Biomass fuels are being increasingly used for domestic heating and power generation

Transcription

1 /05/$ # 2005 Institution of Chemical Engineers Trans IChemE, Part B, November 2005 doi: /psep Process Safety and Environmental Protection, 83(B6): NUMERICAL SIMULATION OF THE BURNING CHARACTERISTICS OF THERMALLY-THICK BIOMASS FUELS IN PACKED-BEDS Y. B. YANG, V. N. SHARIFI and J. SWITHENBANK Sheffield University Waste Incineration Centre (SUWIC), Department of Chemical and Process Engineering, Sheffield University, Sheffield, UK Biomass fuels are being increasingly used for domestic heating and power generation to cut down the net CO 2 emission to the atmosphere. In most cases, those fuels are thermally-thick under packed-bed combustion conditions. In this paper, a doublemesh numerical scheme is proposed and implemented to simulate the detailed combustion processes for a biomass fuel with sizes ranging from 5 mm to 50 mm. Bench-top experiments were also carried out to validate the theoretical simulation. Under the specific conditions of investigation, it is found that a bed packed with particles over 35 mm can develop a temperature gradient over 4008C inside the particles at the flame front, and significant overlapping of moisture evaporation, devolatilization and char burn-out is observed in the bed-height direction; CH 4 emission over the bed top is more focused on the central part of the combustion period for larger particles; CO level in the flue gases increases with decreasing particle sizes and the opposite is true for H 2 emission. The overall air to fuel stoichiometric ratio for the whole combustion period increases significantly with increasing particle sizes, from 0.57 (fuel-rich) at 5 mm to 1.2 (fuel-lean) at 35 mm, but for the constant stage during combustion, the range of ratio narrows to Keywords: combustion; themally-thick-particle; mathematical modelling. INTRODUCTION Thermally-thick particles are defined as those particles where there is a significant temperature gradient inside the particles. The traditional criterion for thermally-thick particles is based on the Biot number (Bi ¼ ad p /k, where a is the external heat transfer coefficient, k the thermal conductivity of the solid particle and d p the particle diameter). Intensified heating condition at the particle s external surface, large particle size or low thermal conductivity of the particle material all can cause the Biot number 1.0, and steep temperature gradient exists inside the particles. Biomass fuels are being increasingly used for domestic heating and power generation to cut down the net CO 2 emission to the atmosphere, especially in some European countries. In most cases, those fuels are in the form of lump-sized particles (d p ¼ mm) and burned upon a grate. The estimated Biot number for those fuel particles ranges between 2.0 and 20.0 under typical combustion conditions (Yang, 2004). Temperature gradient inside the Correspondence to: Dr Y. B. Yang, Sheffield University Waste Incineration Centre (SUWIC), Department of Chemical and Process Engineering, Sheffield University, Mappin Street, Sheffield, S1 3JD, UK. y.b.yang@sheffield.ac.uk particles therefore exists which affects the burning characteristics of a packed fuel bed. For packed bed combustion and gasification, major physical and chemical processes all depend on the local temperature level inside the solid fuels and it is important to understand the effect of temperature gradient when the fuel particles are large or thermally-thick. Conventional experimental methods only measure the overall burning characteristics of the bed (burning rate, flue gas temperature, gaseous compositions and so on) under limited conditions. Employing numerical methods, however, detailed temperature distribution, changes in the solid composition and gaseous concentrations both inside and outside the particles can be simulated, provided that proper mathematical models are assumed. This would greatly help the understanding of the combustion processes of thermally-thick particles in packed beds and improve the combustion technique for operation optimization. In the past, numerical simulations of the pyrolysis and burning processes of single particles or lumps of solid have been made. Blasi (1996) simulated the pyrolysis process of a wooden particle subjected to an assigned external radiation where uniform boundary conditions at the particle surface were assumed and one-dimensional equations in terms of particle radius solved. Similar works were done by Saastamoinen and Richard (1996) for the simulation 549

