Prediction Of Rapid Biomass Devolatilization Yields With An Upgraded Version Of The Bio-CPD Model G.Vizzini 1, A.Bardi 2, E. Biagini 1 , M. Falcitelli 1, and L. Tognotti 3(*) 1. Consorzio Pisa Ricerche –Divisione Energia e Ambiente SCaRL, Lungarno Mediceo 40, 56127 Pisa – ITALY
[email protected],
[email protected],
[email protected] 2. Enel SpA - Ingegneria e innovazione. Area Tecnica Ricerca, Via A. Pisano 120 - 56122 Pisa – ITALY,
[email protected] 3. Università degli Studi di Pisa – Dipartimento di Ingegneria Chimica,Chimica Industriale e Scienza dei Materiali, Via Diotisalvi 2, 56100 Pisa – ITALY
[email protected] 1. Introduction Biomass represents renewable energy source, that curb the emissions of SOx, NOx and heavy metals. Biomass is a near CO2-neutral fuel, that can be used in pyrolysis, gasification, combustion and co-combustion to reduce the consumption of fossil fuels. The direct combustion of biomass feedstocks is problematic and scarcely applied for power generation, especially because of technological (low heating value, flame instability due to the high reactivity and volatile matter content, fouling, slagging, and corrosion phenomena) and economical problems. Other processes usually find a practical application, namely, cocombustion, pyrolysis, and gasification. Devolatilization is the first step in all thermal processes and it produces a large amount and a variety of volatile species and a solid residue. Therefore, besides experimental characterization of biomass devolatilization products, the development of predictive numerical models is still a goal of primary importance for both improving and designing industrial processes. This work introduces an upgraded version of a network devolatilization model, called bio-CPD [1], for the rapid pyrolysis of biomass feedstocks. Respect to the other affirmed network models such as the bio-FG DVC [2] and bio-FLASHCHAIN [3], the formulation of bio-CPD is more sophisticated. Biomass is represented as a chain copolymer of cellulose, hemicellulose and a lignin-like component. The chemical structural and kinetic parameters for the each component have been proposed, based on theory, literature review and curve-fitting. The framework of the kinetic scheme for bond rupture adopted by the CPD [4] was maintained, but the formulation of the mass balances was substantially revised introducing a population balance for the liquid n-mers, and the ability to differentiate the fate of the tar yields in the vapor phase and the liquid metaplast remaining in the particle. In this way the role of the secondary reactions of tar cracking in the gas phase has become more apparent. Further, a method for the prediction of the main chemical species of volatile yields has been proposed. The predictions have been compared with a selection of experimental data from literature, showing fine agreement. 2. Model Development The original approach proposed by Fletcher with the Chemical Percolation Devolatilization Model (CPD) [4] for coals, was used in the development of bio-CPD as starting step. Besides, it was extended to predict rapid pyrolysis of cellulose, hemicellulose and lignin. The input data needed by the present version of bio-CPD model are: ultimate and CHL analysis, thermal history of the particle and operating pressure. The model describes devolatilization of a generic biomass through an idealized description of the structure of Cellulose, Hemicellulose and Lignin. For each component (Ce, He, Li), the fractional change in the mass as a function of time is divided into three parts: light gas, tar precursor fragments, and char. (*)
Corresponding author:
[email protected]
III-4, 1
Italian Section of the Combustion Institute B IO -C P D C O D E MODEL NEEDS U ltim a te A nalys is (C H N O )
C H L A na lysis
D ata N
A zevedo [6] C o rrelation [7] H yp othesis on C ellulose H em ic ellu los e C om position
OUTPUT T herm a l h is tory D a ta a v aila b le?
av ailab le?
