Simulation of coal combustion by AUSM turbulence-chemistry char combustion model and a full two-fluid model

Transcription

1 Fuel 84 (25) Simulation of coal comustion y USM turulence-chemistry char comustion model and a full two-fluid model Yu Zhang a, *, Xiao-Lin Wei a, Li-Xing Zhou, Hong-Zhi Sheng a a Institute of Mechanics, Chinese cademy of Sciences, eijing 18, China The State Key Laoratory of Clean Coal Comustion, Tsinghua University, eijing 184, China Received 28 July 24; received in revised form 1 March 25; accepted 1 March 25 vailale online 21 pril 25 stract n algeraic unified second-order moment (USM) turulence-chemistry model of char comustion is introduced in this paper, to calculate the effect of particle temperature fluctuation on char comustion. The USM model is used to simulate gas-particle flows, in coal comustion in a pulverized coal comustor, together with a full two-fluid model for reacting gas-particle flows and coal comustion, including the su-models as the k-3-k p two-phase turulence model, the EU-rrhenius volatile and CO comustion model, and the six-flux radiation model. new method for calculating particle mass flow rate is also used in this model to correct particle outflow rate and mass flow rate for inside sections, which can oey the principle of mass conservation for the particle phase and can also speed up the iterating convergence of the computation procedure effectively. The simulation results indicate that, the USM char comustion model is more preferale to the old char comustion model, since the later totally eliminate the influence of particle temperature fluctuation on char comustion rate. q 25 Elsevier Ltd. ll rights reserved. Keywords: USM char comustion model; Particle temperature fluctuation; Particle mass outflow rate 1. Introduction CFD has ecome a very powerful tool to simulate complex chemical multiphase in many processes, such as coal comustion, et al. In turulence flow, it is very hard jo for CFD to simulate the closure chemistry reaction rate. In commercial codes such as FLUENT, a simplified PDF model is used. Due to the approximation made in adopting the product of several single-variale PDF s instead of a joint PDF, when using this model to predict NO formation, the difference etween prediction and eriment is rather large [1]. n earlier version of a second-order moment model was used to predict NO formation in coal comustion [2], and the predicted results have not yet een verified y eriment. However, this model has een used to simulate NO formation in methane air comustion [3] and the NO concentration is significantly under-predicted due to the fact * Corresponding author. Tel: C ; fax C address: yuzhang@imech.ac.cn (Y. Zhang) /$ - see front matter q 25 Elsevier Ltd. ll rights reserved. doi:1.116/j.fuel that the approximation of E/RT/1 made in the series ansion of the onential function of temperature leads to the elimination of the higher-order terms, which cannot e neglected for E/RTO5 in the case of NO formation. In order to improve the second-order moment (SOM) models, a SOM-PDF model [4] and a unified SOM (USM) model [5] were proposed. These models are used to simulate methane air comustion and NO formation and oth of them have een well verified y eriment. The results show that the SOM-PDF and USM models are much etter than the EU- rrhenius model, the simplified PDF model and the old version of the SOM model. For the coal comustion conditions, due to the complexity of the processes, an algeraic unified second-order moment (USM) turulence-chemistry model for NO formation has also een proposed [6]. s more attention is focused on the chemical reaction details of the char comustion model [7], there are still no reports aout the influence of particle turulence on char comustion rate. In this paper, The E E model is used to descrie two-phase flow, which means that particle phase is also treated as a fluid phase and the Euler method is used to