2 550 YANG et al. of simultaneous drying and pyrolysis of solid fuel particles, and Chan et al. (1985) for the modelling of physical and chemical processes during pyrolysis of a large biomass particle, among others. But for packed beds, the situation is quite different. A packed bed is formed by many particles resting on each other and supported by a grate. Normally the bed is ignited from the bed top and the flame front travels down to the bottom of the bed as the reactions go on. For numerical simulation, this presents quite a challenge. First, the number of particles in the bed can be well over thousands and it is not feasible to calculate each individual particle by the methods mentioned above, in view of the current computer capacity and speed. Secondly, the boundary conditions at the particle s external surface are non-uniform and highly three-dimensional, so is the temperature distribution inside the particle. This adds to the difficulty of a numerical solution for a packed bed by requiring significantly more computer resources and power. Thirdly, fuel particles normally take irregular shapes and can rotate, switch positions with neighbouring ones due to grate-movement and mechanical disturbance. All these factors make the mathematical modelling and numerical simulation of packed-bed combustion a complicated task. Peters (2003) summarized the previous mathematical models on packed bed combustion. Those models can be generally classified into four categories: continuousmedium models (Behrendt, 1992; Krüll et al., 1998; Shin and Choi, 2000) where the solid bed was treated as a continuous medium; neighbouring-layers models (Goh et al., 1998; Adams, 1980) where the packed bed above the grate was divided into four layers representing fuel, drying, pyrolysis and ash; well-stirred reactor models (Stapf et al., 1997; Beckmann and Scholz, 1995) where the bed was simulated by a cascade of well-stirred reactors; and the 1d þ 1d model (Wurzenberger, 2001) where a onedimensional and transient single-particle model in spherical coordinates was implemented in a transient onedimensional fuel-bed model. Except the 1d þ 1d model, all the above models were based on the assumption of thermally-thin particles in one way or another. Yang et al. (2002, 2003, 2004) did a systematic study of packed bed combustion employing the continuous medium method and the temperature gradient inside the particles was partly dealt with by using computation cells much smaller than the particle diameter. The obtained prediction results were consistent with the experimental data to a large extent. The application of the method is limited to thermally-thin or slightly thermally-thick particles (Biot, 2.0), however, and the particles sizes were no larger than 20 mm (biomass fuels). For larger particles or the Biot number of the particles.2.0, new solving techniques have to be investigated. The 1d þ 1d model of Wurzenberger (2001) provides a good starting point for thermally-thick particle simulation. But the one-dimensional calculation for the single particles in spherical coordinate is not applicable to an actual packed-bed, due to the highly 3-D nature of the boundary conditions at the particle surfaces as mentioned above. Based on the experiences of the authors previous modelling work (Yang et al., 2002, 2003, 2004), a double-mesh computation scheme for thermally-thick particles burning in a packed bed is proposed and implemented in this paper. The maximum particle size simulated is 50 mm although a maximum particle size of 35 mm was used in the experimental validation due to equipment limitation. The work in this paper provides a tool for the modellers to simulate the combustion and gasification of large biomass particles in packed-bed form and helps to understand the detailed burning characteristic of thermally-thick biomass fuels, and optimizing the operating conditions for practical furnace applications. MATHEMATICAL MODEL FOR THE COMBUSTION OF THERMALLY-THICK BIOMASS PARTICLES IN A PACKED BED Gas-Phase The double-mesh scheme separates the gaseous phase from the solid phase in the packed bed. For the gaseous phase, the continuous medium model through a porous entity is used (Yang et al., 2002). The transport equations are as follows: g ) þr(fr g (V g V B )) ¼ S sg (1) where f represents the void fraction of the local bed; V g is the gas velocity and V B the velocity of a moving boundary. The source term S sg is the conversion rate from solid to gas due to moisture evaporation, devolatilization and char combustion. g V g ) þr(fr g (V g V B )V g ) ¼ rp g þ F(v) (2) where F(v) represents resistance of solids to fluid flow in a porous medium (Ergon equation), F(v g ) 8 m k V g >< ¼ >: Species transport 150m(1 f)2 if Re, 10 (Darcy) V dp 2 g 1:75r g(1 f) f3 d p f 3 if Re 10 (Forchheimer) V g jv g j g Y ig ) þr(fr g (V g V B )Y ig ) ¼r(D ig r(r g Y ig ) þ S yig (4) where Y ig is mass fractions of individual species (e.g., H 2, H 2 O, CO, CO 2, C m H n,...). The source term S yig accounts for mass sources of the individual species during evaporation, devolatilization and the combustion of volatile gases and char.

