Three-dimensional combustion modelling of biomass fired pulverized fuel boiler

Transcription

1 Three-dimensional combustion modelling of biomass fired pulverized fuel boiler a M. Stastny, F. Ahnert, H. Spliethoff Section Thermal Power Engineering, Technical University Deljt, the Netherlands. Abstract A Computational fluid dynamic (CFD) model was applied for a 200 MW pulverised fuel boiler. Peat, demolition wood and wood residuals were used as fuel. The computer code FLUENT was used for the modelling of the combustion process inside the boiler. The FWG k-e turbulence model together withwall functions was adapted for characterization of the flue gas behaviour. Reaction between fuel and oxidizer was modelled using the mixture-fractionpdf approach. The CFD calculations were compared with acoustic pyrometry measurements and process data. 1 Introduction Electricity generation by coal combustion is the most spread technology (about 34% in the world) and is one of the biggest sources of air pollution. Therefore the development of clean combustion technologies and an increase of the efficiency is an indispensable task. Biomass combustion can be an economic way to contribute to the reduction of CO*. Biomass fuels are considered environmentally friendly for several reasons. First, there is no net increase in CO2 as a result of burning biomass fuel (in contrast to CO2produced by fossil fuel). Biomass consumes the same amount of CO2 from the atmosphere during growth as is released during combustion. The second reason is that most biomass fuels contain a very small amount of sulphur. In addition, the alkaline ash from biomass is able to capture some of the SO2 produced during combustion [l]. Commercially operated biomass combustion systems, however, suffer from severe operational problems, e.g. slagging, fouling and corrosion. Understanding, detection and diagnosis of slagging are imperative for

2 440 Advmced CompututiodMethods it1 Hwt Trmsfkr countermeasures. Fuel quality and flue gas temperature are key parameter for the slagging tendency. Flue gas temperature measurements with widely used thermocouples give limited information about local temperature on the furnace wall. CFD modelling can give additional information over the flue gas side of the boiler by calculation of the temperature distribution. In this work, a prelimnary study is presented to show the possibility of a CFD model and to provide information about processes inside the boiler. Geometrical and operational data were used to model the first gas path of the boiler. P1-back wall P2-front wall P - Figure 1: Scheme of the modelled area of the boiler This work is part of the EU- project "Slagging and Fouling Prediction by Dynamic Boiler Modelling". The general objective of this project is to apply a model to a biomass fired power plant to provide designers with a tool to predict and recognize slagging and fouling in an early stage of the process. 2 Biomass fuel Properties of biomass, which are important for combustion purposes, include pyrolysis behaviour, tar yield, volatile composition, together with the yield and composition of the char and its reactivity towards 02. A number of these parameters are required as input to the existing CFD particle combustion models, such as devolatilization yields and rates, composition of volatiles, char yield and char burning rates. Clearly, measuring all these properties for each biomass type is time consuming, expensive and imprecise [2]. One option is to use the similarity with coal combustion since biomass fuels follow the same sequence of pyrolysis, devolatilization and combustion as seen in low-rank coal combustion mechanisms. But there are some significant differences between coal and biomass combustion. Coal density is usually in the range from kg/m3 for

