portugal,” Applied Energy, vol. 87, no. 8, pp. 2551–2560, 2010.

[4] P. Carneiro and P. Ferreira, “The economic, environmental and strategic value of biomass,”

Renewable Energy, vol. 44, pp. 17–22, 2012.

[5] S. Ferreira, N. A. Moreira, and E. Monteiro, “Bioenergy overview for Portugal,” Biomass and

Bioenergy, vol. 33, no. 11, pp. 1567–1576, 2009.

[6] European Climate Foundation, Biomass for heat and power – Opportunity and economics,

Technical report, June 2010.

[7] R. Saidur, E. Abdelaziz, A. Demirbas, M. S. Hossain, and S. Mekhilef, “A review on biomass as

a fuel for boilers,” Renewable and Sustainable Energy Reviews, vol. 15, no. 5, pp. 2262–2289,

2011.

[8] E. Butler, G. Devlin, D. Meier, and K. McDonnell, “A review of recent laboratory research and

commercial developments in fast pyrolysis and upgrading,” Renewable and Sustainable Energy

Reviews, vol. 15, no. 8, pp. 4171–4186, 2011.

[9] http://www.fortum.com.

[10] http://www.dynamotive.com.

[11] http://www.btgworld.com.

[12] L. Burhenne, J. Messmer, T. Aicher, and M. P. Laborie, “The effect of the biomass components

lignin, cellulose and hemicellulose on TGA and fixed bed pyrolysis,” Journal of Analytical and

Applied Pyrolysis, vol. 101, pp. 177–184, 2013.

[13] C. Di Blasi, “Modeling chemical and physical processes of wood and biomass pyrolysis,”

Progress in Energy and Combustion Science, vol. 34, no. 1, pp. 47–90, 2008.

[14] M. G. Grønli, G. Várhegyi, and C. Di Blasi, “Thermogravimetric Analysis and Devolatilization

Kinetics of Wood,” Industrial & Engineering Chemistry Research, vol. 41, no. 17, pp. 4201–4208,

2002.

[15] D. K. Seo, S. S. Park, J. Hwang, and T. U. Yu, “Study of the pyrolysis of biomass using thermo-

gravimetric analysis (TGA) and concentration measurements of the evolved species,” Journal of

Analytical and Applied Pyrolysis, vol. 89, no. 1, pp. 66–73, 2010.

[16] M. R. B. Guerrero, M. Marques da Silva Paula, M. M. Zaragoza, J. S. Gutiérrez, V. G. Velderrain,

a. L. Ortiz, and V. Collins-Martínez, “Thermogravimetric study on the pyrolysis kinetics of apple

pomace as waste biomass,” International Journal of Hydrogen Energy, pp. 1–9, 2014.

[17] T. R. Rao and A. Sharma, “Pyrolysis rates of biomass materials,” Energy, vol. 23, no. 11, pp.

973–978, 1998.

40

[18] T. Damartzis, D. Vamvuka, S. Sfakiotakis, and A. Zabaniotou, “Thermal degradation studies and

kinetic modeling of cardoon (Cynara cardunculus) pyrolysis using thermogravimetric analysis

(TGA),” Bioresource Technology, vol. 102, no. 10, pp. 6230–6238, 2011.

[19] Z. Li, W. Zhao, B. Meng, C. Liu, Q. Zhu, and G. Zhao, “Kinetic study of corn straw pyrolysis:

Comparison of two different three-pseudocomponent models,” Bioresource Technology, vol. 99,

no. 16, pp. 7616–7622, 2008.

[20] D. Vamvuka, N. Pasadakis, and E. Kastanaki, “Kinetic Modeling of Coal / Agricultural By-Product

Blends,” Energy & Fuels, no. 17, pp. 549–558, 2003.

[21] T. Mani, P. Murugan, J. Abedi, and N. Mahinpey, “Pyrolysis of wheat straw in a thermogravimetric

analyzer: Effect of particle size and heating rate on devolatilization and estimation of global

kinetics,” Chemical Engineering Research and Design, vol. 88, no. 8, pp. 952–958, 2010.

[22] A. Sharma and T. R. Rao, “Kinetics of pyrolysis of rice husk,” Bioresource Technology, vol. 67,

pp. 53–59, 1999.