3 BURNING CHARACTERISTICS OF THERMALLY-THICK BIOMASS FUELS 551 g H g ) þr(fr g (V g V B )H g ) ¼r(l g rt g ) þ S a h 0 s (T so T g ) þ Q h (5) where H g represents gas enthalpy, l g the thermal dispersion coefficient and Q h, the heat gain of the gas phase due to heat release during combustion. Reactions in the gaseous phase include the burning of volatile gases: CH 4 þ 0:5O 2 ¼ CO þ 2H 2 CO þ 0:5O 2 ¼ CO 2 H 2 þ 0:5O 2 ¼ H 2 O and water-shift or steam reforming: CO þ H 2 O ¼ CO 2 þ H 2 CH 4 þ H 2 O ¼ CO þ H 2 (R1) (R2) (R3) (R4) (R5) under gasification conditions. The combustion of the primary volatile gases is controlled by both kinetics and mixing with oxygen in the under-grate air. The rate of the individual reactions can be referred to the authors earlier work (Yang et al., 2002, 2003). The steam reforming reactions are mostly used in fuel cells and achieved by catalytic medium. In the current simulation, reactions (R4) and (R5) are excluded from the calculation. Solid Phase The solid phase uses a different set of mesh from the gaseous phase. The assumed solid particles are in the shape of cubes (as seen in Figure 1). Volume shrinkage of biomass particles during combustion is calculated by the following equation: V V 0 ¼ 1 a 1 (M 0 M) a 2 (VM 0 VM) a 3 (C 0 C) (6) where V and V 0 represent the current and initial volumes of a particle and a 1, a 2, a 3 are shrinkage factors during moisture evaporation, volatile release and char burnout, respectively. Their values are chosen such that a 1 ¼ 1 represents the volume of the particle shrinking linearly with moisture loss and a 1 ¼ 0 means no volume shrinkage occurring during evaporation. The same principle applies to a 2 and a 3. The bed is assumed to be stationary and a set of particles is piled up to form a slice of the bed. The number of the particles depends on the bed height. The particles are discretized into small cells which are divided into boundary cells and inner cells. Boundary cells are located at the external surface of each of the particles and the energy conservation for the boundary cells is described s H s ) ¼r(r s rt s ) S a h 0 s (T so T g ) þrq r þ Q sh where H s presents the solid-phase enthalpy, l s is the effective thermal conductivity of the solid bed and q r the radiative heat flux. The source term Q sh summarizes heat effects due to moisture evaporation and heterogeneous combustion. For inner cells the above equation is reduced s H s ) ¼r(l s rt s ) þ Q sh (7) (7a) by dropping out the radiation and convection terms. The effective thermal conductivity of the solid material consists of a conductive and a radiative contribution (Blasi, 1996): l s ¼ 1 p l 0 g þ hl wood þ (1 h)l C þ l rad (8) where 1 p is the local porosity of the solid cell and l wood, l C are thermal conductivity of the wood and char. The radiative part is calculated as following (Hottel and Sarofim, 1967) l rad ¼ 4 1 rad s b T 3 s d pore (9) where 1 rad is the emissivity of the pores inside a particle and d pore the diameter of the pores. The solid fuel is assumed to consist of four components: moisture, volatile matter, fixed carbon and ash. The mass conservation equation for those components s Y is ¼ Sy is (10) Figure 1. The idealized model for numerical simulation of the combustion of thermally-thick particles. where Y is represents mass fractions of particle compositions (moisture, volatile, fixed carbon and ash) and Sy is the source term. Sy is accounts for the loss of the individual

4 552 YANG et al. components (moisture, volatile, fixed carbon and ash) during evaporation, devolatilization and char combustion. The incineration process of solid wastes can be divided into four successive sub-processes: evaporation of moisture from the solids, volatile release/char formation, burning of the hydrocarbon volatiles in the gaseous space, and the combustion of char particles. Calculation of the rates for moisture evaporation, devolatilization and char burning can be found in the works of Yang et al. (2002, 2003). The composition of the volatile gases released during devolatilization process is assumed to be 36% CH 4, 34% CO, 9.5% H 2, 13% CO 2 and 7.5% H 2 O by volume, based on both elemental and heat balances so that the total masses of C, H and O in the product molecules equal to those from the ultimate analysis and the total heat from the burning of the combustible gases equals to the calorific value of the volatiles, which can be deduced from the LCV of the fuel. The fuel devolatilization rate parameters are those of Alves and Figueiredo (1988) where k v ¼ exp(9977/t s ) and the kinetic rate of char combustion, k r ¼ 2.3 T s exp( /T s ) (Smoot and Pratt, 1979). The effect of the kinetic rate of fuel devolatilization has been investigated in the authors previous work (Yang et al., 2003) and the parameters selected in the work are believed to best describe the fuels fired here. Radiation Heat Transfer in the Bed Radiation is the major mechanism of heat transfer between solid