2 Y. Zhang et al. / Fuel 84 (25) descrie oth two-phase movements. Since the particle phase is treated like a continuous phase, it has own turulent kinetic energy, stress and so on. If there is two-phase chemical reaction flow, the particle phase should also have own temperature fluctuation. In this paper, a new USM turulence chemistry reaction model of char comustion is proposed, which takes the particle temperature fluctuation into account. The simulation results indicate that, the model proposed in this paper is more reasonale than the old model. In this paper a set of in-house procedures is used. In the gas phase CFD simulation, it is well known that, for the procedure to converge and produce the reasonale predicted results, the outflow rate must e corrected y the principle of mass conservation in the procedure. In this paper, as well as a real coal comustor simulation, an outflow rate correction method for the particle phase in a two-fluid model is also discussed. 2. USM turulence chemistry char comustion model 2.1. Particle temperature fluctuation Fig. 1 shows the particles at different times in the same places. The lack points mean low temperature particles without char comustion, while the white points represent the high temperature particles with char comustion. Clearly, the particle temperatures at different times in the same place are likely to change and fluctuate ecause of char comustion. It can e shown y means of time average method that: T p ðxþ Z 1 T ð tct=2 tkt=2 T p ðx; tþdt (1) T p ðx; tþ Z T p ðxþ CT pðx; tþ (2) where T is the time period, T p is the instantaneous particle temperature, T p is the time averaged particle temperature. The instantaneous gas-particle chemical reaction rate can e ressed as follows: m ZK 1 pd2 pr S Y O2 K E (3) RT p where the,, E are the chemical reaction parameters, d p is the particle diameter, r s is the mean gas density at surface of t t + dt Fig. 1. Particles at different Times t and tcdt. particle, and Y O2 is the oxygen component of the reaction gas at the surface of particle. If the correlation of components is neglected, the time averaged multi-phase chemical reaction rate is as following: _m ZK 1 pd2 pr S Y O2 ;S 1 K E RT p It can e seen that, in turulent flow, there is still a closure prolem of the turulent multiphase chemical reaction rate which is lained in the reference [2] USM turulence chemistry char comustion model ssuming that, there are three chemical reactions on the surface of char C CO 2 /CO 2 ; 2C CO 2 /2CO; C CCO 2 /2CO (5) then _m 1 c;pzk 1 pd 2 E pr S Y O2 ;S 1 :5 K 1 c1 RðT p C1=2TpÞ E C K 1 RðT p K1=2TpÞ ; _m 2 c;pzk 1 pd 2 E pr S Y O2 ;S 2 :5 K 2 c2 RðT p C1=2TpÞ E C K 2 RðT p K1=2TpÞ ; _m 3 c;pzk 1 pd 2 E pr S Y CO2 ;S 3 :5 K 3 c3 RðT p C1=2TpÞ E C K 3 RðT p K1=2TpÞ ð6þ T P Z T p =T g T g (7) Formula (6) gives the USM model, which is proposed in this paper. TP is the particle temperature fluctuation, T p is the mean particle temperature, T g means the gas mean temperature, and Tg represents the gas temperature fluctuation. There is a simple assumption that the ratio of particle temperature to gas temperature equals the ratio of particle temperature fluctuation to gas temperature fluctuation, so what TP can e ressed y formula (7), and may avoid solving the particle temperature conservation equation. Formula (8) gives the old char comustion model, from which it can e seen that, here the effect of particle temperature fluctuation on char chemical reaction rate is (4)

3 18 Y. Zhang et al. / Fuel 84 (25) totally eliminated. _m 1 c;p ZK 1 pd 2 pr S Y O2 ;S 1 K E 1 ; c1 RT p _m 2 c;p ZK 1 pd 2 pr S Y O2 ;S 2 K E 2 ; c2 RT p _m 3 c;p ZK 1 pd 2 pr S Y CO2 ;S 3 K E 3 c3 RT p 3. The full two-fluid model for reacting gas-particle flows and coal comustion For the comprehensive modelling of reacting gasparticle flows and coal comustion, a full two-fluid model [6] is used. The continuity, momentum, energy and turulent kinetic energy equations for gas phase and particle phase are derived and solved in Eulerian coordinates. The su-models are: k 3 k p two-phase turulence model, EU-rrhenius comustion model, six-flux radiation model: and two-equation model of coal devolatilization. The detailed description of this comprehensive model can e found the Ref. [6]. The USM turulence-chemistry model for char formation is incorporated into the comprehensive model. The