Tesina di "Energia da biomasse e rifiuti" - Applicazione di un modello di energia di attivazione distribuita (DAEM) alla pirolisi di diverse biomasse

UNIVERSITÀ DEGLI STUDI DI PERUGIA

DIPARTIMENTO DI INGEGNERIA

Corso di Laurea Magistrale in Ingegneria Meccanica

A.A. 2015/2016

Application of a Distributed Activation Energy Model (DAEM) to the

pyrolysis of different biomasses

Course of Energy from biomass and waste

Students Supervisors

Lorenzo Catanzani Prof. Eng. Francesco Fantozzi

Samuele Trinari Doc. Pietro Bartocci

Course of Energy from biomass and waste

Application of a Distributed Activation Energy Model (DAEM) to the pyrolysis of different biomasses

Students: Lorenzo Catanzani (code: 279775) & Samuele Trinari (code: 280861)

Summary

1. State of the art of biomass pyrolysis modeling pag. 3

References 3

2. Distributed Activation Energy Model description pag. 9

3. TG and DTG diagrams pag. 11

4. Application of the DAEM model to glycerol pellet pag. 12

5. Application of the DAEM model to olive stone pag. 16

6. Conclusions pag. 18

7. Appendix

Appendix A: Dataset used pag. 19

Appendix B: MATLAB code 2

Course of Energy from biomass and waste

Application of a Distributed Activation Energy Model (DAEM) to the pyrolysis of different biomasses

Students: Lorenzo Catanzani (code: 279775) & Samuele Trinari (code: 280861)

1 ­ State of the art of biomass pyrolysis modeling

Kinetics of biomass to predict mass loss evolution is usually determined by TGA and conducted

with small samples at low heating rates to ensure the absence of heat and mass transport

α α

limitations, i.e., to be under a kinetically controlled regime. The reaction rate (d /dt, being

conversion) over temperature in a TGA experiment with pinewood at a constant heating rate of 10

K/min is shown in Fig. 1.1 [1].

Fig.1.1. Normalised mass loss and reaction rate over temperature for pyrolysis of pine wood at 10K/min. Experimental data in

points, model with three pseudo­components in solid line and each pseudo­component in dashed lines (adapted form Anca­Couce et

al [1].

The main peak corresponds to cellulose; the shoulder at lower temperatures, to hemicellulose;

and lignin decomposition covers a wider temperature range, including the tail at high

temperatures. There are two main mathematical approaches used to analyse the data in order to

obtain the kinetics data: model­based (model­modelfitting) and isoconversional (model­free)

methods [2],[3],[4],[5].

Isoconversional (model­free) methods can be used to compute kinetic parameters during

conversion without model­based assumptions, such as an a priori first order reaction. In these

3

Course of Energy from biomass and waste

Application of a Distributed Activation Energy Model (DAEM) to the pyrolysis of different biomasses

Students: Lorenzo Catanzani (code: 279775) & Samuele Trinari (code: 280861)

methods the activation energies are calculated at fixed conversions, taking advantage of the fact

that the reaction rate depends exclusively on the reaction temperature. There are integral

isoconversional methods, such as the Kissinger–Akahira–Sunose (KAS) [6],[7], Flynn–Wall–Ozawa

(FWO) [8],[9] or Vyazovkin [10] methods; and differential methods, such as the Friedman method.

In model­based methods, a reaction model must be postulated first. The most appropriate

reaction model can solely be selected on the basis of the quality of the regression fit. Nonlinear

least squares fitting is the method most commonly employed in the biomass community to fit

experimental data and evaluate the Arrhenius parameters. First and nth order reaction models are

usually selected. It is recommended to employ the reaction rate (as in Fig. 1.1) instead of mass loss

over time/temperature for the fitting because the details of devolatilisation are better shown [11].

The description of biomass pyrolysis with just one component is not precise enough. Biomass

pyrolysis is assumed to be approximately the sum of the inputs of the respective main

components: cellulose, hemicellulose and lignin [12]. Pyrolysis can be described with a parallel

reaction scheme in which three pseudocomponents usually represent the main biomass

components, although more components can be employed. However, the proportions of each

pseudo­component do not correspond exactly to the composition of the real components because

of the influence of mineral matter, different char yields and interactions among the components

[13]. Moreover, the final char yield has to be a priori defined in this scheme; this scheme predicts

the mass loss evolution over time but not the variations in product composition at different

pyrolysis conditions. Some studies obtain the pyrolysis kinetics using model­fitting methods solely

with experiments conducted at one heating rate [14],[12]. Branca et al. [15], Anca­Couce et al. [3]

or Sánchez­Jiménez et al. [16] have criticised this, though. Force fitting models to non­isothermal

data obtained from a single heating rate can generate very inconsistent Arrhenius parameters that

display a strong dependence on the kinetic model selected [2]. Compensation effects can be

avoided by employing several heating rates, i.e., different combinations of pre­exponential factors

and activation energies can describe the same weight loss curve reasonably well. Only one set of

data can predict the behaviour of the material at several heating rates [13]. 4

Course of Energy from biomass and waste

Application of a Distributed Activation Energy Model (DAEM) to the pyrolysis of different biomasses

Students: Lorenzo Catanzani (code: 279775) & Samuele Trinari (code: 280861)

As reviews have indicated, activation energies vary widely in the literature for each

pseudo­component in the parallel reaction scheme [13],[2],[18]. Fig. 1.2 presents an overview

(data from Refs. [14], [12], [19], [15], and [20],[21],[22],[23],[24],[25],[26],[27]).

Fig.1.2. Activation energies reported in literature for the biomass pseudo­components. Data obtained with experiments performed

at several heating rates (except 11): 1[20], 2[21], 3[19], 4[22], 5[15], 6[23], 7[24], 8[25], 9[26], 10[27], 11[12], 12[14]. When several

values are reported in one study, only the mean value is shown (adapted from Anca­Couce et al. [3]

The included values in the figure are only obtained from works in which several heating rates were

investigated, except for the reference work of Gronli et al. [12]. The activation energies of the

pseudo­components in the parallel reaction scheme usually resemble the activation energies of

the original components. Pure cellulose pyrolysis (for low heating rates) can be described with a

first order reaction model with a high activation energy (191–253 kJ/mol according to Antal et al.

[28]). The main component for biomass in the parallel reaction scheme representing the cellulose

peak usually has activation energies in the range of 190–250 kJ/mol, close to the values for pure

cellulose. Furthermore, the literature usually reports a lower activation energy value for the

hemicellulose pseudo­component than for cellulose, but usually the activation energy is still high

(150–200 kJ/mol). The reported range of activation energies for lignin is very broad [13],[18], from

20 to 200 kJ/mol. The widely varying kinetic data reported in recent years in the literature have

sparked concern among researches about the reliability of the reported pyrolysis experiments and

5

Course of Energy from biomass and waste

Application of a Distributed Activation Energy Model (DAEM) to the pyrolysis of different biomasses

Students: Lorenzo Catanzani (code: 279775) & Samuele Trinari (code: 280861)

the analysis of the data [13],[2],[18],[29]. The following recommendations were suggested to

determine kinetics in a consistent way by Anca­Couce et al. [3]:

• To first reproduce the reference experiments with pure cellulose (Avicel PH 105) from Gronli et

al. [30].

• To perform and analyse experiments with different heating rates.

• To employ integral isoconversional methods to verify the reliability of the experiments and to

avoid selecting inappropriate reaction models in a fitting routine.

The first recommendation is done to validate the thermogravimetric analysis and to ensure the

absence of heat and mass transport limitations. The International Confederation for Thermal

Analysis and Calorimetry (ICTAC) has presented further advices for the collection of data in a

kinetically controlled regime [31]. Experiments conducted at different heating rates are

recommended to avoid compensation effects, as previously explained. Khawamand Flanagan [4]

suggested the complementary use of isoconversional and model­based methods to determine

solid state reaction kinetic parameters from experimental data. Activation energies can be first

predicted by using isoconversional methods. The most accurate reaction model can be then

chosen, by using model­based methods, in order to arrive at an activation energy that is close to

the one obtained from the isoconversional analysis. Thus, the selection of the most appropriate

reaction model is in this way potentially more consistent than when the quality of the regression

fit alone is used as a basis [4]. Moreover, quality of the linear fitting in Arrhenius plots of integral

isoconversional methods for the different conversions can verify the reliability of the experiments.

Compensation effects or assuming first order kinetic models can lead to erroneous kinetic

parameters, particularly leading to an underestimation of the activation energy of lignin pyrolysis.

By following the recommendations above, consistent kinetics should be obtained. Isoconversional

methods are not commonly employed by members of the biomass pyrolysis community, but their

popularity has recently increased. Differential isoconversional methods are very sensitive to noise,

therefore integral methods are usually preferred. Generally, high activation energies are obtained,

with variations during conversion [32],[33],[34],[35]. Activation energies at different conversion

levels can be approximately associated to biomass components: hemicellulose at low, cellulose at

medium and lignin at high conversion levels. Different integral isoconversional methods, such as

KAS, FWO or Vyazovkin, deliver very similar results for the same data set. 6

Course of Energy from biomass and waste

Application of a Distributed Activation Energy Model (DAEM) to the pyrolysis of different biomasses

Students: Lorenzo Catanzani (code: 279775) & Samuele Trinari (code: 280861)

The complexity of biomass pyrolysis has led to the employment of more complex models than the

previously presented ones, with one reaction and a single activation energy for each component.

Distributed activation energy models (DAEM) assume that decomposition takes place over a large

number of independent, parallel reactions with different activation energies, which reflects

variations in the bond strengths of species. The difference in activation energies can be

represented by a continuous distribution function. This model was recently reviewed by Cai et al.

[40]. Kinetic parameters can be estimated by distribu