particles in a packed bed and provides one of the source terms for the boundary cells on the right hand of equation (7). The widely used flux model (Smoot and Pratt, 1979) is employed: dixi þ ¼ (k a þ k s )Ixi þ dx þ 1 i 2N k ae b þ 1 2N k X N s i¼1 di xi ¼ (k a þ k s )Ixi dx þ 1 i 2N k ae b þ 1 2N k X N s i¼1 (I þ xi I xi ) (11a) (I þ xi þ I xi ) (11b) þ where N is the total number of space discretization and I xi and I 2 xi (i ¼ 1, N) are the radiation fluxes in the positive and negative directions respectively of each discretization. E b represents the black-body radiation. k a and k s denote the absorption and scattering coefficients. In this work N is set to 1 and k s assumed to be zero while k a is approximated by the following equation (Shin and Choi, 2000): k a ¼ 1 d p ln(f) (12) SOLVING THE EQUATIONS AND MODEL PARAMETERS To simplify the calculation, the fuel cubes are assumed to be arranged in a specific way in the bed as illustrated in Figure 1. The computation domain includes only a slice of the bed because of the periodic conditions across the bed and a single column of particles including neighbouring gaseous space was considered. The solving technique is based on the SIMPLE algorithm (Patankar, 1980) for the governing equations of both gas and solid phases and an existing computer code FLIC was adapted for simulation. The whole computation domain is discretised into around 300 sections along the bed height and 20 sections across. Time-dependant numerical solution is sought for a set of parameters, including gas and solid temperatures, concentration of gaseous species (CH 4, CO, H 2,O 2,CO 2,H 2 O and N 2 ), gas velocity and the four components of the solid phase (water, volatile matter, fixed carbon and ash). Details of the solving technique and general boundary conditions can be found in Yang et al. s earlier work (2002). Particle shrinkage factors a 1, a 2 and a 3 in equation (6) are taken as being 0.8. Other fuel properties employed in the model are: the porosity of the wood particle 1 p ¼ 0.6; wood conductivity l wood ¼ 0.2 Wm 21 K 21 ; char conductivity l C ¼ 0.1 Wm 21 K 21 ; particle pore size d pore ¼ m; and pore cell emissivity 1 rad ¼ 0.9. These data are based on work by Peters (2003). Sensitivity analysis has shown that variation of the particle thermal conductivity, l s from 0.1 to 0.3 Wm 21 K 21 only causes no more than 2% change in the average burning rate of the bed. The fuel is assumed to be ignited by over-board radiation at a temperature of 1173 K with emissivity of 0.8. This radiation source is present for the whole combustion processes. Primary air at 208C enters the bed from under the grate at a rate of 0.1 kg m 22 s 21 unless stated otherwise. Initial bed height is taken at 410 mm. The combustion starts at the bed top and the bed height falls as the flame front travels down towards the grate. The time-dependant solution can be converted to solution along both the bed height and length if a moving bed is employed and travelling speed assumed. EXPERIMENTS A fixed-bed reactor was employed to burn the biomass fuel. This reactor has been used in previous studies (Yang et al., 2002, 2003, 2004). It was a vertical cylindrical combustion chamber suspended from a weighing scale. The height of the chamber was 1.5 m with an inner diameter of 200 mm. It consisted of an interior tube surrounded by a thick layer of insulating material and an external casing. The grate was located at the bottom of the chamber and consisted of a perforated plate made from stainless steel. Thermocouples were used to monitor the temperature inside the bed at different height levels. There was a gassampling probe inside the chamber at 430 mm above the grate. The main components of the gas measurements of interest were O 2, CO, CO 2. A gas burner was placed at a 458 angle toward the wood chip sample at 750 mm above the grate. The gas burner was used to initiate the burning process of the fuel sample and switched off after a steady combustion was observed. Primary air was fed from the bottom of the fixed bed reactor through the grate without preheating. The fuel was pine wood chips cut into four different sizes: 5 mm 5mm 5 mm, 10 mm 10 mm 10 mm, 20 mm 20 mm 20 mm and 35 mm 35 mm 35 mm. Proximate and ultimate analysis of the fuel are shown in Table 1. Density of the fuel material was

5 BURNING CHARACTERISTICS OF THERMALLY-THICK BIOMASS FUELS 553 Table 1. Proximate and ultimate analysis of willowwood fuel used in the experiments. Moisture Volatile Fixed carbon Ash C H O LCV (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (wt%) (MJ kg) measured being 820 kg m 23 and the bulk density in packed state is around 290 kg m 23, giving a bed porosity of For each run 3.8 kg of fuel was used and the initial bed height was around 410 mm. The primary air employed was 0.1 kg m 22 s 21 at room temperature for all the cases. Figure 2. Solid temperature and composition profiles inside packed beds of three different particle sizes.