asic equations of 3D turulent two-phase reacting flows and coal comustion can e ressed in the following generalized form: Gas-phase equations: v vx ðru4þ C v rvr ðrrv4þ C v C v rvr rg 4 v4 vr rvq ðrw4þ Z v vx C v r 2 vq G v4 4 vq v4 G 4 vx CS 4 CS 4p Particle-phase equations: v vx ðr pu p 4 p ÞC v rvr ðrr pv p 4 p ÞC v rvq ðr pw p 4 p ÞZ v vx C v v4 p rg rvr 4p C v vr r 2 vq G v4 p 4p CS vq 4p CS 4pg G 4p v4 p vx ð9þ ð1þ where ( and ( p are the generalized independent variales, and S (, S (p and S (pg are the source items. The meanings of these variales and terms are given in the Ref. [6]. The oundary conditions for the gas phase and particle phase are specified as in usual treatment, fully developed flow conditions at the exit; symmetrical conditions at the axis; non-slip condition for the gas velocity at the wall; and the wall function approximation is used for near-wall grid nodes. The particle-phase conditions at the wall are: Zero Normal Mean Velocity (ZNMV) and zero gradients of other variales 4. Mass flow rate correction method of particle phase in two-fluid model In this paper a set of in-house procedures is used to simulate multiphase flow ased on the two-fluid model. In the pure gas flow simulation, it is will known that the gas mass flow rate must e corrected y the principle of mass conservation: what is the asic requirement of the SIMPLE method [9]. In the two-phase flow simulation, the gas phase mass flow rate still needs to e corrected, which is the same as that for pure gas flow simulation, and can e ressed as follows: UðLP1; J; KÞ Z UðL; J; KÞSUMGSIN=MSSGðLÞ (11) U is the main flow velocity, SUMGSIN is the inlet mass flow, LP1 means outflow oundary section. L represents nearest section to the outlet. MSSG is the mass flow at different sections, and can e ressed as follows: MSSGðIÞ Z :5ðRHOðI; J; KÞ CRHOðI K1; J; KÞÞ UðI; J; KÞ ð12þ In formula (12), RHO means the density of the gas. is the sectional area of the comustor. The outflow rate correction is the asic requirement of the SIMPLE method. Since mass flow at every section is corrected, iteration can converge quickly, and it is often used in real simulation work. In the multiphase flow simulation, if the two-fluid model used, the particle is treated as a continuous phase. The question here arisen as to whatever the particle mass flow rate needed to e corrected or not? What is the most reasonale method? If use gas mass flow rate correcting method is used for particle phase, formulae (11) and (12) can e recasted as follows: UPðLP1; J; KÞ Z UPðL; J; KÞSUMGSINP=MSSPðLÞ (13) MSSPðIÞ Z :5ðRHOPðI; J; KÞ CRHOPðI K1; J; KÞÞ UPðI; J; KÞ ð14þ UP means main flow velocity of particle phase, MSSP is particle mass flow rate at different sections, RHOP is particle concentration. The method, the so called Method 1, is descried y formulae (12) and (14) to calculate mass flow rate. Unfortunately, if the aove method is used to correct the particle mass flow rate at different sections, generally speaking, the iteration will not converge. So, many researchers considered that particle mass flow rate cannot e corrected. Figs. 2 and 3 show the section gas and particle mass flow rates along the reactor height without any chemical reactions (Details of erimental work are introduced in

4 Y. Zhang et al. / Fuel 84 (25) mass flow ratekg/s the next section). Fig. 2 indicates that, along the reactor height the gas flow rate is not changed, which is considered. For particles, the mass flow rate calculated y method 1 increases at first and then decreases along the reactor height (Fig. 3). pparently, this reaks the mass conservation rule. If this curve is used to correct particle mass flow rate, the iteration will not converge. Recast formula (14) is MSSPðIÞ Z MSSPðIÞ CUPðI; J; KÞ RHOPðI K1; J; KÞ jk when Fig. 2. Gas mass flow rate along the reactor height. method 1 ðupði; J; KÞOÞ MSSPðIÞZMSSPðIÞCUPðI;J;KÞRHOPðI;J;KÞ jk ðupði;j; KÞÞ ð15þ concentration continuum equation essentially differs form the gas continuum equation. The gas continuum equation is converted to a pressure correction equation, different from other conservation equations characterized y a central