[23] U. Fernandes and M. Costa, “Potential of biomass residues for energy production and utilization

in a region of Portugal,” Biomass and Bioenergy, vol. 34, no. 5, pp. 661–666, 2010.

[24] S. V. Vassilev, D. Baxter, L. K. Andersen, and C. G. Vassileva, “An overview of the chemical

composition of biomass,” Fuel, vol. 89, no. 5, pp. 913–933, 2010.

[25] S. V. Vassilev, D. Baxter, L. K. Andersen, C. G. Vassileva, and T. J. Morgan, “An overview of

the organic and inorganic phase composition of biomass,” Fuel, vol. 94, pp. 1–33, 2012.

[26] H. Yang, R. Yan, H. Chen, D. H. Lee, and C. Zheng, “Characteristics of hemicellulose, cellulose

and lignin pyrolysis,” Fuel, vol. 86, no. 12–13, pp. 1781–1788, 2007.

[27] Dias J., “Utilização da biomassa: avaliação dos resíduos e utilização de pellets em caldeiras

domésticas,” Instituto Superior Técnico, Universidade de Lisboa, 2002.

[28] A. Demirbas, “Combustion characteristics of different biomass fuels,” Progress in Energy and

Combustion Science, vol. 30, no. 2, pp. 219–230, 2004.

[29] J. Werther, M. Saenger, E. U. Hartge, T. Ogada, and Z. Siagi, “Combustion of agricultural

residues,” Progress in Energy and Combustion Science, vol. 26, no. 1, pp. 1–27, 2000.

[30] B. . Jenkins, L. . Baxter, T. . Miles, and T. . Miles, “Combustion properties of biomass,” Fuel

Processing Technology, vol. 54, no. 1–3, pp. 17–46, 1998.

[31] M. Rabaçal, “Influência das características do combustível no desempenho energético e

ambiental de caldeiras domésticas,” Tese Mestrado, Inst. Super. Técnico, Univ. Lisboa, 2010.

[32] D. López-González, M. Fernandez-Lopez, J. L. Valverde, and L. Sanchez-Silva, “Kinetic analysis

and thermal characterization of the microalgae combustion process by thermal analysis coupled

to mass spectrometry,” Applied Energy, vol. 114, pp. 227–237, 2014.

[33] TA Instruments, “SDT 2960 Simultaneous DSC-TGA including DTA capabilities.” pp. 1–8.

[34] M. Costa and P. Coelho, "Combustão", 2a ediçao. Amadora, 2012.

[35] S. Badzioch and G. W. Hawksley, “Kinetics of thermal decomposition of pulverized coal

particles,” Industrial & Engineering Chemistry Process Design and Development, vol. 9, no. 4,

pp. 521–530, 1970.

41

[36] J. Cai, W. Wu, and R. Liu, “An overview of distributed activation energy model and its application

in the pyrolysis of lignocellulosic biomass,” Renewable and Sustainable Energy Reviews, vol.

36, pp. 236–246, 2014.

[37] A. Rostami, M. R. Hajaligol, and S. E. Wrenn, “A biomass pyrolysis sub-model for CFD

applications,” Fuel, vol. 83, no. 11–12, pp. 1519–1525, 2004.

[38] E. Ranzi, A. Cuoci, T. Faravelli, A. Frassoldati, G. Migliavacca, S. Pierucci, and S. Sommariva,

“Chemical kinetics of biomass pyrolysis,” Energy and Fuels, vol. 22, no. 6, pp. 4292–4300, 2008.

[39] http://sci.waikato.ac.nz/farm/content/plantstructure.html.

[40] http://www.scielo.br/scielo.php?pid=S0104-14282013000300002&script=sci_arttext

[41] A. Geletu, “Solving Optimization Problems using the Matlab Optimization Toolbox - a Tutorial,”

Math. Program., no. 8, p. 48, 2007.

[42] http://www.mathworks.com.

[43] O. Authier, E. Thunin, P. Plion, and L. Porcheron, “Global Kinetic Modeling of Coal

Devolatilization in a Thermogravimetric Balance and Drop-Tube Furnace,” Energy & Fuels, vol.

29, no. 3, pp. 1461–1468, 2015.

[44] J. Cai and L. Ji, “Pattern search method for determination of DAEM kinetic parameters from

nonisothermal TGA data of biomass,” Journal of Mathematical Chemistry, vol. 42, no. 3, pp. 547–

553, 2007.

[45] C. Sheng and J. L. T. Azevedo, “Modeling biomass devolatilization using the chemical

percolation devolatilization model for the main components,” Proceedings of the Combustion

Institute, vol. 29, no. 1, pp. 407–414, 2002.

[46] J. S. Lim, Z. Abdul Manan, S. R. Wan Alwi, and H. Hashim, “A review on utilisation of biomass

from rice industry as a source of renewable energy,” Renewable and Sustainable Energy

Reviews, vol. 16, no. 5, pp. 3084–3094, 2012.

[47] J. M. Cai and L. S. Bi, “Kinetic analysis of wheat straw pyrolysis using isoconversional methods,”

J. Therm. Anal. Calorim., vol. 98, no. 1, pp. 325–330, 2009.

[48] K. Slopiecka, P. Bartocci, and F. Fantozzi, “Thermogravimetric analysis and kinetic study of

poplar wood pyrolysis,” Third International Conference on Applied Energy, pp. 1687–1698, 2011.

[49] D. Li, L. Chen, X. Zhang, N. Ye, and F. Xing, “Pyrolytic characteristics and kinetic studies of three

kinds of red algae,” Biomass and Bioenergy, vol. 35, no. 5, pp. 1765–1772, 2011.

[50] M. Lanzetta and C. Di Blasi, “Pyrolysis kinetics of wheat and corn straw,” Journal of Analytical

and Applied Pyrolysis, vol. 44, no. 2, pp. 181–192, 1998.

[51] J. J. M. Orfão, F. J. A. Antunes, and J. L. Figueiredo, “Pyrolysis kinetics of lignocellulosic

materials—three independent reactions model,” Fuel, vol. 78, no. 3, pp. 349–358, 1999.

[52] J. Ballester and S. Jiménez, “Kinetic parameters for the oxidation of pulverised coal as measured

from drop tube tests,” Combustion and Flame, vol. 142, no. 3, pp. 210–222, 2005.

[53] M. Stenseng, A. Jensen, and K. Dam-Johansen, “Investigation of biomass pyrolysis by

thermogravimetric analysis and differential scanning calorimetry,” Journal of Analytical and

Applied Pyrolysis, vol. 58, pp. 765–780, 2001.

42

[54] D. H. L. H. Yang, R. Yan, H. Chen, C. Zheng and D. T. Liang, “In-Depth Investigation of Biomass

Pyrolysis Based on Three Major Components: Hemicellulose, Cellulose and Lignin,” Industrial &

Engineering Chemistry Research, vol. 20, pp. 388–393, 2005.

[55] C. Di Blasi and C. Branca, “Kinetics of Primary Product Formation from Wood Pyrolysis,”

Industrial & Engineering Chemistry Research, vol. 40, no. 23, pp. 5547–5556, 2001.

[56] J. J. M. Órfão and J. L. Figueiredo, “A simplified method for determination of lignocellulosic

materials pyrolysis kinetics from isothermal thermogravimetric experiments,” Thermochimica

Acta, vol. 380, no. 1, pp. 67–78, 2001.

[57] Y. Fan, Y. Cai, X. Li, H. Yin, N. Yu, R. Zhang, and W. Zhao, “Rape straw as a source of bio-oil

via vacuum pyrolysis: Optimization of bio-oil yield using orthogonal design method and

characterization of bio-oil,” Journal of Analytical and Applied Pyrolysis, vol. 106, pp. 63–70, 2014.

[58] G. Toscano, D. Duca, G. Rossini, C. Mengarelli, and a. Pizzi, “Identification of different woody

biomass for energy purpose by means of Soft Independent Modeling of Class Analogy applied

to thermogravimetric analysis,” Energy, vol. 83, pp. 351–357, 2015.

[59] D. López-González, M. Fernandez-Lopez, J. L. Valverde, and L. Sanchez-Silva, “Gasification of

lignocellulosic biomass char obtained from pyrolysis: Kinetic and evolved gas analyses,” Energy,

vol. 71, pp. 456–467, 2014.