6 554 YANG et al. The weight of the total rig was recorded automatically by electronic weighing equipment, from which the mass loss rate of the burning bed was drawn. RESULTS Figure 2 presents the solid temperature and composition profiles inside packed beds of three different particle sizes at a certain time instant during their steady combustion. The graph shows only a slice of the bed (a width of twoparticle-diameters) along the overall bed width, due to the periodic character of the burning bed (see Figure 1), but reveals burning details for the whole bed-height from bottom to top. A normalized dimensionless variable is defined to facilitate the illustration: U ¼ f f min f max f min (13) where f represents solid temperature T s, moisture fraction in the solids M, remaining volatile matter VM and fixedcarbon C in the ash, respectively. For each profile plot in Figure 2, there are two sub-plots which are attached with a total width of two particle diameters (2d p ). They are exactly the same results but the left sub-plot is a continuous plot and the right is a contour plot. The two types of illustration are used for clearer view and easy understanding of the results. The time instances chosen are 1000 s, 1400 s and 1700 s for the initial particle sizes of 10 mm, 35 mm and 50 mm, respectively. The criteria for these selections are such that they are all within the steady-state combustion stage and comparable to each other in terms of the bed height. The steady-state combustion stage is the period after the bed ignition and before the final char burnout where the burning rate, gas temperature and composition of the flue gases are more or less constant (Yang et al., 2003, 2004). For the initial particle size of 10 mm, the simulated result shows that the profile of the solid temperature in the bed is nearly one dimensional against the bed height but quite uniform in the horizontal or width direction. At the same bed height, variation in the width direction of the solid temperature is within 10% of its maximum difference (1000 K roughly) from the centre of a particle to its surface. The unreleased volatile matter and fixed-carbon are also relatively uniformly distributed across the bed width. The moisture profile is less uniform in the width direction, however and the variation of water content in the solid can reach 50% of the maximum difference from the particle center to its external surface at the flame front. For the initial particle size of 30 mm, Figure 2 shows that the cross-width temperature variation from particle centre to surface at the flame front reaches 40% of the maximum difference and roughly a temperature gradient of 400 K exists inside an individual particle. This produces the specific pattern of the moisture content profile across the bed where the areas near the particle surface are dried at elevated temperatures while regions near the particle centres are still filled with the original water at room temperature. For the volatile matter and fixed-carbon in the solids, the width-direction-variances of their respective levels are comparable to the respective height-direction-variances and the burning processes in the solid phase at the flame front demonstrate a multidimensional feature. As the initial particle size further increases to 50 mm, the temperature and solid composition profiles in the bed become more multi-dimensional and the flame front is prolonged in the bed height direction. The combustion processes are more individual-particle-based and steep gradients exist inside the particles in terms of temperature, moisture, residual volatile and formed char. Figure 3 demonstrates the individual process rate distribution along the bed height for the three cases of different particle sizes. The respective time instances are the same as those in Figure 2. For the initial fuel size of 10 mm [Figure 3(a)], the zone depths of the moisture evaporation, devolatilization and char burning are around 35 mm, 15 mm and 10 mm, respectively. The overlapping depth is 5 mm for moisture evaporation and devolatilization and Figure 3. Process rate profiles along bed height for three different particle sizes: (a) 10 mm; (b) 35 mm; (c) 50 mm.

7 BURNING CHARACTERISTICS OF THERMALLY-THICK BIOMASS FUELS mm for devolatilization and char burn-out. The total reaction depth excluding the moisture evaporation is around 21 mm and the peak process rates for moisture evaporation, devolatilization and char burning are 0.5, 6 and 3 kg m 23 s 21, respectively. For the initial particle size of 35 mm, Figure 3(b) shows the zone depths of the moisture evaporation, devolatilization and char burning increase to around 100 mm, 55 mm and 20 mm, respectively. The overlapping depth rises to 40 mm for moisture evaporation and devolatilization and 13 mm for devolatilization and char burn-out. The total reaction depth excluding the moisture evaporation is 70 mm, much thicker than that for the 10 mm particles and the peak process rates for moisture evaporation, devolatilization and char burning are reduced to 0.12, 1.4 and 0.93 kg m 23 s 21, respectively. For the even bigger particles of 50 mm [Figure 3(c)], the devolatilization zone is further extended to 100 mm and the char-burning zone to 30 mm in depth. The total reaction zone thickness excluding the moisture evaporation is around 120 mm. Figure 4 illustrates CH 4 profiles both inside the bed and in the over-bed area immediately adjacent to the bed top for three different particle sizes. White line indicates the bed top boundary. For the 10 mm particles, a high CH 4 period in the bed-top space follows the initial unsteady ignition-stage. The over-bed CH 4 concentration then gradually falls to a stable level until all the volatile matter in the solids is released as combustion approaches the finishing point. The maximum CH 4 level is 1.9% by mass. For the initial particle size of 35 mm, a series of spikes in the over-bed CH 4 concentration are