N
Y
C H L m o del
H em icellu lo s e C e llu lo se L ig n in
Y
P artic le T an d H eatin g R a te
A ss um p tio n of C haracteristic T em perature C H E M K IN
L ign in C om position from e le m ent balan ce
M ateria l b ala nce
M o dific ation o f m ain equ atio ns: O rig in al C P D
M ac ro -p ro d u cts C h ar T ar G as (C H N O )
¾ B ridg es’ scission k inetics ¾ F rag m e nts’ popu la tion cou nt ¾ F lash vapo rization ¾ C ross-link in g ¾ T ar-crack ing
G as s p ec ia tio n (C O C O 2 H 2 C H 4 NH3 HCN N2 … )
Fig. 1 Calculation flow scheme for the bio-CPD code. Algebraic expressions are obtained for each part using percolation lattice statistics. In accounting for mass in the metaplast (tar precursor fragments), the part that vaporizes is treated in a manner similar to flash vaporization, where it is assumed that the finite fragments undergo vapor/liquid phase equilibration on a time-scale that is rapid with respect to the bridge reactions. As an estimate of the vapor/liquid that is present at any time, a vapor pressure correlation based on a simple form of Raoult’s law is used. For the part of the metaplast that reattaches to the coal lattice, a crosslinking rate expression is used. In order to make the model capable to describe biomass devolatilization, it was necessary to reformulate most of CPD equations; the following modifications were introduced: ¾ A population balance was formulated in order to distinguish between the fate of the side chains leaving the particle with tar vapors, and those remaining in the liquid metaplast. In fact only the latter are gas precursor during percolative depolymerization. ¾ Evaluation and optimization of structural and kinetic parameters, for the three components, were carried out. ¾ A vapor pressure correlation was added for the three components. ¾ Secondary reactions of tar-cracking were introduced, assuming first order kinetics and Arrhenious equations [5]. ¾ A ChemKin routine was introduced for performing light gas equilibrium calculations. It is important to stress that fast pyrolysis yields may depends on the thermal history and not just on the heating rate and final temperature. This happen if the process has any limits such that inadequate heat transfer or coarse dimension of the particles. In the present approach if the temperature profile is missing, it is calculated by another assessed code, the so called CHL model [6], as schematized in Fig 1. The latest upgrading of bio-CPD concerns the capability to predict the yields of main gas species: a procedure based on hypothesis of speciation for char, metaplast and tar released from cellulose, hemicellulose and lignin was proposed. In this way bio-CPD provides, through closing of material balance, light gas elemental speciation. Then a ChemKin routine is used for calculating the gas species at a characteristic equilibrium temperature. 2.1. Structural and kinetic parameters for Cellulose, Hemicellulose and Lignin For the development of bio-CPD, the basic assumption that each biomass fuel consists of cellulose, lignin and hemicellulose (hard- and softwood) is assumed. Thus the problem was to III-4, 2
Italian Section of the Combustion Institute
assess kinetic, structural parameters and vapour pressure correlation for each component. This work was accomplished by the following method [1]: another assessed code “CHL” [6] was used, as generator of “ideal” experiments for the individual biomass components; initial values for structural and kinetic parameters were guessed, based on theory and literature review; the set of the bio-CPD parameters was tuned with an optimization program, which looks for the best matching between CPD and CHL predictions. The resulting structural and kinetic “best” parameters are shown in the Table 1. Structural Parameters MCL Mδ σ+1 p0 c0
Cellulose 162 15.96 2.005 0.9998 0.0
Hemicellulose 173.45 10.28 2.7 0.79 0.267
Lignin HW 207.5 39 3.5 0.71 0.1
Lignin SW 186 34 3.5 0.71 0.1
93718 157.46 0.58
93718 157.46 0.58
87058 299 0.59
87058 299 0.59
2.14E+15 54069 2662 3.02 0.0 1.19E+08 26743 879
7.96E+14 47636 1945 1.62 0.0 1.69E+08 18467 1802
2.60E+15 54000 3972 3.9 0.0 3.00E+15 6600 4776
2.60E+15 54000 3972 3.9 0.0 3.00E+15 6600 4776
3.0E+06 26169
1.49E+06 26169
1.49E+06 26169
1.49E+06 26169
3.00E+15 65000
3.00E+15 65000
3.00E+15 55680
3.00E+15 55680
Vapor Pressure Correlation
Piv = α exp(-β MWiγ / T ) α (atm.) β (g-γ moleγ K) γ Kinetic Parameters Ab (s-1) Eb (cal/mol) σb (cal/mol) ρ (kδ/kc) Ec (cal/mol) Ag (s-1) Eg (cal/mol) σg (cal/mol) Tar cracking Atc (s-1) Etc (cal/mol) Crosslinking Acr (s-1) Ecr (cal/mol)
Table 1 Structural and kinetic parameters optimized for Cellulose, Hemicellulose and Lignin. Cellulose is the most abundant component of a biomass, and it is a linear polymer with few branches. Its monomer is levoglucosane (C6H10O5), dehydrated form of the glucose and its fragmentation degree is very low, therefore, the value of p0 was assigned approximately at 1 (0.999). The coordination number σ+1 is near to 2, because each monomeric unit is linked through two oxygen bridges. For cellulose, the cluster unit was selected as levoglucosane, so the selected value for MCL is 162 u.m.a, which is its molecular weight [7]. The molecular weight per side chain (Mδ) was supposed to be small, because it has to be considered a mean value expressed on a per monomer basis of the mass of the terminal chains and of the lattice amorphous defects. The higher value of Mδ obtained by fitting, perhaps doesn’t correspond to a real presence of lateral chain, but it reflects the possibility for the some fragments of metaplast to decompose before becoming tar vapor. Finally, cellulose pyrolysis produces much lower char yields, when compared with hemicellulose and lignin, so the initial value for the population of char bridges (c0) is set to 0. The exact values of all the structural parameters, together with kinetic parameters need to be fitted to data on light gases, tar and char yields over a meaningful range of temperatures. To that purpose CPD model was used together with an optimization program to find the best parameter values. Hemicellulose is a heteropolymer which contains several sugar monomers (C5 and C6). It is a more branched polymer and more
III-4, 3
Italian Section of the Combustion Institute
fragmented than cellulose. For this reason, a value between 2 and 3 for σ+1 and a value of 0.79 for p0 were assigned. An average molecular weight was assigned for MCL, calculated using all molecular weights of monomers C5 and C6 (MCL=173). As well as for cellulose, the value of Mδ for hemicellulose was obtained by fitting. Lignin is similar to low rank coal with a three-dimensional lattice structure of complex racemic polymer. Its structure is not well defined and depends on the biomass type. One of the main problems when studying lignin is the impossibility of extracting it from the biomass without chemically modifying it. Using 13 CNMR analysis and theory research, previous investigations [8] proposed that for the CPD model coniferyl, sinapyl and p-coumaryl alcohols are the base clusters. The set of values for the structural parameters recently proposed in the above citation was adopted in the present work as initial “guess” for a further optimization. 2.2. Predictive method for volatile yield speciation and composition of light gases In the present work, a general method is proposed to provide the speciation of volatile products released from biomass, including the yields of all main gas species. The starting step is the elemental compositions of cellulose and hemicellulose: for the cellulose elemental composition of levoglucosane (C6H10O5) is assigned without uncertainty. For hemicellulose, several thermogravimetric analysis have shown that the main product from pyrolysis is Xylose (C5H8O4), therefore its elemental composition was assigned to the hemicellulose of parent biomass. The same reasoning was applied to assign composition of char, metaplast and tar from cellulose and hemicellulose: when cellulose undergo to thermal degradation at low temperatures (200 °C), it already suffers a depolymerization mechanism that proceeds through molecular severance. Therefore, the main product from pyrolytic cellulose is levoglucosane. Subsequently, this species may be involved in several degrading reactions that proceed toward further intermediary products [9]. Elemental composition of levoglucosane (C6H10O5) and xylose (C5H8O4) are assigned to both the metaplast and tar released from cellulose and hemicellulose respectively, which are here considered fragments n-mers whose repetitive units are levoglucosane and xylose [9]. The composition of char was assumed as carbon. This is explained by crosslinking mechanism, that generates a chemical rearrangement of the metaplast on the solid structure, through the occurrence of stable carbon-carbon bonds. However, the presence of heteroatoms in the organic matrix is exclusively ascribed to the residual metaplast, which does not undergo crosslink, and to the unfragmented residual mass. Composition of parent components is obviously assigned to this residual mass. The composition of lignin is a direct consequence of the assignments already established for the other two components. It is determined by difference from biomass ultimate analysis closing elemental mass balance. The composition of lignin for each biomass results different agreeing with its polymorphous character. Whereas elemental composition of a pseudo-species with 11 atoms of carbon, 13 atoms of hydrogen and 3 atoms of oxygen is assigned to the metaplast and tar in every case. The elemental composition of this virtual compound is the average of the three main hydroxycinnamyl alcohols constituting the lignin structure. Now, starting from component and ultimate analysis of a generic biomass, bio-CPD gives the complete distribution of the element in the final pyrolysis products: char, tar and gas. The elemental composition of light gases are calculated by difference. Then gas species distributions at the local equilibrium state is predicted by minimizing the free-Gibbs energy, by the routine EQUIL [10] of the ChemKin code. 3. Results and Discussion The predictions of bio-CPD are evaluated with the experimental dataset on flash pyrolysis reported by Scott et al. [11]. This dataset was selected because the mass and elemental closures were satisfied within tight tolerances, so the product distribution was complete and
III-4, 4
Italian Section of the Combustion Institute
self-consistent. Two feed materials, Avicel cellulose and red maple sawdust, were used in these tests, with properties shown in Table 2. Tests were carried out using two type of rectors with heating rate of 14000 K/s and gas residence time of 0.5. Particle residence times are not as precisely known and varied with the reactor configuration and type. The temperature profile for each test was calculated with CHL code [6] and then used in bio-CPD calculations; the solid residence time resulted from 0.5-0.7 s and 10-12 s respectively at higher and lower temperatures. The comparison between the final products calculated with the experimental data are shown in Fig. 2 and Fig. 3. Biomass Characterization
Eastern Red Maple (wt% daf)
Cellulose
43.51
Hemicellulose
23.28
Lignin
23.21
Avicel Cellulose (C6H10O5)n Ultimate Analysis (wt% daf) C 100.0 H 0.0 O N 0.0 VM
Eastern Red Maple (wt% daf) 48.5 6.1 44.9 0.5 82.3
Avicel Cellulose (C6H10O5)n (wt% daf) 44.4 6.2 49.4 0.0 -
Table 2 Chemical characterization of Eastern Red Maple and Avicel Cellulose. For cellulose, char, tar and gas yields predicted by bio-CPD agree better with experimental data (Fig. 1). Light gases yields are shown in the Fig 3., they are calculated using a chemical equilibrium approach: depolymerization with flash distillation produces three macro-species (char, tar, gas) and bio-CPD also predicts their elemental speciation. 120 100
tar exp gas exp char exp
70
tar bio-CPD gas bio-CPD char bio-CPD
60 50
Yields [wt %]
Yields [wt %]
80 60 40 20
CO exp CO2 exp C2H4+C2H2 exp CH4 exp bio-CPD equi1 bio-CPD equi2
40 30 20 10
0
0 500
600
700
800
900
Temperature [°C]
500
600
700
800
900
Temperature [°C]
Fig. 2 Pyrolysis yields for cellulose. Left: macro-products . Right: light gas species. Symbols denote experimental results from [11], lines denote bio-CPD predictions. Furthermore the model uses the speciation of light gases and thermodynamic database of ChemKin code and finally provides mass fractions of major gas species. Profiles of CO, CO2, CH4 and hydrogenated products yields were calculated at two equilibrium temperatures: in the case called “equi1” the equilibrium temperature is the final temperature of pyrolysis, whereas in the case called “equi2”, the equilibrium temperature was frozen at 900 K for the tests carried out at high temperature (1000-1100 K). The results obtained suggests that at high temperature kinetic mechanisms become very important in the pyrolysis, because they can slow gas species evolution toward the equilibrium state.
III-4, 5
Italian Section of the Combustion Institute 100
tar exp gas exp char exp
CO exp CO2 exp bio-CPD equi1
40
60
Yields [wt %]
Yields [wt %]
80
50
tar bio-CPD gas bio-CPD char bio-CPD
40
20
C2H4+C2H2 exp CH4 exp bio-CPD equi2
30
20
10
0
0
400
500
600
700
800
Temperature [°C]
400
500
600
700
800
Temperature [°C]
Fig. 3 Pyrolysis yields for Eastern Red Maple. Left: macro-products. Right: light gas species. Symbols denote experimental results from [11], lines denote bio-CPD predictions. 1. Final remarks and future work An upgraded version of a network devolatilization model, called bio-CPD, for the rapid pyrolysis of biomass feedstocks was introduced. Given the proximate and ultimate analyses and/or the component analysis (in terms of cellulose, hemicellulose and lignin), the thermal history and pressure, bio-CPD predicts the complete distribution of primary devolatilization products from a wide set of biomasses, including the yields of tar and all major gas species, the elemental compositions of tars and chars, and the molecular weight distribution of tar. The chemical structural and kinetic parameter for each component (cellulose, hemicellulose and lignin) were developed based on theory literature review and curve fitting. Secondary reactions of tar-cracking and predicting methods for the gas speciation were introduced. The predictions were compared with a selection of experimental data from literature, showing fine agreement. Wider experimental dataset and more detailed kinetic schemes are needed to refine the predictions of the minor light gas species. Future work will be devoted towards two fields: 1) the integration of bio-CPD in more complex kinetic schemes, including secondary reactions for tar cracking and soot formation; 2) the assessment of the devolatilization predictions for the coals in gasification environment at elevated pressure, adopting the improved formulation of the CPD model. References 1.
Biagini E., Falcitelli M., Tognotti L. 29° Meeting on Combustion – Italian Section of the Combustion Institute 14-17 June 2006 Pisa, Italy. 2. Chen, Y., Charpenay, S., Jensen, A., Wojtowicz, M.A. and Serio, M.A.. (1998). 27th International Symposium on Combustion. The Combustion Institute. pp. 1327-1334. 3. S. Niksa, Proc. 28th Int. Symp. on Combustion, Edinburgh, UK, 2000, 2727-2733. 4. Fletcher, T., Kerstein, A. R.: Technical report, Combustion Research Facility, Sandia National Laboratories, 1992 5. Rath, J. Staudinger, G.: Fuel, 80:1379–1389, 2001. 6. Biagini, E.: PhD thesis, Chemical Engineering - University of Pisa - Italy, 2003. 7. Sheng, C., Azevedo, J.L.T.: Proceedings of the Combustion Institute, Volume 29,2002/pp. 407-414. 8. Pond, H. R., T. H. Fletcher, and L. L. Baxter, 3rd Annual Joint Meeting of the U.S. Sections of the Combustion Institute, Chicago, IL (March 16-19, 2003). 9. Cuoci A., Faravelli T., Frassoldati A., Granata S., Migliavacca G., Ranzi E., Sommaria S. Proceedings of The Combustion Institute (2007). 10. A.E. Lutz, F.M. Rupley, R.J. Kee, W.C. Reynolds, EQUIL: a program for computing chemical equilibria, Technical report, Sandia National Laboratories, 1996. 11. Scott D.S., Piskorz J., Bergougnou M.A., Graham R., Overend R.P.: Ind. Eng. Chem. Res.1988,27, pp. 8-15.
III-4, 6