difference scheme. The particle concentration equation is a normal conservation equation, and is characterized y an up-wind difference scheme. For this reason, the methods to calculate section mass flow rate of gas and particles are different, and using method 1 to calculate particle mass flow rate will lead to the wrong results. It worth mentioning that, the particle phase mass flow rate correction has a different meaning from the gas phase mass flow rate correction. For the gas phase, the outflow rate correction is indispensale, which is an essential requirement of the SIMPLE method. n inside mass flow rates correction at different sections is just used to save calculation time. For the particle phase, oth outflow rate correction and inside mass flow rate correction are dispensale, eing used only for saving calculation time. Therefore, even if the mass flow rate correction is not used for the particle phase, the procedure run smoothly, ut converges slowly. 5. Experimental section The trial tests were conducted in an entrained flow comustion reactor (EFCR) (see Fig. 4). The electrically heated reactor has five regulated heating zones. The carrying air with pulverized coal enters the urner center, The method to calculate the mass flow rate y formula (5) called method 2, gives reasonale results. The particle mass flow rate is not great changed along the reactor height. When using calculated results to correct particle outflow rate or mass flow rate at inside sections, the iteration will converge. The main difference etween method 1 and method 2 is the way in which particle concentration is treated. The former uses the central difference scheme while the later takes the up-wind difference scheme. Indeed, the particle φ 8 φ12 φ18 φ3 φ34 coal + carrying air primary air secondary air T1 T2 urner exit (. m) mass fow rate kg/s method 1 method urner reactor tue (2 mm) to filter and chimney T3 T4 T5 reactor exit (2.5 m) to analyser Fig. 3. Particle mass flow rate along the reactor height. Fig. 4. Schematic of the coal comustor.

5 182 Y. Zhang et al. / Fuel 84 (25) Tale 1 Parameters of the asic calculation Parameters Units Values Coal mass flow kg/h 1. Wall temperature 8C 125 Volume flow of coal carrying air Nm 3 /h 1.5 Temperature of coal carrying air 8C 2 Primary aircsecondary air Nm 3 /h 8. Primary air:secondary air 1:2 Temperature of primary air 8C 25 Temperature of secondary air 8C 35 Mean diameter of particle mm 16, 52, 16, 35 % 3, 35, 25, 1 Tale 2 Coal analysis data of coal type1 Proximate analysis (wt%, raw asis) Moisture Volatile sh Fixed caron Ultimate analysis (wt%, raw asis) C H N S O Coal Coal surrounded y the primary air and the secondary air. gravimetric screw conveyor supplies a constant coalfeeding rate. The furnace (ceramic tue) has a length of 2.5 m and an internal diameter of 2 mm [8]. Tale 1 gives the parameters of the asic calculation. Tale 2 gives the two ituminous and analysis data. set of in-house CFD procedure is used to simulate coal comustion and NO formation, non-uniform staggered grid which nodes of 537 are used. Running a example on a Pentium-4-2.G personal computer takes aout 4 5 days. 6. Results and discussion Fig. 5 gives the indicated and erimental oxygen concentration result for different coals. Model 1 means the USM turulent char comustion model, while the is the old char comustion model. From Fig. 5, it can e seen that, the predicted result of is more accurate compared with, From while the result is higher than the erimental result. Fig. 6 shows the caron dioxide concentration for different coals. pparently, the predicted result of is etter than that of, which is lower than the erimental data. oth Figs. 5 and 6 indicate that, for the sake of eleminating the effect of fluctuations in the particle temperature the old model underestimates the char comustion rate, so its predicted oxygen result is higher than in, while the predicted caron dioxide result is lower. The USM model takes the particle temperature fluctuation into account most Fig. 5. The O 2 concentration along the reactor height ( coal 1 coal 2) Fig.6. TheCO 2 concentrationalong thereactor height( coal1 coal2).

