Accepted Manuscript Characteristics and kinetic study on pyrolysis of five lignocellulosic biomass via thermogravimetric analysis Zhihua Chen, Mian Hu, Xiaolei Zhu, Dabin Guo, Shiming Liu, Zhiquan Hu, Bo Xiao, Jingbo Wang PII: DOI: Reference:

S0960-8524(15)00735-X http://dx.doi.org/10.1016/j.biortech.2015.05.062 BITE 15033

To appear in:

Bioresource Technology

Received Date: Revised Date: Accepted Date:

29 March 2015 18 May 2015 19 May 2015

Please cite this article as: Chen, Z., Hu, M., Zhu, X., Guo, D., Liu, S., Hu, Z., Xiao, B., Wang, J., Characteristics and kinetic study on pyrolysis of five lignocellulosic biomass via thermogravimetric analysis, Bioresource Technology (2015), doi: http://dx.doi.org/10.1016/j.biortech.2015.05.062

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Characteristics and kinetic study on pyrolysis of five lignocellulosic biomass via thermogravimetric analysis Zhihua Chena, Mian Hu a, Xiaolei Zhua, Dabin Guoa, Shiming Liu a, Zhiquan Hua,1, Bo Xiaoa, Jingbo Wangb a

School of Environmental Science & Engineering, Huazhong University of Science

and Technology, Wuhan 430074, China b

Hainan Research Academy of Environmental Sciences, 98 Baiju Avenue, Haikou,

Hainan Province, 571126, China

Abstract Pyrolysis characteristics and kinetic of five lignocellulosic biomass pine wood sawdust, fern (Dicranopteris linearis) stem, wheat stalk, sugarcane bagasse and jute (corchorus capsularis) stick were investigated using thermogravimetric analysis. The pyrolysis of five lignocellulosic biomass could be divided into three stages, which correspond to the pyrolysis of hemicellulose, cellulose and lignin, respectively. Single Gaussian activation energy distributions of each stage are 148.50-201.13 kJ/mol with standard deviations of 2.60-13.37 kJ/mol. The kinetic parameters of different stages were used as initial guess values for three-parallel-DAEM model calculation with good fitting quality and fast convergence rate. The mean activation energy ranges of hemicellulose, cellulose and lignin were 148.12-164.56 kJ/mol, 171.04-179.54 kJ/mol and 175.71-201.60 kJ/mol, with standard deviations of 3.91-9.89, 0.29-1.34 and 23.22-27.24 kJ/mol, respectively. The mass fractions of hemicellulose, cellulose and lignin in lignocellulosic biomass were respectively estimated as 0.12-0.22, 0.54-0.65 and 0.17-0.29.

1

Corresponding author. Tel./fax: +86 27 87557464. E-mail address: [email protected] (Z. Hu).

1

Keywords: Lignocellulosic biomass; kinetics; Non-isothermal thermogravimetric analysis; Three-parallel-DAEM model

1 Introduction Lignocellulosic biomass represents a renewable and largely untapped source of raw feedstock for thermochemical conversion (pyrolysis, gasification and combustion) into liquid, gas fuels and other energy-related end products. Large amount of lignocellulosic biomass such as agricultural residues (approximately 0.73 billion tons ) and forest residues (approximately 37 million m3) are produced per year in China (Ma et al., 2012) . Wheat straw, pinewood sawdust, sugarcane bagasse, jute stick, and fern (Dicranopteris dichotoma) residues are most common agricultural and forest wastes in Southern China which can be used as potential sources of renewable energy. Comparing with gasification and combustion, pyrolysis method has some advantages: (a) lower temperature requirement,(b) oxygen-absent condition, and (c) production of higher-quality oil (Bridgwater et al., 1999; Damartzis et al., 2011). Kinetic modeling of pyrolysis can help to describe practical conversion processes and optimize the design of efficient reactors (Di Blasi, 2008; Ceylan and Topcu, 2014). Non-isothermal thermogravimetric analysis (TGA) is a powerful technique for biomass pyrolysis kinetics study. Many studies (Oyedun et al., 2014; Damartzis et al., 2011; Gai et al., 2013) had been focused on this topic by using various model-free and model-fit methods. However, such model-free and model-fit methods, which are only suitable for one-step reaction process, are not applicable to the simulation of biomass pyrolysis process due to its complexity (Vyazovkin et al., 2011). Multi-step reaction model (MSRM) is more suitable to simulate biomass pyrolysis kinetics (Vyazovkin et al., 2011). 2

Among many MSRMs, the parallel reaction model (PRM) is most commonly used in different biomass pyrolysis kinetics studies (Li et al., 2008; Anca-Couce et al., 2014). The simplest PRM is three- parallel-reaction (TPR) model which assumes that pseudo components (hemicellulose, cellulose and lignin) react independently and simultaneously with nth order (n=1 or n≠1) reaction. Although very good fit quality on experimental data can be obtained using TPR model, the activation energy of each pseudo component was quite different from that of the pure hemicellulose (xylan), cellulose and lignin (Anca-Couce et al., 2014). The main reason may be that the mineral catalysis and diffusion effect are not taken into account in this model (Varhegyi et al., 2009; Anca-Couce et al., 2014). Another PRM is the distributed activation energy model (DAEM) which assumes the activation energy is not a fixed value but follows a continuous distribution when the number of parallel reactions trends to infinity. And then the mineral catalysis and diffusion effect can be ignored reasonably in DAEM. Since the pyrolysis of three pseudo components are independent from each other, pyrolysis behavior of each component can be further described by a DAEM. Currently, the three-parallel-DAEM model has been considered to be one of the most suitable kinetic models for the simulation of lignocellulosic biomass pyrolysis (Varhegyi et al., 2011; Cai et al., 2013). The unknown parameters of the three-parallel-DAEM model were usually determined by evaluating the experimental TG or DTG data using the nonlinear least squares (NLS) method (Cai et al., 2013; Varhegyi et al., 2011). When the NLS method is used, the determination of the initial values of kinetic parameters and fitting object (TG or DTG data) is very important ( Opfermann, 2000; Vyazovkin et al., 2011). According to ICTAC (International Confederation for Thermal Analysis and Calorimetry), the DTG data tends to magnify noise and thus converting into a 3

systematic error in the values of kinetic parameters compared with the TG data (Vyazovkin et al., 2011). So, the kinetic parameters estimated by fitting raw TG data are more accurate than those by fitting DTG data. The thermal degradation profiles of lignocellulosic biomass are interpreted as the addition of the independent degradations of their main components (hemicellulose, cellulose and lignin) (Sanchez-Silva et al., 2012). The reaction system for the overall conversion is assumed to be a set of the single reactions corresponding to the successive solid conversions. The kinetic parameters according to the degree of solid conversions can be determined by the DAEM method (Shen et al., 2011). Since the kinetic parameters of each stage are similar to those of pseudo components, they can be chosen as initial guess values when performing three-parallel-DAEM model simulation through NLS. To the author’s knowledge, so far no information is available on using such method to determine the initial guess values for kinetic parameters estimation by three-parallel-DAEM model. This study aims to investigate the characteristics and kinetics of five lignocellulosic biomass (pine wood sawdust, fern stem, wheat straw, sugarcane bagasse and jute stick) via thermogravimetric analysis. The three-parallel-DAEM model based on the nonlinear least squares method was used to simulate biomass pyrolysis.

2. Methods 2.1 Materials In this study, five biomass feedstocks: pinewood sawdust (PW), fern (Dicranopteris linearis) stem (FS), wheat stalk (WS), sugarcane bagasse (SB) and jute (corchorus capsularis) stick (JS) were used. All five biomass feedstocks were 4

obtained in Guiping city, Guangxi province, China. Each material was firstly dried under the sun for 30-40 days and then crushed, milled and sieved to achieve a particle size below 0.074 mm. The undersize products were firstly used as the samples for proximate analysis. After dried in 105℃ for 48 hours, the undersize products were then used as ultimate and TGA analysis samples. 2.2 Biomass characterization and thermogravimetric analysis The proximate analysis for the samples was performed according to related ASTM standards (Ahmaruzzaman, 2008). The ultimate analysis of five biomass samples was performed using a CHNS/O analyzer (Vario Micro cube, Elementar, Germany). The weight percent of carbon, hydrogen, nitrogen and sulfur in samples can be detected simultaneously, and the weight percent of oxygen was determined by difference. Thermogravimetric analysis (TGA) tests were performed using a Pyris1 TGA instrument (Perkin Elmer Co., Ltd, USA). Approximately 5.0 mg of each sample was paved uniformly in a platinum crucible and heated linearly from ambient temperature to 800℃ under nitrogen (purity of 99.99%, flow rate of 120 ml/min) atmosphere at a heating rate of 10K/min. All runs at given heating rate were repeated three times to eliminate the vibration error. 2.3. Kinetic modeling 2.3.1 Theoretical background of DAEM Reaction rate equation of solid-state pyrolysis in non-isothermal TGA can be interpreted as the homogeneous kinetics combined with Arrhenius equation, thus the final equation obtained is expressed as follows: r=

dα A  E  = exp  −  f (α ) dT β  RT 

(1)

5

where r is the reaction rate, 1/K; α is the conversion, dimensionless; T is the absolute temperature, K; A is the apparent pre-exponential factor, 1/s; E is the apparent activation energy, J/mol; R is the ideal gas constant, 8.3145 J/(mol·K); β is the linear heating rate, K/min; f(α) is conversion dependence function (or algebraic reaction model) and its independent variable, α, can be defined as

α=

m0 − mt m0 − m f

(2)

In Eq.(2), the m0 (%), mt (%) and mf (%) is the initial mass, instant mass and final mass of the sample, respectively. The integral form of Eq.(1) is given as g (α ) = ∫

α

0

1 A T A  E  d α = ∫ exp  − dT = ψ ( E , T )  T f (α ) β 0 β  RT 

(3)

where g(α) is the integral of f(α); ψ(E,T) is the temperature integral which has no analytical solution but always can be replaced with approximate expression. Combined with first-order reaction, f(α)=1-α, Eq.(3) becomes

 A  1 − α = exp  − ψ ( E, T )   β 

(4)

The distributed activation energy model (DAEM) assumes infinite first-order reactions happen and all such reactions share a constant pre-exponential factor (A) and the activation energy (E) are thought to obey a continuous distribution during pyrolysis process (Cai et al., 2014). According to Eq.(4), the conversion function of DAEM is given as: ∞  A  1 − α = ∫ exp  − ψ ( E , T )  f ( E )dE 0  β 

(5)

where f(E) is the probability density function (PDF) of the continuous distribution of activation energies. Several PDFs were reported in the literature including Gaussian, Weibull, and Gamma distribution (Cai et al., 2014). Among these PDFs, the Gaussian 6

PDF (Eq.(6)) is the most widely used one. f (E ) =

 ( E − E0 )2  1 exp  −  2σ 2  σ 2π 

(6)

where E0 and σ is the mean value and standard deviation of Gaussian PDF, respectively. 2.3.2 The multi-stage model The multi-stage model is firstly applied in this study. The corresponding kinetic parameters of each divided stage will be used as the initial value for nonlinear least squares calculation for three-parallel-DAEM model. 2.3.2.1 The division of pyrolysis stages The pyrolysis behavior of lignocellulosic biomass in TGA can be divided into three stages by using second derivative method (Gronli et al., 2002). This method gives the characteristic temperature of separate stages through the second derivative of TG curve or differential DTG (DDTG) curves. The first stage of DTG profile mainly corresponds to hemicelluloses decomposition and its end point is the local minimum of DDTG curve. And the third stage is mainly ascribed to lignin degradation and its initial point on DDTG curve corresponding to where the

−d 2 m / dT 2 values no longer change or change very little. The second stage, which shares the peak of DTG curve and lies in the middle of the first and third stage, mainly corresponds to cellulose pyrolysis. 2.3.2.2 DAEM kinetics of different stages Combining Eq.(5) with Eq.(6), the conversion function according to Gaussian DAEM for each stage is given by :

α j = 1−

1

σ j 2π



+∞

0

 ( E − E0, j ) 2   A  exp  − j ψ ( E , T )  exp  −  dE .  2σ 2j  β   

7

(7)

where j ( j=1, 2, 3) denotes number j stage. The unknown parameters Aj, E0,j and σj can be estimated by using nonlinear least squares method to minimize the following objective function: Nj

O.F1 = ∑ (α j , Exp − α j ,Calc )2

(8)

i =1

where αj,Exp, αj,Calc and Nj represents the experimental conversion, calculated conversion and the number of experimental points of jth stage, respectively. 2.3.3 Three-parallel-DAEM model In this model, the lignocellulosic biomass is regarded as the sum of three pseudo components: hemicellulose, cellulose, and lignin. For each pseudo component, infinite first-order reactions happen during its pyrolysis process based on DAEM. The mass lose rate expression for every such first-order reaction is as follows: −

dM p,i dT

=

 E  exp  − p ,i  M p,i β  RT 

Ap ,i

(9)

where Mp,i, Ap,i and Ep,i is the instantaneous mass, pre-exponential factor and activation energy of number p (p∈(0,+∞)) reaction for ith pseudo component, respectively. The 1 st, 2nd and 3rd pseudo component is assigned to hemicellulose, cellulose and lignin, respectively. According to DAEM, the total mass of ith pseudo component at given temperature (T) is given as: ∞

mi = ∫ M p,i dE

(10)

0

where mi is the mass of ith pseudo component. Mp,i can be obtained from the integral of Eq.(9) and given as:

 A  M p ,i = M p,i ,0 exp  − p ,i ψ ( E , T )   β 

(11)

where Mp,i,0 is the initial value of Mp,i. A constant pre-exponential factor (Ap,i ) is 8

further assumed and the Ep,i obeys a continuous distribution when the reaction number p tends to infinity. Then the mass change function of ith pseudo component can be obtained from the integral of Eq.(11) and given as: ∞  A  mi = m0,i ∫ exp  − i ψ ( E , T )  fi ( E )dE 0  β 

(12)

where Ai is the constant pre-exponential factor of ith pseudo component. The m0,i and distribution function of activation energies, fi(E), is respectively expressed as (Scott et al., 2006): ∞

m0,i = ∫ M p,i,0 dE

(13)

0

and fi ( E ) =

M i ,0





0

(14)

M i ,0 dE

By assuming the activation energies obey the common Gaussain distribution, the f(E) of each pseudo component is expressed as :

 ( E − E0,i )2  1 fi ( E ) = exp  −  2σ i2  σ i 2π 

(15)

In Eq.(15), σi and E0,i respectively refers to the normal distributed standard deviation and mean activation energy of jth pseudo component. The mass of char residue, mchar, is assumed to be constant when the pyrolysis completes. Thus the mass change function is given as: 3

m = ∑ mi + mchar

(16)

i =1

with  mi = mi ,0 and mchar = 0   mi = 0 and m = mchar

when T = T0 when T = T f 9

(17)

where T0 and Tf is the intial temperature and final temperature of pyrolysis process, respectively. Then the initial mass of ith can be given as: m0,i = ci (1 − mc ,tol ) .

(18)

where ci means the fraction of ith pseudo component in the biomass and ∑ci=1. The combination of Eq.(12), Eq.(16) and Eq.(18) gives the mass change function of lignocellulosic biomass as follows: 3 ∞  A  m = mc ,tol + ∑ ci (1 − mc,tol )∫ exp  − i ψ ( E, T )  f i ( E )dE 0 i =1  β 

(19)

The unknown parameters of three-parallel-DAEM model (Ai, σi, E0,i and ci ) can be estimated by using nonlinear least squares to minimize the objective function: N

O.F2 = ∑ (mExp − mCalc )2

(20)

i =1

where mExp is the experimental normalized mass; mCalc is the calculated normalized mass according to Eq.(19); N is the number of experimental points. 2.4 Calculation details 2.4.1 Treatment of temperature integral The 4th degree rational approximate expression (Eq.(21)) (Senum and Yang, 1977) which is one of the most accurate expressions with low error is used for the calculation of temperature integral ψ(E,T) in this study. T

E exp(− x)  −E  π ( x)  dT = R x  RT 

ψ ( E , T ) ≈ ∫ exp  0

(21)

where x=E/(RT) and

x3 + 18x 2 + 88 x + 96 π ( x) = 4 x + 20 x3 + 120 x 2 + 240 x + 120

(22)

2.4.2 Treatment of DAEM’s improper integral To deal with improper integral (dE integral from 0 to infinity) in Eq.(7) and 10

Eq.(19), the Lobatto Quadrature method (the built-in solver quadl of MATLAB® ( version R2012a)) will be employed. The limit of dE integral is from 0 to E0+30σ. 2.4.3 Nonlinear least squares In the present study, the nonlinear least squares method was performed by using lsqnonlin (a built-in optimization tool of MATLAB®) which is based on the Levenberg-Marquardt algorithm for the minimization of O.F1 (Eq.(8)) and O.F2 (Eq.(20)). The fit quality of nonlinear least squares method for three-parallel-DAEM model is calculated by the following expression (Gomez et al., 2004):

Fit (%) = 100

S / Np max(mExp )

= 100 S / N p

(23)

where S represents the objective function value of O.F2; Np is the number of unknown parameters.

3 Results and discussion

3.1 Characterization of fuels The proximate and ultimate analysis results, as well as the high calorific value (HHV) of the five lignocellulosic biomass are listed in Table 1. High amount of volatiles ranging from 71.80 % to 83.27% of these lignocellulosic biomass can be observed. It indicates that the five kinds of biomass could be considered as desirable feedstocks for pyrolysis, combustion or gasification processes. The moisture content (10.34%) of fern stem (FS) is the highest one, which may be due to the fact that the fern can only grow in areas where there are suitably moist conditions (Liang et al., 2008). Furthermore, the ash contents for PS (0.51%), FS (0.65%) and JS (0.81%) were quite lower than those of WS (7.99%) and SB (6.73%). The high ash content can be a disadvantage for pyrolysis process because it will result in fouling production on 11

the reactor wall and thus lead to disposal problems and reduced energy conversion efficiency (McKendry, 2002). The HHV of WS is the lowest one owing to the highest ash content and lowest volatiles content among the five kinds of biomass (Damartzis et al., 2011). The ultimate analysis results also showed that the five studied biomass had low nitrogen and sulfur content (ranging from 0.08 to 0.93%). Biomass feedstocks with low nitrogen and sulfur contents can be good candidates for thermochemical conversion. 3.2 Thermal decomposition characteristics The TG/DTG curves of the five studied lignocellulosic biomass pyrolysis at heating rate of 10K/min were exhibited in Fig. 1(a) and (b). One peak, one shoulder and one long tailing can be observed from each DTG curve. The shoulder caused by peaks overlaps represents the fastest conversion of hemicellulose. The peak corresponds to the decomposition of cellulose and the tailing mainly corresponds to the pyrolysis of lignin. In addition, the pyrolysis processes can be divided into three zones from DTG curves. The initial temperature (Ti) and final temperature (Tf) of different zones are listed in Table 2. The first zone (Zone I) from ambient temperature to about 141.47-172.40℃corresponds to the mass loss due to evaporation of little adsorbed water and light volatiles. The second zone (Zone II), ranging from 141.47-172.40℃ to about 554.85-588.00℃ is the main pyrolysis stage which is caused by the devolatilization. In the third zone (Zone III, mainly 554.85℃ to 800.00℃), very little mass loss is observed which is probably attributed to the degradation of carbonaceous in the residues. Table 2 also shows that Zone II is the main process of pyrolysis with high mass loss proportion (MLP, from 86.95 to 95.22%). However, different biomass has different Ti, Tp and Tf, which may be attributed to the strength of cross between 12

hemicellulose, cellulose and lignin in different biomass (Perez et al., 2002). The MLPs of FS, SB and JS in Zone II exceeded 90%, which were larger than those of PS (88.52%) and WS (86.95%). The reason lies in the fact that the volatile matter contents in FS, SB and JS are larger than those in PW and WS, as shown in Table 1. The degradation of hemicellulose, cellulose and lignin took place in Zone II (Gronli et al., 2002). In this zone, a shoulder, a peak and a long tail can be observed for all five lignocellulosic biomass. In general, the main peak of DTG profile corresponds to cellulose degradation and the shoulder at lower temperature corresponds to the decomposition of hemicelluloses and lignin, which covers a wider temperature range (Anca-Couce et al., 2014). By using second derivative method, Zone II can be further divided into three stages. Fig.1(c) gives the TG, DTG and differential DTG (DDTG) curves of the Zone II for pine wood. As can be seen, the peak of DTG curve represents the temperature at which the maximum weight loss rate is reached; in DDTG curve, this temperature is corresponding to the point of DDTG when (-d 2m/dT2=0). The shoulder at lower temperature is mainly ascribed to the degradation of hemicellulose and its temperature is corresponding to the local minimum value of DDTG (Gronli et al., 2002), about 304.12℃. The temperature range for stage I can be regarded as from Ti of Zone II to the temperature of local minimum of DDTG curve. The next section (temperature range is from 304.12 to 395.67℃) of DDTG is observed as sinusoid liked curve. This region can be regarded as Stage II which mainly corresponds to cellulose degradation. The final temperature of Stage II is the corresponding temperature that -d2m/dT2 values no longer change or change very little. And the rest region can be seen as the Stage III, which is mainly attributed to lignin degradation. The initial temperature (Ti) and final temperature (Tf) of different stages for five studied biomass are summarized in Table 2. There exist 13

some differences in the behavior of these five materials. The Ti and Tf of different stages for the five biomass are different from each other. The differences for all lignocellulosic samples can be attributed to content variations of hemicellulose, lignin and cellulose (Sanchez-Silva et al., 2012). 3.3 Single DAEM kinetics of different stages Stage I, II and III mainly corresponds to the pyrolysis of hemicellulose, cellulose and lignin, respectively (Gronli et al., 2002). To investigate the kinetic parameters (E0,j, Aj and σj) of single DAEM, the nonlinear least squares method (Eq.(8)) had been used and the initial guess value for these parameters were referred to Zhang et al., 2014. The expected ranges of E0,j and Aj were selected from 0 to infinity, σj was selected from 0 to 50 kJ/mol according to Cai et al., 2013.The calculated kinetic parameters are given in Table 3. The fitting R2 of different stages for five lignocellulosic biomass are shown in Fig.2 (c). It can be observed that high R2 are obtained for all stages especially for Stage II. The DAEM simulation of Stage I (FS), Stage II (PW) and Stage III (JS) are shown in Fig.2 (a). Fig.2 (a) shows that the simulation of PW in Stage II gives a nice fit to the experimental data (R2=0.9995). The fitting qualities of DDS (Stage I) and CCS (Stage III) are good, too ( R2=0.9783 and 0.9826, respectively). Mean value of E0,j is 148.50, 169.15 and 201.13kJ/mol, respectively. It is obvious that the standard mean activation energies follows the order of E0,III > E0,II > E0,I, which indicates that the activation energy of for lignin decomposition is the highest one and subsequently are cellulose and hemicellulose. The order of E0,j is in good agreement with the simulation results reported by Cai et al., 2013. The standard deviation σj follows the order of σIII > σI> σII. It indicates that lignin has the widest while cellulose has the most narrow distribution range as shown in Fig.2 (b). The 14

deviations of observed E0,j and σj for different stages are very low indicating that these parameters are stable in different stages. At the same time, the constant pre-exponential factor ( Aj ) of different biomass ranges from 2.83E+11 to 4.67E+13 s-1, which is in the reasonable parameter range (10E+11 to 10E+16 s-1) according to transition-state theory (de Jong et al., 2007). The comparisons between the estimated parameters and the literature results are given in Fig.2 (d). The results showed that the E0,j of stage II were lower than those of pure hemicellulose (xylan) (179.84 kJ/mol) and cellulose (240.23 kJ/mol) but E0,j values of stage III were higher than those of pure lignin reported by Zhang et al., 2014. It may be due to the temperature difference of different stages. The final temperature of stage I is only the peak temperature of hemicelluloses but not the termination temperature of its reaction. It is well known that thermal degradation of lignin takes place over a wide temperature range and that of Stage III only occupys the high temperature part of the whole temperature range of lignin degradation. In other words, the lower temperature range for lignin decomposition was ignored thus leads to a higher E0,j value than that of the pure lignin. The difference between estimated σj (as shown in Fig.2 (d)) and the results of Zhang et al. can also be explained for this reason. However, the E0 values range of Stage I (E0,I ), Stage II (E0,II) and Stage III (E0,III) respectively are similar to the E0 value of pseudo (not pure ) hemicellulose (160-195 kJ/mol), cellulose (175.6 kJ/mol) and lignin (237.1-266.6 kJ/mol) components of lignocellulosic biomass reported in previous studies (Cai et al., 2013; de Jong et al., 2007). It means that the Gaussian DAEM kinetics parameters of different stages can be used as good initial guess value for the three-parallel-DAEM model calculation. 3.4 Kinetics of three-parallel-DAEM model 15

According to Eq.(15) and Eq.(19),there are 12 unknown parameters (E0,i, Ai, σi and ci, where i=1,2,3) should be estimated for different biomass, Np=12 in Eq.(23). Since ∑ci=1, the ci value of lignin was replaced with

c3 = 1 − c1 − c2

(24)

Then the number of unknown parameters is Np=11. The single DAEM kinetic parameters of different stages were used as the initial guess value of three-parallel-DAEM model calculation. So the E0,j, Aj and σj value of stage I, stage II and Stage III were used as the corresponding E0,i, Ai and σi for hemicellulose, cellulose and lignin, respectively. The c1=0.20 and c2= 0.50 have been used as the initial value of ci (i=1, 2), which is similar to the data reported by Cai et al., 2013. The expected ranges of E0,i and Ai were selected from 0 to infinity and the range for σi is 0-50kJ/mol, which is in accordance with previous study (Cai et al., 2013). The estimated parameters are listed in Table 4 and the fitting results are plotted in Fig.3. As shown in Fig.3 (a-e), the three-parallel-DAEM model represents a good fit to the experimental results (Fit(%)=0.37-0.69, Table 4). As shown in Table 4, the convergence rates of nonlinear least squares method for different lignocellulosic biomass are very fast with 56-100 iterations. The corresponding iterative processes are exhibited in Fig.3 (f). In addition, the obtained final objective function value is very low (1.512E-4 to 5.317E-4) which leads to good fit quality. It means that the single DAEM kinetic parameters of different stages are very good for the three-parallel-DAEM model calculation by using nonlinear least squares method. By using numerical differential method the reaction rate (DTG curves in -dm/dT form) of pseudo components and their sum rates were obtained and are shown in Fig.4. As also can be seen, very good deconvolution of DTG curves has been obtained. This results are in conformity with the assumes of the three-parallel-DAEM model 16

that main peak of DTG curve corresponds to cellulose decomposition, the shoulder at lower temperatures corresponds to hemicellulose pyrolysis and the lignin decomposition covers a wider temperature range. As depicted in Table 4, the mean activation energy (E0,i) range for hemicellulose, cellulose and lignin is 148.12-164.56 kJ/mol, 171.04-179.54 kJ/mol and 175.71-201.60 kJ/mol, respectively. The mean activation energies for hemicellulose (E0,1), cellulose (E0,2) and lignin (E0,3) follow the order of E0,3 > E0,2 > E0,1. Our result is similar to other studies (e.g. Varhegyi et al. (2009), Cai et al. (2013) and (Cai et al. (2013)) and agree with the single DAEM results. The E0,i order of three pseudo components indicates that hemicellulose is easier to decompose than cellulose, and cellulose is easier to decompose than lignin because the activation energy can be thought as the height of the potential barrier for the reaction. Low activation barrier allows a reaction to happen quickly and easily but high activation barrier makes a reaction go more slowly. As shown in Fig.4, the first peak of each pseudo component is associated with the pyrolysis of hemicelluloses which appears at lower temperature range. It further proves that hemicellulose is easier to degrade than cellulose and lignin. The second peak attributed to the decomposition of cellulose and appears after the peak of hemicellulose. The temperature range of lignin is widest indicating that the lignin is harder to degrade than hemicellulose and cellulose. However, the mean activation energy of hemicellulose and cellulose is lower than that of pure hemicellulose (179.84 kJ/mol, (Zhang et al., 2014)) or xylan (178.331 kJ/mol, (Cai et al., 2013)) and pure cellulose (210.037 kJ/mol, (Cai et al., 2013)), respectively. And the mean activation energy of lignin is higher than that of pure lignin (165.61 kJ/mol, (Zhang et al., 2014)). In addition, the kinetic parameters of PW, SB and WS are a little lower than the values reported by Cai et al., 2013. This is 17

attributed to the fact that the matrices of amorphous hemicellulose, cellulose and lignin may vary in different lignocellulosic biomass (Perez et al., 2002). In this study, the pyrolysis kinetics of two novel lignocellulosic biomass, fern (Dicranopteris linearis) stem (FS) and jute (Corchorus capsularis) stick (JS) were first determined. A fundamental understanding of pyrolysis behaviors and kinetics of FS and JS is essential to their efficient thermochemical conversion. The estimated standard deviations (σi) of the activation energy distribution is 3.91-9.89, 0.29-1.34 and 23.22-27.24 kJ/mol for the hemicellulose, cellulose and lignin, respectively. Fig.5 shows the Gaussian activation energy distribution of the three pseudo components of five studied lignocellulosic biomass. It is obvious that the activation energy distribution for lignin has the widest distribution for each feedstock. On the contrary, the cellulose has the narrowest distribution. The activation energy distribution width of the three pseudo components is similar to the reported results of Varhegyi et al. (2011) and Cai et al. (2013). The wide distribution of the lignin indicates that the pyrolysis of lignin occurs cover an extensive temperature range which also can be seen in the deconvolution of DTG curves (Fig.4). 3.5 Mass fractions of pseudo components The present three-parallel-DAEM model is based on the fitting objective of TG curve. Through this model the initial fraction of pseudo components can be estimated based on the assumption that each pseudo component can be decomposed completely at final temperature. As shown in Table 4, the estimated mass fractions for hemicelluloses, cellulose and lignin respectively is 0.12-0.22, 0.54-0.65 and 0.17-0.29. The estimated mass fractions of pine wood, wheat straw and sugarcane bagasse in this study are very similar to those obtained in the recent studies (Shi and Wang, 2014; Su et al., 2015). Mass fractions of pseudo components of JS is also in accordance with 18

the results of previous study (Yao et al., 2008). And the mass fractions of pseudo components of FS is similar to those of the fern plant reported by Carrier et al., 2011. The above results demonstrated it is a novel and useful method to estimate the mass fractions of pseudo components in different biomass feedstocks using three-parallel-DAEM model. 4 Conclusions

The pyrolysis of five lignocellulosic biomass can be divided into three stages which mainly corresponds to hemicelluloses, cellulose and lignin decomposition, respectively. The kinetic parameters of different stages can provide referential values for three-parallel-DAEM model simulation with fast convergence rate and good fitting quality. Mean activation energies for hemicelluloses, cellulose and lignin are 148.12-164.56 kJ/mol, 171.04-179.54 kJ/mol and 175.71-201.60 kJ/mol, with standard deviations of 3.91-9.89, 0.29-1.34 and 23.22-27.24 kJ/mol, respectively. Mass fractions of hemicelluloses, cellulose and lignin is estimated to be 0.12-0.22, 0.54-0.65 and 0.17-0.29, respectively.

Acknowledgements

The authors would like to acknowledge the financial supports of Jiangsu province science and technology support program, China (BE2013127), Fundamental Research Funds for the Central Universities (2013TS070), the National High Technology Research and Development Program (863 Program) of China (No. 2012AA101809), Natural Science Foundation of Hainan Province (Project No. 414194) and SRF for ROCS, SEM, China. The first author also sincerely acknowledge Prof. Shuiqing Li ( Particle And Combustion Engineering (PACE) Research Group, Tsinghua University). In addition, the author would like to thank the 19

Analytical and Test Center of Huazhong University of Science and Technology for carrying out the analysis of biomass samples. References

1. Ahmaruzzaman, M., 2008. Proximate analyses and predicting HHV of chars obtained from cocracking of petroleum vacuum residue with coal, plastics and biomass . Bioresource Technology, 99, 5043–5050. 2. Anca-Couce, A., Berger, A., Zobel, N., 2014. How to determine consistent biomass pyrolysis kinetics in a parallel reaction scheme , pp. 230–240. 3. Bridgwater, A.V., Meier, D., Radlein, D., 1999. An overview of fast pyrolysis of biomass. Organic Geochemistry, 30, 1479-1493. 4.

Cai,

J.,

Wu,

W.,

Liu,

R.,

2013.

Sensitivity

analysis

of

three-parallel-DAEM-reaction model for describing rice straw pyrolysis , pp. 423–426. 5. Cai, J., Wu, W., Liu, R., 2014. An overview of distributed activation energy model and its application in the pyrolysis of lignocellulosic biomass . Renewable and Sustainable Energy Reviews, 236–246. 6. Cai, J., Wu, W., Liu, R., Huber, G.W., 2013. A distributed activation energy model for the pyrolysis of lignocellulosic biomass. Green Chemistry, 1331-1340. 7. Cao, H., Xin, Y., Wang, D., Yuan, Q., 2014. Pyrolysis characteristics of cattle manures using a discrete distributed activation energy model. Bioresource Technology, 172, 219-225. 8. Carrier, M., Loppinet-Serani, A., Denux, D., Lasnier, J., Ham-Pichavant, F., Cansell, F., Aymonier, C., 2011. Thermogravimetric analysis as a new method to determine the lignocellulosic composition of biomass. Biomass & Bioenergy, 35, 298-307. 20

9. Ceylan, S., Topcu, Y., 2014. Pyrolysis kinetics of hazelnut husk using thermogravimetric analysis. Bioresource Technology, 156, 182-188. 10. Channiwala, S.A., Parikh, P.P., 2002. A unified correlation for estimating HHV of solid, liquid and gaseous fuels. Fuel, 81, 1051-1063. 11. Damartzis, T., Vamvuka, D., Sfakiotakis, S., Zabaniotou, A., 2011. Thermal degradation studies and kinetic modeling of cardoon (Cynara cardunculus) pyrolysis using thermogravimetric analysis (TGA) . Bioresource Technology, 102, 6230–6238. 12. de Jong, W., Di Nola, G., Venneker, B.C.H., Spliethoff, H., Wojtowicz, M.A., 2007. TG-FTIR pyrolysis of coal and secondary biomass fuels: Determination of pyrolysis kinetic parameters for main species and NOx precursors. Fuel, 86, 2367-2376. 13. Di Blasi, C., 2008. Modeling chemical and physical processes of wood and biomass pyrolysis . Progress In Energy And Combustion Science, 34, 47–90. 14. Gai, C., Dong, Y., Zhang, T., 2013. The kinetic analysis of the pyrolysis of agricultural residue under non-isothermal conditions . Bioresource Technology, 298–305. 15. Gomez, C.J., Manya, J.J., Velo, E., Puigjaner, L., 2004. Further applications of a revisited summative model for kinetics of biomass pyrolysis. Industrial & Engineering Chemistry Research, 43, 901-906. 16. Gronli, M.G., Varhegyi, G., Di Blasi, C., 2002. Thermogravimetric analysis and devolatilization kinetics of wood. Industrial & Engineering Chemistry Research, 41, 4201-4208. 17. Li, Z., Zhao, W., Meng, B., Liu, C., Zhu, Q., Zhao, G., 2008. Kinetic study of corn straw pyrolysis: Comparison of two different three-pseudocomponent models . Bioresource Technology, 99, 7616–7622. 21

18. LIANG, T., DING, S., SONG, W., CHONG, Z., ZHANG, C., LI, H., 2008. A review of fractionations of rare earth elements in plants . Journal Of Rare Earths, 26, 7–15. 19. Ma, L., Wang, T., Liu, Q., Zhang, X., Ma, W., Zhang, Q., 2012. A review of thermal-chemical conversion of lignocellulosic biomass in China, pp. 859-873. 20. McKendry, P., 2002. Energy production from biomass (part 1): overview of biomass. Bioresource Technology, 83, 37-46. 21. Opfermann, J., 2000. Kinetic analysis using multivariate non-linear regression - I. Basic concepts. JOURNAL OF THERMAL ANALYSIS AND CALORIMETRY, 60, 641-658. 22. Oyedun, A.O., Tee, C.Z., Hanson, S., Hui, C.W., 2014. Thermogravimetric analysis of the pyrolysis characteristics and kinetics of plastics and biomass blends . Fuel Processing Technology, 471–481. 23. Perez, J., Munoz-Dorado, J., de la Rubia, T., Martinez, J., 2002. Biodegradation and biological treatments of cellulose, hemicellulose and lignin: An overview. International Microbiology, 5, 53-63. 24. Sanchez-Silva, L., López-González, D., Villaseñor, J., Sánchez, P., Valverde, J.L., 2012. Thermogravimetric–mass spectrometric analysis of lignocellulosic and marine biomass pyrolysis . Bioresource Technology, 163–172. 25. Scott, S.A., Dennis, J.S., Davidson, J.F., Hayhurst, A.N., 2006. An algorithm for determining the kinetics of devolatilisation of complex solid fuels from thermogravimetric experiments. Chemical Engineering Science, 61, 2339-2348. 26. Senum, G.I., Yang, R.T., 1977. Rational approximations of the integral of the Arrhenius function. Journal of thermal analysis, 11, 445-447. 27. Shen, D.K., Gu, S., Jin, B., Fang, M.X., 2011. Thermal degradation mechanisms 22

of wood under inert and oxidative environments using DAEM methods. Bioresource Technology, 102, 2047-2052. 28. Shi, X., Wang, J., 2014. A comparative investigation into the formation behaviors of char, liquids and gases during pyrolysis of pinewood and lignocellulosic components. Bioresource Technology, 170, 262-269. 29. Su, H., Liu, G., He, M., Tan, F., 2015. A biorefining process: Sequential, combinational lignocellulose pretreatment procedure for improving biobutanol production from sugarcane bagasse. Bioresource Technology, 187, 149-160. 30. Varhegyi, G., Bobaly, B., Jakab, E., Chen, H., 2011. Thermogravimetric Study of Biomass Pyrolysis Kinetics. A Distributed Activation Energy Model with Prediction Tests. Energy & Fuels, 25, 24-32. 31. Varhegyi, G., Chen, H., Godoy, S., 2009. Thermal Decomposition of Wheat, Oat, Barley, and Brassica carinata Straws. A Kinetic Study. Energy & Fuels, 23, 646-652. 32. Vyazovkin, S., Burnham, A.K., Criado, J.M., Pérez-Maqueda, L.A., Popescu, C., Sbirrazzuoli, N., 2011. ICTAC Kinetics Committee recommendations for performing kinetic computations on thermal analysis data , pp. 1–19. 33. Yao, F., Wu, Q., Lei, Y., Guo, W., Xu, Y., 2008. Thermal decomposition kinetics of natural fibers: Activation energy with dynamic thermogravimetric analysis. Polymer Degradation And Stability, 93, 90-98. 34. Zhang, J., Chen, T., Wu, J., Wu, J., 2014. A novel Gaussian-DAEM-reaction model for the pyrolysis of cellulose, hemicellulose and lignin, pp. 17513-17520.

23

Table Captions

Table 1 Ultimate analyses and proximate analyses of the five biomass. M, moisture content; A, ash content; V, volatile matter content; F, fixed carbon content; HHV, high calorific value. a

b

calculation by difference.

HHV=0.3491C+1.1783H+0.1005S-0.1034O-0.0151N-0.0211A

(Channiwala and Parikh, 2002). Table 2 Characteristic temperatures of the different zones and Stages for the five biomass during pyrolysis process. “~” denotes ambient temperature; MLP is the mass loss proportion of Zone II to total mass loss in TGA sets. Ti and Tf is the initial and final temperature of each zone and stage, respectively; Tp is peak temperature of DTG curves. Table 3 Single DAEM kinetic parameters values of different stages. where Mean and Dev is the mean value and deviation of parameters for different stages and biomass. Table 4 Simulation results of three-DAEM-reaction model for the five lignocellulosic biomass. where H is hemicellulose, C is cellulose and L is lignin. Figure Captions

Fig.1 (a) TG curves of five kinds of biomass; (b) DTG curves five kinds of biomass; (c) Characteristic of temperature of Zone II for pine wood by using second derivative method. mu masked beside the right vertical coordinate of (a) and (b) means the initial normalized mass of TG sets; Tc masked beside the right vertical coordinate of (b) means Celsius temperature. Fig.2 Single DAEM kinetic results of Stage I, II and III: 24

(a) Comparison between simulation and experiment results of FS (Stage I), PW (Stage II) and JS (Stage I); (b) Gaussian f(E) of Stage I (FS), Stage II (PW) and Stage III (WS); (c) Fitting R2 of single DAEM of different lignocellulosic biomass in different stage; (d) Comparison between DAEM kinetic parameters and reference (Ref., (Zhang et al., 2014)). Fig.3 TG simulation results of three-parallel-DAEM model for five lignocellulosic biomass: (a) Pine wood (PW), (b) Fern dicranopteris dichotoma stem (FS), (c) wheat straw (WS), (d) sugarcane bagasse (SB), (e) Jute (corchorus capsularis ) stick (JS) and (e) Iterative processes of the nonlinear least square method for five lignocellulosic biomass. Fig.4 Comparison between simulated DTG and experimental DTG curve of three-parallel-DAEM-reaction model for five lignocellulosic biomass: (a) pine wood (PW), (b) Fern (dicranopteris dichotoma) stem (FS), (c) wheat straw (WS), (d) sugarcane bagasse (SB) and (e) Jute ( corchorus capsularis ) stalk (JS). Fig.5 Gaussian f(E) of pseudo components: (a) Gaussian f(E) of hemicellulose, (b) Gaussian f(E) of cellulose, (c) Gaussian f(E) of lignin and (d) Gaussian f(E) of three components for FS and JS.

25

100

80

TG: mu (%)

(a)

PW FS WS SB JS

60

Zone III

Zone I Zone II

40

20 100

200

300

400

500

600

700

800

Temperature (oC) 1.2

o

DTG: −dmu /dTc (%/ C)

(b)

PW FS WS SB JS

1.0 0.8 0.6 Zone I

Zone III

Zone II

0.4 0.2 0.0 100

1.0

200

300

400

500

600

Temperature (oC)

×10-3 10

700

800

-3

×10 0.2

Tp

(c)

0.9 8

0.1

DDTG

0.5 0.4 0.3

2

2

− d m/dTc =0

Shoulder

2

0.6

0.0 6

-0.1

Shoulder

4

Stage I

Stage III

-0.2

2

-0.3

TG

DTG

0.2

Stage II

0 100

200

300

DDTG: − d m/dTc

TG: m

0.7

DTG: −dm/dT

2

0.8

-0.4 400

500

600

o

Temperature ( C)

Fig.1 (a) TG curves of five kinds of biomass; (b) DTG curves five kinds of biomass; (c) Characteristic of temperature of Zone II for pine wood by using second derivative method. mu and Tc masked beside the vertical coordinate means the initial normalized mass of TG sets and Celsius temperature, respectively.

26

0.16

1.0

(a)

(b) 2

R =0.9995

2

R =0.9783

Simulation: FS PW JS

Stage II Stage III

0.04

Stage I

0.2

Experiment: FS PW JS

0.0 400

0.08

500

600

700

800

0.00 100

900

150

200

250

Hemicellulose

Cellulose

Lignin

250

1.00

(c)

40

(d) Ref.

200

E0,j (kJ/mol)

0.98

R2

300

E(kJ/mol)

Temperature (K)

0.96

30

150

20 Black color: E0 value Blue color : σ value

σ j (kJ/mol)

α

0.6

0.4

Stage I for FS Stage II for PW Stage III for JS

0.12 2

R =0.9826

f(E) (mol/kJ)

0.8

Ref.

0.94

10

100 PW

FS

WS

SB

JS

0.92 50

Stage I

Stage II

Stage III

Stages

Stage I

Stage II

Stage III

0

Stages

Fig.2 Single DAEM kinetic results of Stage I, II and III: (a) Comparison between simulation and experiment results of FS (Stage I), PW (Stage II) and JS (Stage I); (b) Gaussian f(E) of Stage I (JS), Stage II (PW) and Stage III (WS); (c) Fitting R2 of single DAEM of different lignocellulosic biomass in different stage; (d) Comparison between DAEM kinetic parameters and reference (Ref., (Li et al., 2008)).

27

1.0

(a)

TG: m

0.6

Experiment: Experimental data

PW 0.4

FS

0.6

0.2

Simulation: Hemicellulose Cellulose Lignin Fit data

(b)

0.8

TG: m

0.8

1.0

Simulation: Hemicellulose Cellulose Lignin Fit data

Experiment: Experimental data

0.4

0.2

0.0

0.0

500

600

700

800

400

500

600

1.0

0.6

WS

TG: m

TG: m

0.8

800

900

1.0

Simulation: Hemicellulose Cellulose Lignin Fit data Experiment: Experimental data

(c)

700

Temperature (K)

Temperature (K)

0.4

0.8

(d)

0.6

SB

Simulation: Hemicellulose Cellulose Lignin Fit data Experiment: Experimental data

0.4

0.2

0.2

0.0

0.0 500

600

700

800

400

900

500

600

700

800

900

Temperature (K)

Temperature (K)

0.8

(e)

0.6

JS

(f)

8

Simulation: Hemicellulose Cellulose Lignin Fit data

Objective function

TG: m

1.0

Experiment: Experimental data

0.4

0.2

PW FS WS SB JS

6

4

2

0

0.0 400

500

600

700

800

900

Temperature (K)

0

10

20

30

40

50

60

Iteration

70

80

90 100 110

Fig.3 TG simulation results of three-parallel-DAEM model for five lignocellulosic biomass: (a) Pine wood (PW), (b) fern (dicranopteris dichotoma ) stem (FS), (c) wheat straw (WS), (d) sugarcane bagasse (SB), (e) jute (corchorus capsularis ) stick (JS) and (e) Iterative processes of the nonlinear least square method.

28

×10-3

×10-3

(a)

Simulation: Hemicellulose Cellulose Lignin Fit data:

DTG: −dm/dT (/K)

8

6

PW

Experiment: Experimental data:

4

(b)

6

DTG: −dm/dT (/K)

10

Simulation: Hemicellulose Cellulose Lignin Fit data

FS 4

Experiment: Experimental data

2

2 0

0 500

600

700

400

800

500

×10-3

12

(c)

6

WS 4

Experiment: Experimental data

2

900

(d) Simulation: Hemicellulose Cellulose Lignin Fit data

8

SB

6

Experiment: Experimental data

4

0 500

600

700

800

900

Temperature (K)

400

500

600

700

800

900

Temperature (K)

×10-3

(e)

12

Simulation: Hemicellulose Cellulose Lignin Fit data

10

DTG: −dm/dT (/K)

800

2

0

14

700

×10-3

10

Simulation: Hemicellulose Cellulose Lignin Fit data

DTG: −dm/dT (/K)

DTG: −dm/dT (/K)

8

600

Temperature (K)

Temperature (K)

JS

8

Experiment: Experimental data

6 4 2 0 400

500

600

700

800

900

Temperature (K)

Fig.4 Comparison between simulated DTG and experimental DTG curve of three-parallel-DAEM model for five lignocellulosic biomass: (a) pine wood (PW), (b) fern (dicranopteris dichotoma) stem (FS), (c) wheat straw (WS), (d) sugarcane bagasse (SB) and (e) jute (corchorus capsularis) stalk (JS).

29

(a)

f(E) (mol/kJ)

0.08

Hemicellulose

0.06

(b)

1.2

PW FS RS SB JS

f(E) (mol/kJ)

0.10

0.04

PW FS WS SB JS

0.8

Cellulose

0.4

0.02 0.00

0.0 100

125

150

175

200

225

165

170

E(kJ/mol)

180

185

0.6

0.020 PW FS WS SB JS

0.016

0.012

0.008

Hemicellulose FS JS

(d)

0.5

f(E) (mol/kJ)

(c) f(E) (mol/kJ)

175

E(kJ/mol)

Cellulose FS JS

0.4 0.3

Ligin FS JS

0.2

Lignin 0.004

0.1

0.000

0.0 50

100

150

200

250

300

100

E(kJ/mol)

150

E(kJ/mol)

200

250

Fig.5 Gaussian f(E) of pseudo components: (a) Gaussian f(E) of hemicellulose, (b) Gaussian f(E) of cellulose, (c) Gaussian f(E) of lignin and (d) Gaussian f(E) of three components for FS and JS.

30

Table 1 Ultimate analyses and proximate analyses of the five biomass. Ultimate analyses (wt.%) C H S N Oa PW 49.37 6.12 0.19 0.08 44.24 FS 46.74 5.52 0.46 0.27 47.01 WS 42.16 6.02 0.31 0.53 50.98 SB 45.02 6.17 0.14 0.21 48.46 JS 47.74 5.70 0.81 0.93 44.82 Proximate analyses (wt.%) M A V F HHVb(MJ/kg) PW 9.17 0.51 72.87 17.45 19.88 FS 10.34 0.65 83.27 5.74 17.99 WS 5.93 7.99 71.80 14.28 16.10 SB 3.60 6.73 81.89 7.78 17.98 JS 4.72 0.82 79.35 15.11 18.80 M, moisture content; A, ash content; V, volatile matter content; F, fixed carbon content; HHV, high calorific value. a calculation by difference. b HHV=0.3491C+1.1783H+0.1005S-0.1034O-0.0151N-0.0211A (Channiwala and Parikh, 2002).

31

Table 2 Characteristic temperatures of the different zones and stages for the five biomass during pyrolysis process. Zone I Zone II Zone III MLP(%) Ti (℃) Tf (℃) Ti (℃) Tp (℃) Tf (℃) Ti (℃) Tf (℃) PW ~ FS ~ WS ~ SB ~ JS ~ Zone II Stage I

172.40 141.47 154.82 157.40 157.22

172.40 141.47 154.82 157.40 157.22

340.37 321.79 326.60 350.79 340.38

Stage II

567.24 588.00 581.16 573.17 554.85

567.24 588.00 581.16 573.17 554.85

798 796 811 797 796

88.52 90.40 86.95 95.22 93.76

Stage III

Ti (℃) Tf (℃) Ti (℃) T f ( ℃) Ti (℃) Tf (℃) PW 172.40 304.15 304.15 395.67 395.67 567.24 FS 141.47 304.45 304.45 403.32 403.32 588.00 WS 154.82 297.32 297.32 380.33 380.33 581.16 SB 157.40 301.29 301.29 407.73 407.73 573.17 JS 157.22 276.15 276.15 402.70 402.70 554.85 “~” denotes ambient temperature; MLP is the mass loss proportion of Zone II to total mass loss in TGA sets. Ti and Tf is the initial and final temperature of each zone and stage, respectively; Tp is peak temperature of DTG curves.

32

Table 3 Single DAEM kinetic parameters values of different stages. Stage I Bioma ss

E0,j

Stage II -1

Aj(s )

(kJ/mo

σj(kJ/mo

E0,j(kJ/mo

l)

l)

5.02

158.22

Stage III -1

Aj(s )

σj(kJ/mo

E0,j(kJ/mo

l)

l)

2.67

199.88

Aj(s-1)

σj(kJ/mo l)

l) PW

FS

WS

SB

JS

Mean Dev

152.03

1.80E+1 2

149.72

9.59E+1

3 5.37

172.07

1 147.80

1.47E+1

149.14

1.08E+1

5.52

167.98

2.87E+1

5.30

177.65

5.79E+1

1.34E+1

3.0389

/

196.46

6.20

169.84

2.96E+1

169.15

1.19E+1

0.4405

7.1079

/

2

2.83E+1

2.66

195.18

6.98E+1

3.00

202.36

8.27E+1

14.68

14.64

1 13.38

1 3.28

211.77

2 5.48

13.80

1

2

2 148.50

1.40

2

2 1.39E+1

1.27E+1

8.74E+1 1

2

2

143.80

4.67E+1

6.01E+1

10.36

2 2.60

201.13

0.7195

6.5859

3

1.74E+1

13.37

2 1.7730

where Mean and Dev is the mean value and deviation of parameters for different stages and biomass, respectively.

33

Table 4 Simulation results of three-DAEM-reaction model for the five lignocellulosic biomass.

PW

FS

WS

SB

JS

H C L H C L H C L H C L H C L

Ai (s-1) 7.51E+12 8.29E+12 1.27E+13 1.23E+12 2.13E+12 9.42E+10 1.74E+12 1.37E+13 2.11E+13 1.13E+12 8.78E+12 1.78E+13 2.56E+12 1.60E+13 9.66E+12

E0,i (kJ/mol) 164.56 176.40 187.99 160.82 171.04 175.71 148.12 173.45 192.85 151.67 179.54 189.69 155.42 178.33 201.60

σi (kJ/mol) 7.75 0.29 24.81 9.18 0.69 24.07 4.37 1.34 27.24 3.91 0.98 24.12 9.89 0.72 23.22

ci 0.22 0.54 0.24 0.12 0.62 0.26 0.17 0.54 0.29 0.20 0.59 0.21 0.18 0.65 0.17

Iteration 75

O.F2 Np 3.099E-4 11

Fit(%) 0.53

56

1.608E-4 11

0.39

59

2.980E-4 11

0.52

100

1.512E-4 11

0.37

61

5.317E-4 11

0.69

where H is hemicellulose, C is cellulose and L is lignin.

34

Highlights

Pyrolysis characteristics of five lignocellulosic biomass were investigated. Pyrolysis process was divided into three stages using second derivative method. A three-parallel-DAEM model was developed to simulate the pyrolysis process. A new method was used to choose initial values for three-parallel-DAEM simulation. Mass fractions of pseudo components were directly estimated by three-parallel-DAEM.

35

Characteristics and kinetic study on pyrolysis of five lignocellulosic biomass via thermogravimetric analysis.

Pyrolysis characteristics and kinetic of five lignocellulosic biomass pine wood sawdust, fern (Dicranopteris linearis) stem, wheat stalk, sugarcane ba...
848KB Sizes 0 Downloads 11 Views