observed through most of the combustion period and the maximum over-bed CH 4 level is 1.7% by mass. For the initial particle size of 50 mm, a very different over-bed CH 4 profile is observed. Instead of a more or less uniform distribution of CH 4 over the time range, it is found that high CH 4 emission is more concentrated on the middle part of the combustion processes, with the maximum level reaching 2.0% by mass. At the both ends of the combustion processes, CH 4 level is dramatically reduced. Figure 5 demonstrates the comparison between the model (FLIC) prediction and measurements of the gas temperature profiles vs. reaction time for particle size 35 mm. The maximum temperature is correctly predicted, which is around 1370 K or (11008C). But there are two areas where discrepancy exists. The first one is the predicted temperature hot area around t ¼ 1300 s when the bed has just been ignited. Measurements did not show such a hot area. This is perhaps because of the relatively cooler reactor wall at the initial stage of combustion and the bare thermocouples could give a reading more than 1008C lower than the actual level. The second discrepancy is in the late stage of combustion where measurement shows a longer period of higher temperature while the model (FLIC) prediction demonstrates a shorter period of higher temperature. The last stage of combustion mainly comprises the char burnout and the kinetic data for biomass char combustion have not been well established. The used char-burning data in the model may not consistent with the actual situation. Figure 6 demonstrates the burning rate vs. reaction time for different particle sizes and comparison between model (FLIC) predictions and experiments. The whole combustion period can be divided into three stages: ignition stage, steady combustion stage and final burn-out stage. Model predictions show that the smallest particle size (5 mm) produces the quickest bed ignition by having the shortest ignition period while the biggest particles (35 mm) result in the longest ignition period before the steady-state-burning is reached. The ignition periods for the four particle sizes investigated are 210 s, 330 s, 660 s and 1080 s, respectively. For the steady-state burning, the 10 mm particle size gives the highest mass loss rate around kg m 22 s 21, followed by particles sizes of 5 mm, 20 mm and 35 mm. The 35 mm particles give the lowest mass loss rate around 0.05 kg m 22 s 21, which is 25% lower than the maximum. For the last burn-out stage, smaller particle groups of 5 mm and 10 mm show a temporary surge and then a sharp fall in the instant mass loss rate before the combustion is completed. The bigger particle groups of 20 mm and 35 mm show, however, a steady and gradual fall in the instant mass loss rate. Comparison of the model predictions to the measurements is reasonably satisfactory. Figure 7 shows both the model predictions and the experimental results of CO concentration in the out-ofbed flue gases as a function of reaction time. Three general trends have been observed: first, there is a substantial part of the time period where the CO concentration in the flue gases keeps relatively constant; second, this stable level of CO increases with decreasing particle size, from 12% at 35 mm to 18% at 5 mm and third, for smaller particles, there is a substantial increase in the flue-gas CO (around 30% over the respective stable level) in the late stage of combustion before the combustion is completed. This peak of CO can reach 25% for 5 mm particles and 21% for 20 mm particles. There is, however, no obvious CO peak observed in the late stage for the larger particles of 35 mm. Instead, a modest CO peak was observed in the early stage of combustion just (after the ignition) during experiments. Figure 7 demonstrates that the general CO trends predicted by the mathematical model (FLIC) are in agreement with experimental observations. There are regions where discrepancy exists, however. First, the measurements show a more constant CO profile for most part of the combustion period than do the model predictions. Second, the experimentally-observed CO peak in the late stage is not well predicted for particle size group of 10 mm and 20 mm. There discrepancies need further investigation. Figure 8 show the predicted H 2 concentration in the outof-bed flue gases as a function of reaction time. It is noted that H 2 concentration increases sharply after the bed has been ignited and then falls to a stable level. In the late stage of combustion, H 2 level increases again before combustion is completed. Those peaks are not obvious for larger particles, however. It is also noted that contrary to the situation of CO level in relation to the particle size, the H 2 level during the stable stage increases with increasing particle size, from 2% at 5 mm to 7% at 35 mm. Figure 9 show the instant air to fuel stoichiometric ratio of the bed as a function of reaction time for the beds of

8 556 YANG et al. Figure 6. Burning rate vs. reaction time for different particle sizes: (a) FLIC prediction; (b) experimental data. Figure 4. Calculated CH 4 profiles (in vol%) both inside the bed and in the over-bed area immediately adjacent to the bed top for three different particle sizes. White line indicates the bed top boundary. Primary air flow rate: 0.19 kg m 22 s 21. different particle sizes. It is seen that during the initial ignition stage, combustion is mainly fuel-lean and the air to fuel stoichiometric ratio greater than 1.0. In the following stable combustion stage, the combustion shifts to fuel-rich conditions and the air to fuel equivalent ratio drops to During the late stage of combustion, Figure 5. Gas temperature profiles vs. reaction time for particle size 35 mm: (a) FLIC prediction; (b) experimental data. Figure 7. CO concentration in the out-of-bed flue gases vs. reaction time for different particle sizes: (a) FLIC prediction; (b) experimental data.

9 BURNING CHARACTERISTICS OF THERMALLY-THICK BIOMASS FUELS 557 Figure 8. H 2 concentration in the out-of-bed flue gases vs. reaction time for different particle sizes. Figure 9. Air to fuel stoichiometric ratio vs. reaction time for different particle sizes: (a) FLIC prediction; (b) experimental data. the air to fuel ratio increases to over 1.0 again and the combustion becomes fuel-lean as combustion approaches the finishing point. In terms of the particle size effect, model predictions [Figure 9(a)] shows that the stable-stage air to fuel equivalent ratio is slightly proportional to the fuel size, from 0.29 at 5 mm to 0.35 at 35 mm. But experiments [Figure 9(b)] show that the stable-stage air to fuel ratio remained relatively the same (around 0.32) for the three particle groups of 5 mm, 10 mm and 20 mm, but increased slightly to 0.35 for the largest particles of 35 mm. The overall combustion stoichiometry for the whole combustion process, however, is significantly affected by the particle sizes and the obtained values are 0.57, 0.60, 0.80 and 1.2 for particle sizes of 5 mm, 10 mm, 20 mm and 35 mm, respectively. DISCUSSION Particle size affects the combustion processes in the bed on four different aspects. First is the two-phase heat and mass transfer between the gas and solid. Particles of a smaller size (hence large surface area) can enhance the moisture evaporation and char burning rates in the bed and result in higher burning rate as illustrated in Figure 6. The second effect is on the radiation heat transfer where the bed absorption coefficient to radiation flux is inversely proportional to particle diameter [see equation (12)]. This means that a bed of smaller particles absorbs radiation more quickly. But the radiation flux, on the other hand, is less absorbed by a packing of larger particles and hence penetrates a longer distance in the bed. This can produce a thicker flame front in the bed and affects the temperature and gas concentration profiles as a consequence. This explains the results shown in Figure 3 where the reaction zone thickness for 35 mm particles is several times thicker than that for 10 mm particles. The third effect is on the turbulent dispersion of energy and gas species in the packed bed. Both thermal and fluid dispersion coefficients are proportional to particle diameter. Larger particles produce larger-scale turbulence in the local bed structure and facilitate the cross-flow and inflow mixing in the gas phase. This has the same effect as the radiation penetration on the reaction zone thickness. The fourth effect is on the burning rate of volatile gaseous fuel both in the voids of the bed and in the immediate area above the bed. The process is basically diffusion-controlled. The mixing rate of the fuel gases with air flow from under the grate is a function of particle size (Yang et al., 2002, 2003) and small particles produce higher rate the other conditions being the same and hence intensity the combustion in the gas phase. The obtained results for CO, H 2 and CH 4 concentrations in the out of bed flue gases are just a balance between productions of these gases (through volatile release and intermediate reactions) and their consumption by oxygen in the primary air. For CO, smaller particles result in higher volatile release rate as well as higher char burning rate and both these two processes generate CO. On the other hand, smaller particles also favour good mixing of the produced CO with oxygen from the primary air and increase the consumption of CO. The balanced results favour higher CO in the flue gases. Similar principal and explanation are applied to CH 4 (Figure 4) where smaller particles also favour higher concentration at the bed top. However, the predicted H 2 concentration has the opposite trend to CO and CH 4 and larger particles favour higher H 2 concentration at the bed top (Figure 8). This shows that the decrease in H 2 consumption by oxygen outpaces the decrease in H 2 production as particle size increases. SUMMARY Combustion of thermally-thick biomass fuel particles in a packed bed has been investigated by both numerical method and experimental measurements under constant primary air flow rate. The major conclusions are:. thermally-thick biomass fuels over 35 mm can develop a temperature gradient over 4008C inside the particles at the flame front under ordinary combustion conditions;

10 558 YANG et al.. the total reaction zone thickness excluding the moisture evaporation increases with increasing particle size and is about two times the initial particle diameter;. significant overlapping of moisture evaporation, devolatilization and char burn-out in the bed-height direction is predicted;. CH 4 emission over the bed top is more focused on the central part of the combustion process for larger particles;. CO emission in the flue gases increases with decreasing particle sizes while H 2 emission behaviours in the opposite way;. under the specific conditions of investigation, the overall air to fuel stoichiometric ratio for the whole combustion period increases significantly with increasing particle sizes, from 0.57 (fuel-rich) at 5 mm to 1.2 (fuel-lean) at 35 mm; but for the constant stage during combustion, the range of ratio narrows to NOMENCLATURE C constant; molar fractions of species (fuel, oxygen, fixed carbon) C fixed carbon fraction C 0 fixed carbon fraction in fresh fuel D ig dispersion coefficients of the species Y ig,m 2 s 21 d p particle diameter, m d pore pore diameter in particle, m I radiation heat flux, W m 22 H g gas enthalpy, J kg 21 H s solid-phase enthalpy, J kg 0 h s convective heat transfer coefficient between solid and gas, W m 22 K 21 k a absorption coefficient of radiation, m 21 k s scattering coefficient of radiation, m 21 k v rate constant of devolatilization, s 21 M moisture fraction in solid M 0 initial moisture fraction in solid p g gas pressure, Pa Q h heat loss/gain of the gases, W m 23 Q sh thermal source term for solid phase, W m 23 q r radiative heat flux, W m 22 S a particle surface area, m 2 S sg conversion rate from solid to gases due to evaporation, devolatilization and char burning, kg m 23 s 21 Sy ig mass sources due to evaporation, devolatilization and combustion, kg m 23 s 21 Sy is source term, kg m 23 s 21 t time instant, s T g gas temperature, K T s solid temperature, K T s0 solid temperature at the particle surface, K V B velocity of a moving boundary, m s 21 V g superficial gas velocity (vector), m s 21 VM volatile matter in solid VM 0 initial volatile matter in solid Y ig mass fractions of gaseous compositions (CO, CO 2, O 2,CH 4,H 2 and so on) Y is mass fractions of particle compositions (moisture, volatile, fixed carbon and ash) Greek letters 1 p solid emissivity 1 p porosity inside a particle 1 rad emissivity of pores inside a particle k permissivity of the packed bed l c thermal conductivity of char, W m 21 K 21 l g thermal dispersion coefficient, W m 21 K 21 l s solid-phase thermal conductivity in bed, W m 21 K 21 l wood effective thermal conductivity of the wood material, W m 21 K 21 m gas viscosity, kg m s 21 ; ratio of the current to the initial biomass mass r g gas density, kg m 23 r s solid material density in the bed, kg m 23 s b Boltzmann radiation constant, Wm 22 K 24 f void fraction in the bed REFERENCES Adams, T., 1980, A simple fuel bed model for predicting particulate emissions from a wood waste boiler, Combustion and Flame, 39: Alves, S. S. and Figueiredo, J.L., 1988, Pyrolysis kinetics of lignocellulosic materials by multistage isothermal thermo-gravimetry, J Analyt Appl Pyrolysis, 13: 123. Beckmann, M. and Scholz, R., 1995, Simplified mathematical model of combustion in stoker systems. The Third European Conference on Industrial Furnaces and Boilers, Porto, Portugal, Behrendt, Th., 1992, Thermodynamische Modellierung des Betriebsverhaltens einer Hausmüllverbrennungsanlage am Beispiel TAMARA. Fortschrittberichte Reihe 15: Umwelttechnik (VDI Verlag, Düsseldorf). Blasi, C.D., 1996, Heat, momentum and mass transport through a shrinking biomass particle exposed to thermal radiation, Chemical Engineering Science, 51 (7): Chan Wai-Chun, R., Kelbon, M. and Krieger, B.B., 1985, Modelling and experimental verification of physical and chemical processes during pyrolysis of a large biomass particle, Fuel, 64: Goh, Y.R., Siddall, R.G., Nasserzadeh, V., Zakaria, R., Swithenbank, J., Lawrence, D., Garrod, N. and Jones, B., 1998, Mathematical modelling of the waste incinerator burning bed, Journal of the Inst of Energy, 71: Hottel, H.C. and Sarofim, A.F., 1967, Radiative Transfer (McGraw-Hill Book Company, New York). Krüll, F., Kremer, H. and Wirtz, S., 1998, Feuerraumsimulation einer Müllverbrennungsanlage bei gleichzeitiger Simulation der Verbrennung auf dem Rost. VDI Bericht 1390, Modellierung und Simulation von Dampferzeugern und Feuerungen, Patankar, S.V., 1980, Numerical Heat Transfer and Fluid Flow, (Hemisphere, London, Washington). Peters, B., 2003, Thermal Conversion of Solid Fuels (WIT Press, Southampton, Boston). Saastamoinen, J. and Richard, J.-R., 1996, Simultaneous drying and pyrolysis of solid fuel particles, Combustion and Flame, 106: Shin, D. and Choi, S., 2000, The combustion of simulated waste particles in a bed, Combustion and Flame, 121: Smoot, L.D. and Pratt, D.T., 1979, Pulverized-Coal Combustion and Gasification (Plenum Press, New York, USA). Stapf, M., Kircherer, A., Kolb, Th., Wölfert, A. and Seifert, H., 1997, Verbrennung im Drehrohrofen: Modellierung, Betriebs und Technikumsversuche; Volume 15 of VDI-Berichte 1313 (VDI Verlag, Düsseldorf). Wurzenberger, J.C., 2001, A combined packed bed and single particle model applied to biomass combustion, PhD thesis, Graz University of Technology. Yang, Y.B., 2004, Simulation of thermally-thick particles and latest development of packed-bed modelling. EPSRC (UK) Supergen Progress Report, Loughborough, November 3 5. Yang, Y.B., Goh, Y.R., Zakaria, R., Nasserzadeh, V. and Swithenbank, J., 2002, Mathematical modelling of MSW incineration on a travelling bed, Waste Management, 22: Yang, Y.B., Yamauchi, H., Nasserzadeh, V. and Swithenbank, J., 2003, Effects of fuel devolatilisation on the combustion of wood chips and incineration of simulated municipal solid wastes in a packed bed, Fuel, 82: Yang, Y.B., Sharifi, V.N. and Swithenbank, J., 2004, Effect of air flow rate and fuel moisture on the burning behaviours of biomass and simulated municipal solid wastes in packed beds, Fuel, 83: ACKNOWLEDGEMENT This study was funded by UK Engineering, Physics and Science Research Council (EPSRC) SuperGen Biomass Grant. The manuscript was received 13 October 2004 and accepted after revision 14 February 2005.