6 8 Y. Zhang et al. / Fuel 84 (25) r(m) r(m) particle mass flow rate (kg/s) x (m) x (m) Fig. 7. The particle temperature map ( coal 1 coal 2) reasonale, so its predicted results are etter. From Figs. 5 and 6 can e seen that, in the inlet region, the difference etween the predicted results and erimental results is comparatively large. The reason for this might e a measurement error. Near the inlet of the reactor, it is hard to keep the analysis tue at the exact central position in the inlet and the measurement data ecome very sensitive to the position of the proe. Fig. 7 gives the particle temperature map; it can e seen that in the inlet region the particle temperature rises very rapidly. Fig. 8 shows the particle mass flow rate along the reactor height; it can e seen that the particle reaction rate predicted as USM is higher than that of old model. In other words, particle temperature fluctuation will enhance the particle comustion process. 7. Conclusions 1. The old char comustion model totally eliminates the influence of particle temperature fluctuation on char comustion rate. Its predicted oxygen concentration is higher than the erimental data while the predicted caron dioxide level is lower than erimental. 2. The USM turulence char comustion model takes the particle temperature fluctuation into account will, oth the predicted oxygen and caron dioxide are etter than the predicted results y the old char comustion model, especially in the inlet region, where the particle temperature fluctuation is very high. 3. The gas mass outflow rate correction should e considered in two-fluid model simulation. For particle phase, the outflow rate correction is dispensale and may e used only to save calculation time. It should e noted that, the mass flow rate correction methods for gas and particle are essential different cknowledgements particle mass flow rate (kg/s) Fig. 8. The coal mass flow rate along the reactor height ( coal 1 coal 2). Financial Support from the Chinese Special Funds for Major State asic Research Projects (G ) and National Natural Science Foundation of China (No ) is acknowledged. References [1] Zhou LX. dvances in numerical modeling of turulent reaction of NOx formation. dv Mech (China) 2;3: [2] Zhou LX, Guo YC, Lin WY. Two-fluid models for simulation reacting gas-particle flows, coal comustion and NO x Formation. Comust Sci Technol 2;(15): [3] Liao CM, Liu ZN, Zheng XQ, Liu CQ. NOx prediction in 3D turulent diffusion flames y using implicit multi-grid methods. Comust Sci Technol 1996;119:219 6.

7 184 Y. Zhang et al. / Fuel 84 (25) [4] Zhou LX, Chen XL, Zheng CG, Yin J. Second-order moment turulence-chemistry models for simulating NOx formation in gas comustion. Fuel 2;79: [5] Zhou LX, Chen XL, Zheng CG, Yin J. Second-order moment turulence-chemistry models for simulating NOx formation in gas comustion. Fuel 2;79: [6] Zhou LX, Zhang Y, Zhang J. Simulation of swirling coal comustion using a full two-fluid model and an USM turulence-chemistry model. Fuel 23;82:11 7. [7] Eaton M, Smoot LD, Hill SC, Eatough CN. Coponents, formulations, solutions, evaluation, and application of comprehensive comustion models. Prog Energy Comust Sci 1999;25: [8] Wei X, Han X, Schnell U, Maier J, Woerner H, Hein KRG. The effect of HCl and SO 2 on NO x formation in coal flames. Energy Fuels 23; 17(5): [9] Patankar SV, Patankar D. calculation procedure for heat, mass and momentum transfer in three-dimensional paraolic flows. Int J Heat Mass Transfer 1972;15: