3300
Energy & Fuels 2009, 23, 3300–3306
Development of a Variable Activation Energy Model for Biomass Devolatilization Enrico Biagini,*,† Ludovica Guerrini,‡ and Cristiano Nicolella‡ Consorzio Pisa Ricerche, DiVisione Energia e Ambiente, Lungarno Mediceo 40, Pisa, Italy and Dipartimento di Ingegneria Chimica, Chimica Industriale e Scienza dei Materiali, UniVersita` di Pisa, Via DiotisalVi 2, Pisa, Italy ReceiVed February 20, 2009. ReVised Manuscript ReceiVed April 9, 2009
A thermogravimetric balance is used in this work to characterize different classes of biomass fuels: residues (rice husks, olive cake, cacao shells), woods (poplar, beech, pellets), and grasses (mischantus). The effect of the heating rate is evaluated in the range 10-80 K/min providing significant parameters for the fingerprinting of the fuels. Kinetic parameters are obtained by applying traditional isoconversional methods. The activation energy as a function of the conversion reveals the multistep nature of the biomass devolatilization. Although average values allow the reactivity of different fuels to be compared, a first-order reaction model can hardly predict the biomass devolatilization in the whole range of conversions. A VEB (variable activation energy model for biomass devolatilization) model is developed, based on the results of the kinetic analysis, maintaining a simple kinetic scheme. A good agreement is obtained for the biomass residues in all HR runs in the entire range of temperatures. The multistep mechanism can be studied without assuming any chemical components or pseudocomponents, thus limiting the number of model parameters. Similarities in the optimized VEB curves for the fuels studied in this work give useful generalization parameters for biomass devolatilization modeling.
1. Introduction Biomass fuels represent a renewable energy source, which may be abundant in specific local areas and cause disposal problems. Thermochemical processes (cocombustion, gasification, and pyrolysis) constitute interesting options for biomass conversion into energy (i.e., heat and power), fuels, and chemicals.1-5 In all these processes, the continuous heating of the fuel causes the breakage of chemical bonds of the natural polymers (cellulose, hemicellulose, and lignin) and the release of volatile matter (condensable and light gases). This devolatilization is a key step in any thermochemical process of biomass conversion. Several consecutive and parallel reactions occur in a narrow range of temperatures, resulting in complex kinetic mechanisms, which require fundamental studies for their characterization. The thermogravimetric (TG) technique is the basis for a fundamental investigation and gives the fingerprinting of the fuel. It provides important data (temperature of devolatilization, volatile matter released), reaction kinetics, and significant parameters for characterizing different classes of solid fuels (see, for instance, ref 6). When biomass devolatilization is studied in a thermogravimetric balance, a composite profile in the * Corresponding author: e-mail:
[email protected]; phone: +390502217850; fax: +390502217866. † Divisione Energia e Ambiente. ‡ Universita ` di Pisa. (1) Abbas, T.; Costen, P. G.; Lockwood, F. C. Symp (Int.) Combust. 1996, 26, 3041–3058. (2) Spliethoff, H.; Hein, K. R. G. Fuel Process. Technol. 1998, 54, 189– 205. (3) Sami, M.; Annamalai, K.; Wooldridge, M. Prog. Energy Combust. Sci. 2001, 27, 171–214. (4) Bridgwater, A. V. Chem. Eng. J. 2003, 91, 87–102. (5) Wall, T. F. Symp (Int.) Combust. 2007, 31, 31–47. (6) Kok, M. V.; Pokol, G.; Keskin, C.; Madara´sz, J.; Bagci, S. J. Therm. Anal. Calorim. 2004, 76, 247–254.
devolatilization curve can be observed, which makes single firstorder reaction models unsuitable for kinetic modeling. Devolatilization reactions are commonly described in terms of chemical components (cellulose, hemicellulose, and lignin) or pseudochemical components. These summative models are based on the hypothesis that the devolatilization of each component proceeds independently.7,8 The reactions of components can be combined in parallel,9-15 serial,16 or complex schemes.17,18 More advanced models increases the number of components for improving the accuracy of the predictions (see, for instance, ref 19). The number of parameters in this kind of approach is very high, and it can be hardly generalized, as the resulting parameters of chemical components or pseudocomponents are remarkably different from biomass to biomass. This is because (7) Koufopanos, C. A.; Maschio, G.; Lucchesi, A. Can. J. Chem. Eng. 1989, 67, 75–84. (8) Raveendran, K.; Ganesh, A.; Khilar, K. C. Fuel 1996, 75, 987–998. (9) Caballero, J. A.; Conesa, J. A.; Font, R.; Marcilla, A. J. Anal. Appl. Pyrol. 1997, 42, 159–175. (10) Varhegyi, G.; Antal, M. J.; Jakab, E.; Szabo´, P. J. Anal. Appl. Pyrol. 1997, 42, 73–87. (11) Gronli, M. G.; Varhegyi, G.; Di Blasi, C. Ind. Eng. Chem. Res. 2002, 41, 4201–4208. (12) Manya, J. J.; Velo, E.; Puigjaner, L. Ind. Eng. Chem. Res. 2003, 42, 434–441. (13) Hu, S.; Jess, A.; Xu, M. Fuel 2007, 86, 2778–2788. (14) Orfao, J. J. M.; Antunes, F. J. A.; Figueiredo, J. L. Fuel 1999, 78, 349–358. (15) Teng, H.; Wei, Y. C. Ind. Eng. Chem. Res. 1998, 37, 3806–11. (16) Varhegyi, G.; Antal, M. J.; Szekely, T.; Szabo, P. Energy Fuels 1989, 3, 329–35. (17) Bradbury, A. G. W.; Sakai, Y.; Shafizadeh, F. J. Appl. Polym. Sci. 1979, 23, 3271–80. (18) Miller, R. S.; Bellan, J. Combust. Sci. Technol. 1996, 126, 97– 137. (19) Muller-Hagedorn, M.; Bockhorn, H.; Krebs, L.; Muller, U. J. Anal. Appl. Pyrol. 2003, 68-69, 231–49.
10.1021/ef9001499 CCC: $40.75 2009 American Chemical Society Published on Web 04/23/2009
ActiVation Energy Model for Biomass DeVolatilization
Energy & Fuels, Vol. 23, 2009 3301
Table 1. Proximate and Ultimate Analyses and Heating Values of the Materials proximate analysis (% as received)
ultimate analysis (% dry ash free)
fuel
moisture
VM
FC
ash
C
H
N
S
HHV [MJ/kg]
olive cake rice husks miscanthus wood pellets beech wood poplar wood
5.64 10.74 3.86 4.90 4.92 4.79
70.74 59.83 79.39 75.34 80.33 80.58
21.41 15.85 12.28 15.36 12.57 10.55
4.95 13.57 4.47 4.40 2.18 4.08
55.13 51.18 50.24 51.80 49.78 52.97
7.43 7.34 5.77 6.40 6.24 6.63
1.87 0.8 5 0.57 1.04 0.28 0.26
0.23 0.18 0.12 0.17 0.10 0.05
18.5 15.5 17.9 19.0 19.3 19.3
the devolatilization of biomass components depends on the way they are structured in the original biomass, on their interactions, and on the presence of ash. Other approaches in the devolatilization of solid fuels consider distributed activation energy. In these cases the variation in the activation energy simulates a series of independent reactions. The distribution function may assume different forms.20 The Gaussian distribution was successfully used for the thermal decomposition of coals (DAEM model21 and further developments) and adapted to biomasses.22,23 The aim of this work is (i) to perform a kinetic study on the devolatilization of biomass fuels of different origin and characteristics and (ii) to develop an innovative model, based on the results of the fundamental study. The simplicity of the single first-order reaction model is preserved and few parameters are required in the generalization elaborated in this work. The variable activation energy model for biomass devolatilization (VEB) is validated in the range 10-80 K/min for agricultural residues, woods, and grass. It allows an accurate simulation of biomass devolatilization and can be included in comprehensive codes for combustion and gasification. 2. Experimental Section 2.1. Materials. Six biomass fuels of different origin are studied: grass (miscanthus), wood (beech, poplar, pellets), and residues (rice husks, exhausted olive cake). They were crushed and sieved, and the fraction 90-125 µm was used for the experimental runs. The proximate and ultimate analyses and the heating values of all materials are listed in Table 1. 2.2. Equipment. A thermogravimetric balance (TG Q500 V6.1 of TA Instruments) is used for all tests. The temperature range is from ambient to 1270 K, with heating rate from 0.1 to 100 K/min. The isothermal temperature precision is 0.1 K. The thermocouple is positioned immediately above the sample. The thermal history is determined by the analyzer controller after a standard calibration procedure in the proper conditions. The weighing capacity is 1.0 g with a sensitivity of 0.1 µg. The purge gas is metered by mass flow controllers and set to 100 mL/min (STP). Pure nitrogen is used for the devolatilization tests.
The sample is loaded and dried at 380 K directly in the TG balance before each run. Usually, 5-10 min are sufficient to attain a constant weight of the sample. A constant heating rate is then programmed up to 1220 K. After cooling down the sample to 1070 K, the gas flow is switched to air and the carbonaceous residue is burnt. In all cases, two runs in the same conditions are carried out to confirm the good reproducibility of the results. Negligible discrepancies between the two runs are observed in the results reported in this work. Five heating rate (HR) programs are studied: 10, 20, 40, 60, and 80 K/min. Preliminary runs were carried out to find the optimal conditions. In particular, the initial amount of sample is a crucial parameter. It should be optimized to meet two needs: (i) to study a representative sample of the material (which is of heterogeneous nature) and (ii) to avoid significant heat transfer limitations in the entire range of HR. 5-6 mg was found to be the optimal initial sample mass in this study.
3. Results 3.1. Experimental Results. The weight loss curves are used to calculate the conversion R of devolatilization for all materials under different heating rates: R(t) ) 1 -
w(t) - w∞ , T ) T0 + HRt w0 - w∞
(1)
where t and T are the time and the temperature, respectively, of the experimental run (linearly related with the heating rate); w is the actual weight of the sample; w∞ is the final weight of the sample; and w0 is the initial weight of the dry sample. The derivative weight loss curves (dtg) are used to provide significant parameters of devolatilization. Curves obtained under all heating rates are compared in Figure 1 for the case of beech wood. Similar comparisons can be made with other biomasses. The composite nature of these materials can be observed in the subpeaks of the dtg curves. In general, the devolatilization profile is complex with a well-defined main peak, a shoulder at earlier temperatures, and a long tail zone. The analysis of subpeaks is generally performed by using deconvolution procedures, and each peak is ascribed to the decomposition of chemical
Figure 1. Comparison of conversion and dtg curves for beech wood under different heating rates.
3302
Energy & Fuels, Vol. 23, 2009
Biagini et al.
Table 2. Characterization of Biomass Devolatilization (HR 20 K/min) fuel
Tonset [K]
Tmax [K]
Toffset [K]
dtgmax [%/s]
Rmax
R770K
olive cake rice husks miscanthus wood pellets beech wood poplar wood
484 516 527 525 531 513
608 624 635 644 647 644
747 646 659 664 666 662
0.160 0.195 0.320 0.251 0.309 0.285
0.538 0.587 0.648 0.639 0.673 0.663
0.875 0.882 0.932 0.905 0.935 0.907
Table 3. Devolatilization Parameters at Different Values of HR Tonset [K]
Tmax [K]
Toffset [K]
dtgmax [%/s]
Rmax
R770K
HR HR HR HR HR
10 20 40 60 80
K/min K/min K/min K/min K/min
517 527 534 537 542
620 635 646 654 656
642 659 672 680 683
0.172 0.320 0.600 0.856 1.080
0.644 0.648 0.656 0.659 0.660
0.934 0.932 0.933 0.931 0.930
HR HR HR HR HR
10 20 40 60 80
K/min K/min K/min K/min K/min
510 525 531 536 539
630 644 658 666 668
0.135 0.251 0.473 0.682 0.914
0.645 0.639 0.658 0.645 0.640
0.909 0.905 0.900 0.885 0.878
miscanthus
Wood Pellets 649 664 677 687 690
components or pseudocomponents (extractives, cellulose, hemicellulose and lignin) for obtaining kinetic parameters.12-14 Important information can be drawn from the analysis of these curves for the comparison of different biomass fuels, based on the definition of similar parameters that characterize the behavior of different materials and quantify the effect of the heating rate on the devolatilization: (1) the onset temperature of devolatilization (Tonset) quantifies the temperature at which the devolatilization can be considered to start and is defined as the temperature of the earliest relative maximum in the second derivative of the weight loss curve during the devolatilization of the dry material; (2) the maximum weight loss rate (dtgmax); (3) the temperature at the maximum weight loss rate (Tmax), which, together with dtgmax, is useful for comparing different materials, as they are a measure of the fuel reactivity and the basis for some kinetic analysis (e.g., Kissinger method24); (4) the conversion at the temperature Tmax (Rmax); (5) the offset temperature of devolatilization (Toffset) is defined at the latest relative minimum in the second derivative of the weight loss curve and quantifies the temperature at which the main devolatilization can be considered complete; (6) the conversion at 770 K (R770K), related to the importance of the tail zone of devolatilization. This temperature was chosen because it is higher than Toffset for all biomasses studied in this work and thus allows the comparison of the extent of the final devolatilization for different materials. The values of these parameters are listed in Table 2 for all materials at 20 K/min. The devolatilization starts at temperature between 500 and 520 K, whereas the temperature for the maximum of the dtg (which can be roughly approximated with the temperature of devolatilization of cellulose) is between 610 and 640 K. These parameters depend on the HR. In general, the higher the heating rate, the higher the values of Tonset, Tmax, and Toffset. The results for mischantus and wood pellets are listed in Table 3. Similar results are obtained for the other materials, from a qualitative point of view. The conversion also weakly (20) Burnham, A. K.; Braun, L. R. Energy Fuels 1999, 13, 1–22. (21) Anthony, D. B.; Howard, J. B. AIChE J. 1976, 22, 625–656. (22) Lewellen, P. C.; Peters, W. A.; Howard, J. B. Symp (Int.) Combust. 1977, 16, 1471–1480. (23) Rostami, A. A.; Hajaligol, M. R.; Wrenn, S. E. Fuel 2004, 83, 1519–25. (24) Kissinger, H. E. Anal. Chem. 1957, 29, 1702–1706.
increases with HR. In all cases, the conversion at 770 K is around 0.9, and this means that 10% of volatiles will devolatilize at higher temperatures. 3.2. Kinetic Analysis. The kinetic analysis of solid decompositions is usually based on a single-step kinetic equation: E dR ) A exp f (R) dt RT
( )
(2)
where R is the extent of conversion; A and E are the Arrhenius parameters, pre-exponential factor and activation energy, respectively; f (R) is the reaction model, which is a function of R. Generally, the triplet A, E, and f (R) is needed to define the kinetics of a reaction. A single first-order reaction (SFOR) model assumes f (R) ) 1 - R. Model-free methods allow for evaluating the Arrhenius parameters without choosing the reaction model. In particular, the isoconversional methods (e.g., Friedman,25 Flynn,26 Kissinger24) yield the effective activation energy as a function of the extent of conversion. In the case of nonisothermal runs, they require a set of experimental tests at constant heating rate. This eliminates the compensation effects in determining the kinetic parameters, avoiding the possibility of different couples of preexponential factor and activation energy to describe reasonably well the same weight loss curve.27 The knowledge of the dependence of E on R assists in both detecting multistep processes and drawing certain mechanistic conclusions, and it is sufficient to predict the reaction kinetics over a wide temperature region. The Friedman method has been selected in this work as the starting point for the VEB model development. According to Friedman,25 E/R can be obtained for a given value of R by plotting ln(dR/dt) against 1/T. This is a modelfree method that can be applied to data sets obtained at different heating rates:
(
ln HR
dR E ) ln A + ln f (R) dT RT
)
(3)
This is a differential method, which can be applied to integral data (e.g., TG data) only after their numerical differentiation. (25) Friedman, H. L. J. Polym. Sci. 1965, 50, 183–195. (26) Flynn, J. H.; Wall, L.A. J. Res. Nat. Bur. Stand. 1966, 70A, 487– 493. (27) Agrawal, R. K. Thermochim. Acta 1985, 90, 347–51.
ActiVation Energy Model for Biomass DeVolatilization
Energy & Fuels, Vol. 23, 2009 3303
biomass fuels. So, a further step in the kinetic model should be done to give a better simulation of biomass devolatilization. 4. Development of the VEB Model and Discussion
Figure 2. Example of calculating the activation energy (reported in kJ/mol) at different conversions by the Friedman method (beech wood).
Figure 3. Comparison of the activation energy obtained by using the Friedman method for the devolatilization of all biomasses.
In his original analysis, Friedman considered only the possibility of a single nth-order reaction. If the value of E varies with the extent of conversion, then the results should be interpreted in terms of multistep reaction mechanisms. Let Eisoc(R) define the function obtained from the isoconversional method. The temperature-conversion data are elaborated according to the Friedman method for the kinetic analysis. An example of the application of the method is reported in Figure 2. The linear fit of isoconversional points obtained under different heating rates gives the kinetic parameters. Similar graphs are obtained for all materials. The results of the activation energy Eisoc(R) are reported in Figure 3 for all materials as a function of the conversion. In general, E varies with the extent of conversion, and three ranges can be identified: (1) at low conversions E increases, to become practically constant for values higher than 10%; (2) a plateau can be observed in a wide range of conversions; (3) for conversions higher than 70% E increases and then decreases, with an irregular trend. As a matter of fact, the linear fit of isoconversional points is good only in the first two ranges of conversion, whereas it is not acceptable at high conversion (see Figure 2). This can be due to uncertainties in the experimental data as well as steep variations in the activation energy, which invalidate the hypothesis of the Friedman method, making it not applicable. The characteristic equations of the isoconversional methods are indeed derived assuming a constant activation energy. This assumption introduces some systematic error in estimating Eisoc(R), when this latter varies significantly. Average values of E in the plateau range are compared in Table 4 for all materials. These data can represent the reactivity of devolatilization and can be used to compare different fuels. Although they may be sufficient for a preliminary approximation of biomass devolatilization, a SFOR model with constant activation energy can hardly describe the composite nature of
Biomass devolatilization consists of a complex system of reactions. Biomass fuels contain reactive components responsible of the initial steps of devolatilization, and this can be observed in the small values of E at low conversions. The intermediate values of conversion are characterized by quasiconstant values of E, while the final tail of devolatilization can be caused by the less-reactive structure of the remaining solid after the main devolatilization and thus a high value of E should be expected at high conversions. For this reason, a SFOR model (substituting the average value of the activation energy to E and f (R) ) 1 - R in eq 2) can give only a representation of the main devolatilization step, but can hardly give a description of the earlier devolatilization and the long final tail. The agreement (not reported here) between experimental and SFOR model results was proved to be not accurate (kinetic parameters are reported in Table 4). However, the SFOR model is the simplest approach, easy to be included in comprehensive codes for combustion and gasification, and therefore some modifications may improve the approach to benefit these features. Preliminarily, some trials were done by introducing in eq 2 the activation energy Eisoc(R) directly obtained by the isoconversional method. We assumed the first-order reaction by substituting f (R) ) 1 - R. The values of the pre-exponential factor A were determined from the best fit of the experimental results by using Eisoc(R) in the entire range. The best fit was obtained by minimizing the deviation with respect to the experimental results:
dev(%) )
∑ HR
∑ |R
mod
N
N
- Rexp | × 100
(4)
where Rmod and Rexp are the model and experimental conversions, respectively; N is the number of results considered for each curve obtained under a certain HR; and the sum is extended to all runs under different heating rates (10-80 K/min). The results (reported elsewhere28) were not satisfying. Observing the curves of Figure 3, the value of Eisoc(R) in the earlier steps is very low, giving unrealistically low onset temperatures. Vice versa, in the tail zone, the trend of Eisoc(R) is irregular, causing unrealistic behavior in the final zone of conversion made of unlikely subpeaks. In some cases, too high values of the activation energy in this zone made the devolatilization stop and caused an underestimation of the final conversion. As a matter of fact, only at the intermediate values of conversion could the agreement be considered good. This was the basic observation to develop a VEB. The function Eisoc(R) can be modified to give a more satisfying fit of the experimental results. The following points describe the development of the VEB model. The reference scheme is in Figure 4. (1) The intermediate values of Eisoc(R) are preserved, selecting an interval [R1,R2] of the main devolatilization where the experimental noise is low and the activation energy is a weak function of the conversion (see the plateau in Figure 3). Let’s say this function is Eplat(R). A preliminary value of the pre(28) Guerrini L. Kinetic Analysis of Thermochemical Conversion Processes of Biomasses; BSC Thesis; University of Pisa: Italy, 2008.
3304
Energy & Fuels, Vol. 23, 2009
Biagini et al.
Table 4. Comparison of Kinetic Parameters and Deviation Obtained by Different Approaches for All Biomasses SFOR model E constant
VEB model E overallb
VEB model E specific
mean E plateau
[kJ/mol]
ln A
ln A
deva
ln A
deva
olive cake rice husks wood pellets beech wood miscanthus poplar wood
180.2 187.0 173.2 190.3 185.7 168.9
31.8 32.5 28.6 31.9 31.6 28.1
33.10 33.12 29.39 32.50 33.33 28.72
5.38 8.35 4.23 7.42 4.21 9.70
31.9 31.41 30.38 30.34 30.91 30.88
20.8 9.34 9.51 15.4 15.3 13.4
a The deviation is calculated between the model and experimental results according to eq 4. b The global deviation of all runs and materials is 83.8; the average deviation of each material is 13.9.
Figure 5. EVEB curves for all materials (symbols). The solid curve represents Eoverall for the generalized approach.
Figure 4. Scheme of the VEB model development.
exponential factor Ap is calculated from the minimization of the deviation of eq 4, limiting the comparison to this interval of R. (2) Two values of E are fixed for R ) 0 (Eon) and R ) 0.9 (Eoff) and, correspondingly, two cubic functions are defined to interpolate (0, Eon) and the points of Eplat and the points of Eplat and (0.9, Eoff), respectively. Let define these functions Ecub1(R) in the range [0, R1] and Ecub2(R) in the range [R2,1]. (3) EVEB(R) is defined in three intervals as shown in the scheme of Figure 4. Equation 2 is then numerically integrated with EVEB(R) in place of E, and the results are compared with the experimental curves. The procedure is optimized by varying Eon and Eoff and considering the widest interval [R1,R2] for each material for minimizing the deviation of eq 4. (4) Finally, the value of A can be further optimized to minimize the deviation in the entire range of R for all runs. The resulting functions EVEB(R) are reported in Figure 5 for all materials. The correspondent values of A and the deviations
are compared in Table 4. The difference of EVEB(R) with respect to Eisoc(R) is pronounced at the earlier conversions and in the tail zone, with EVEB(R) exhibiting milder variations. The interval [R1,R2] is wider for some materials and narrower for those materials with an irregular trend in Eisoc(R) (e.g., miscanthus and olive cake). For all materials, Eon results between Eisoc(0) and Eisoc(R1), while Eoff results higher than Eisoc(R2). Finally, the comparison of the experimental results and the results of the VEB model is reported in Figure 6a for some materials. The agreement is good for all runs and all materials. It is worth noting that the specific behavior of the material in the devolatilization curves is guaranteed by adopting Eisoc(R) in the main devolatilization interval [R1,R2], which is a result of the fundamental study reported in the results section. This is not really constant, but varies weakly with the conversion allowing the composite shape of the main devolatilization to be well-predicted. Since many similarities can be observed in the curves of Figure 5, a generalization of the activation energy for all biomasses can be achieved. This is done by: (1) defining a function Eoverall(R), calculated as a cubic curve, which will be valid for all biomasses and (2) optimizing the pre-exponential factor A, which will be specific for each material. These two points are iteratively met to minimize the deviation extended to all runs and all biomasses:
dev(%) )
∑ ∑
biomass HR
∑ |R
mod
N
N
- Rexp | × 100
(5)
The Eoverall(R) obtained is shown in Figure 5 as a solid curve among the specific EVEB curves of all biomasses. The values of A are reported in the last columns of Table 4 along with the deviation calculated for each material. Finally, the comparison of experimental curves and those obtained with the generalized approach is reported in Figure 6b. When comparing these results with those of the VEB model optimized for each material, the biggest discrepancy can be observed in the final zone of the conversion. The higher values
ActiVation Energy Model for Biomass DeVolatilization
Energy & Fuels, Vol. 23, 2009 3305
Figure 6. Comparison of the experimental conversions and those obtained by (a) the VEB model optimized for each material and (b) the generalized VEB model with Eoverall(R). Symbols: experimental results; continuous lines: model results.
of the deviation for the case of the Eoverall with respect to the case of EVEB with specific optimization (Table 4) denote a lessaccurate approach. Obviously, the generalization of the activation energy precludes the possibility of simulating the peculiar behavior of subpeaks of each material. However, it represents a better approximation with respect to a SFOR model with constant activation energy. It is worth noting that Eoverall is calculated for all biomasses studied here, so a single parameter (A) has to be determined for each material to obtain an enhanced level of approximation, sufficient for many applications. As a matter of fact, the number of materials studied here is not so high. Further biomasses can be included according to the method developed here, so that the Eoverall(R) curve can be improved for generalizing a higher number of materials. Indeed,
the biomasses can be separated in groups (woods, grasses, residues) for a better step of approximation. A better definition of Eoverall(R) can be made instead of a simple cubic curve. These suggestions will be pursued in future work. Here it is worth remarking that an innovative approach in the kinetic model of biomass devolatilization has been developed and validated on the basis of the experimental results obtained on biomasses of different origin and characteristics. The results of the VEB model are good for the specific approach (with EVEB optimized for each biomass) and derive mainly from a fundamental study. A simple kinetic scheme is adopted without assuming any chemical or pseudo components, even though the multistep mechanism can be studied. The generalized approach (with Eoverall obtained for all materials) is noteworthy, because,
3306
Energy & Fuels, Vol. 23, 2009
once a significant number of runs and materials are studied for determining the overall curve of the activation energy, a single parameter allows us to simulate the devolatilization of all biomasses. 5. Conclusions An innovative approach in the kinetic model of biomass devolatilization has been developed and validated on the basis of the experimental results obtained on biomasses of different origin and characteristics. The results of the VEB are good for the specific approach (with EVEB optimized for each biomass) and derive mainly from a fundamental study. The basis of this latter is the characterization of biomasses in a thermogravimetric balance and the isoconversional method of Friedman for the kinetic analysis. The VEB model adopts a simple kinetic scheme
Biagini et al.
without assuming any chemical components or pseudocomponents, however the multistep mechanism can be studied. Studying the similarities in the obtained results, a generalized approach (with Eoverall obtained for all materials) has been attained. The agreement in this case was obviously less accurate but noteworthy because, once the variable activation energy function was optimized for a significant number of runs and materials, a single parameter allowed us to simulate the devolatilization of all biomasses. Acknowledgment. This work collects part of the results of the FISR Italian Project on “Integrated Systems for Hydrogen Production and Use in Distributed Generation”. EF9001499