3 Advancc.d Compututiod Mcthods in HatTrumfc.r 44 1 low-rank coal to 2330 kg/m3for high-density pyrolytic graphite. In contrary the density of biomass fuels is in the range from 100 kg/m3 for straw to 500 kg/m3 for forest wood. Biomass usually consists of % volatile matter whereas coal consists of % volatile matter. The heating values of biomass fuels are appreciably lower than that of coals because of the higher degree of oxidation [l]. These differences can change temperature field in the boiler significantly as well as it has influence on the formation of pollutants. 3 Boilerdescription The boiler of the 200 MW pulverized fuel power plant Fyris (Uppsala, Sweden) has been modelled. The boiler geometry and modelled combustion chamber are shown in figure 1. There are 13burners installed, 9 (three rows) on the front wall and 4 (two rows) on the back wall. However, normally not more than 10burners are used at one time (most often the three burners in the upper front wall row are inactive). Each burner is rated (full-load) 39.5 MW on bio fuels. Injection ports at two different levels are installed. The ports at the lower level (UB and UF in figure 1) are used to inject over-fire air (OFA) and the upper level ports (UFB and UFH in figure 1) are used for recirculated flue gases and limestone injection. 4 Modelling approach The mathematical and physical model of the boiler is solved with numerical models developed for Computational Fluid Dynamics. CFD analysis involves fluid flow, heat transfer and chemical reactions. The modelling of the real boiler can be divided in the following parts: predeveloped by Fluent Ltd., has been chosen to perform the processing stage (geometry of the boiler, mesh), setting up the models (physical and chemical models), defining the boundary conditions based on the process data, solving and post processing stage (analysis of results, comparison with experimental data). It must be mentioned that the verification of the computational simulation in a real boilers is a difficult task. Local, detailed data for model verification are expensive to obtain, but even if the user has experimental data available, the comparison between these data and predictions leaves the user uncertain about both the errors in the model and experimental data [3]. Nevertheless, using the experience obtained by a CFD model can significantly improve the operation of a boiler, regarding stability, NO, and local material temperature of the walls. In this work, a commercial multipurpose CFD application, Fluent 5.7 simulation. 5 Governing equations For this study, an Eulerian approach has been used for the continuum phase and a stochastic Lagrangian description has been adopted for the biomass particles. The gas phase is described by the Navier-Stokes equations, coupled with appropriate equations for density and viscosity. A standard k-f model and a RNG

4 442 Advmced CompututiodMethods it1 Hwt Trmsf?r k-e model were used for modelling turbulence [4,5,6,7]. The standard k-e model was used for the preliminary solution and the final calculation was performed using the RNG k-e model. The reason for using the RNG k-e model was that the standard k-e model describes well flows without swirling or without sharp change within the computational region. It is necessary to use a more efficient turbulent model in case of swirling burners. It must be mentioned as well that there is no need to introduce wall functions when an RNG k-r model is used [4]. It is necessary to include energy conservation equation in the set of governing equations since the simulated problem is non-isothermal. This equation can be solved in terms of temperature adding the appropriate state equation (density expressed as a function of temperature and pressure) and the constitutive relationship between enthalpy, temperature and pressure. Biomass combustion is described in two global processes. The first step is assumed as homogeneous. During this step volatile matters escape from the biomass particle (devolatilization) and the combustion in the gaseous phase takes place, leading to the generation of the volatile combustion products. The second stage is assumed heterogeneous because the combustion of solid phase occurs, in this case char, giving off gaseous products of the heterogeneous combustion [3]. The volatile release modelling is based on a single kinetic rate model [2, 51. This model states that the rate of production of volatile gases is given by a first order reaction and the rate constant is expressed in an Arrhenius form, which correlates rates of weight loss with temperature. Regarding volatile gas combustion, the mixture fractiodf'df approach is used. The mixture fractiodpdf modelling approach involves the solution of transport equations for one or two conserved scalars (the mixture fractions). In this approach, transport equations for individual species are not solved. Instead, individual component concentrations for the species of interest are derived from the predicted mixture fraction distribution. The reacting system is treated using infinitely fast chemistry, chemical equilibrium calculations, or nonequilibrium (flamelet) calculations. In this study a chemical equilibrium is assumed. Physical properties of chemical species and equilibrium data are obtained from the chemical database. Finally, time-averaged values are obtained by weighting the instantaneous values with a probability density function, based on the mean mixture fraction and its variance. Here, a beta function was assumed [5,7]. The char combustion has to be modelled as well. Char oxidation is a much slower process than devolatilization and it therefore determines the burnout time of pulverized biomass in the boiler [2].In this case, a kinetiddiffusion surface reaction rate model was applied for the modelling of the char combustion. The kinetiddiffusion-limited rate model assumes that the surface reaction rate is determined either by kinetics or by a diffusion rate [5]. The resulting rate of reaction is termed global since it incorporates the influence of the pore surface area. When the biomass combustion is modelled, it is necessary to combine the combustion models with a particle transport calculation. In this case, a Lagrangian approach has been adopted. Within the particle transport modelling, the total mass flow of biomass particles has been modelled by tracking a number

5 183: 2002 WIT Press, Ashurst Lodge, Southampton, SO40 7AA, UK. All rights reserved. Advancc.dCompututiodMcthods in HatTrumfc.r 443 of trajectories much slower than the real number of particles, assuming that each simulated particle represents a sample of the real number of particles. In this case, the total number is limited to trajectories. Primary air mass flow Table 1: Boundary conditions Secondary air mass flow Tertiary air mass flow Transport air mass flow 1 Purge air mass flow OFA mass flow Total fuel mass flow ~ Primary air temperature Nm3/ h Nm3/h Nm' /h Nm3/ h 1 1 8SONm3/h Nm3/h 5/ S 1 Rec. flue gas mass flow Nm3/h 1 Secondary air temperature Tertiary air temperature 1 1 Purge air temperature OFA temperature Rec. flue gastemperature., Swlrl angle of the primary air 340 C 1 27 "C 340 "C C 1 Transport air temperature 1 50 "C 1 Swdangle of the secondary ax Wall temderature 1i; 430 "C Wall emissivity 0.8 Turbulence intensity at the 10 % velocity inlets l Radiative heat transfer which is the predominant mechanism in coal or biomass fired industrial furnaces was simulated using the P-1 radiation model [S]. The P-1 model is the simplest approximation of the P-N model, which is based on the expansion of the radiation intensity into an orthogonal series of spherical harmonics. 6 Boundary conditions When the problem is defined in terms of necessary equations, then the definition of the boundary conditions is required. Inlets are defined as the velocity inlet and outflow conditions are expressed as outlet with the external radiation temperature. This temperature is equal to the wall temperature and approximately corresponds with the vaporization temperature of water in the

6 444 Advmced CompututiodMethods it1 Hwt Trmsfkr evaporator. In this case, the temperature of the walls is equal to 703 K with an emissivity of 0.8.The summary of the boundary conditions is shown in table I. Table 2: Fuel analysis of burned biomass I Proximate analysis (%) 1 Ultimate analysis (dry-ash-free) (%) Moisture Fixed Carbon 5.6 C 23.6 H Volatiles N I 2.2 Ash Lower Heating Value (MJ / kg) S 37.1 The supply of fuel is modelled as injection (16 injection points per burner). The number of active burners was ten. The density of burned biomass was calculated with 400 kg/m3 and the specific heat with 1300 Jkg-K [9]. The analysis of the biomass fuel currently being burned is shown in table 2. The particle sizes were simulated and the measured mass fractions for various size classes are shown in table 3. The calculation on these sizes was performed by means of the Rosin-Rammler expression [5]: 0 where D is the mean diameter and n is the size (spread) distribution parameter. In this case, D =0.211m and n = Table 3: The percentage of various size classes of pulverized peat E Diameter ' range [mm] Mass fraction in range [%] Solution procedure The computer code FLUENT is used for solving the governing differential equations in conjunction with the second order upwind technique and underrelaxation factors. The coupling between velocity and pressure is achieved by the SIMPLEC method. The mesh was generated using tetrahedral elements. The initial number of cells was and the final amount of cells after adaptation was The calculation strategy started with solving the gas flow-field equations assuming that the particles, energy and radiation equations are switched-off. After establishing the flow field, the radiation equations, energy balance and trajectories of the particles are solved. Using the flow field, the temperature of

7 Advancc.dCompututiodMcthods in HatTrumfc.r 445 the particles and the burnout histories are determined. The mass, momentum and energy equation for each cell is calculated. The source terms are included in the gas phase equations and the flow field is then recalculated. Level [m Table 4: Results for different levels of the boiler Tavg [K1 Tmax W1 wavg [dsl wmax [m/sl Result and discussion The calculation was performed on a-linux platform (SuSe 7.2) with an AMD Athlon 1.4 GHz processor and 1 GB RAM. The post processing tools of the FLUENT code was used for the examnation of the results Figure 2: Particle trajectories tracked for different inlets

8 225ebO2 2 lx%*02 158aQZ 7 3bC "302 : a k O 2 ~.ale+ol 6.75RUli I l w e 4 1 i!675s+ul 4 54S&o! 225a+m s v 3 a Figure 3: Velocity flow field Process data were available for the comparison with the calculated results. However, there were some limitations. Additional temperature measurements were performed by acoustic pyrometry, to measure the temperature inside the furnace. The measurement devices were installed on one level below the superheater section at a height of m (figure 1). Due to slagging problems not all theoretical possible temperature paths could be measured and no complete temperature distribution field could be obtained. Nevertheless some single paths give already valuable information about the condition in this cross-section. The mean temperature is in the range from 1250 "C to 1270 "C.The mean temperature from the calculation is equal to 1149 "C which is slightly lower than the mean temperature from the measurement. The maximum temperature at this cross-section is 1517 "C, the mean velocity is 11.5 d s and the maximum velocity is 32.4 ds.the summary of the flow field conditions at different levels is presented in table 4. The particle trajectories tracked for different inlets (burner 5, 7 and 13 in accordance with figure l) are shown in figure 2 and the velocity flow field for several cross-sections through the burners is presented in figure 3. A strong flow recirculation is observed in the ash pit zone in both figures. The particles hit the furnace wall at some places, especially in the upper part of the boiler near the contraction of the first path of the boiler. The swirling area is also located at this place. This can increase abrasive wearing of the wall or can cause slagging when the particles melt.

9 Advancc.dCompututiodMcthods in HatTrumfc.r 447 The temperature distribution, at the same cross-sections as in figure 3, is shown in figure 4. The maximum temperature in the boiler is 2301 "C. As above the mean temperature in the measured area agrees with the measured temperature. Nevertheless it can be supposed that the temperature in the region of the superheaters and reheaters is not correctly predicted since these heat exchangers have not been modelled. The mixing of the fuel with the combustion air can be observed close to the burners and most particles burnout in this region. Thus it is guaranteed that the temperature peaks are located in this area. : : : : 1 IJ.aoe42 Figure 4: Temperature distribution 9 Conclusion and outlook A prelimnary study of the CFD modelling for a pulverized fuel power plant was carried out. This modelling applied a similar model for coal combustion to biomass. The simulation could predict the furnace exit temperature. The used models are reasonable and can be applied for further simulations without higher demands on CPU and memory. Further work should examine some process states in the boiler for different operational conditions. This can lead to recommendations for the optimization of the combustion process in order to maximise efficiency, to decrease propensity of slagging and to minimise pollutant emissions.

10 Acknowledgements The authors would like to thank the European Commission for supporting the research project Slagging and Fouling Prediction by Dynamic Boiler Modelling (contract ERKS-CT-I ). References [l] Sam,M., Annamalai K.,M. Wooldridge, Co-firing of coal and biomass fuel blends. Energy and Combustion Science, 27, pp ,2001. [2] Williams, A., Pourkashaninan, M., Jones, J.M., Combustion of pulverised coal and biomass. Energy and Combustion Science, 27, pp , [3]Iranzo, I., Domingo, E. CortQ, C., kauzo, I., Combustion characterisation of a pulverised coal utility boiler based on CFD techniques, 6th International conference on technologies of combustion for a clean environment, Porto, pp ,2001. [4] Fan, J., Qian, L., Ma, Y., Sun, P., Kefa, K., Computational modeling of pulverized coal combustion process in tangentially fired furnaces, Chemical Engineering Journal, 81, pp ,2001. [5]Fluent 5.5 documentation, Fluent, Inc., 1999 [6] Kjaldman, L., Numerical simulation of combustion and nitrogen pollutants in furnaces, VTT Finland, [7] Versteeg, H.K., Malalasekera, W., An introduction to computational fluid dynamics- Thefinite volume method, Longman Group Ltd., [S] Baukal, Ch.E., Gershtein, V.Y., Li, X., Computational fluid dynamics in industrial combustion, CRC Press LLC, [9] Dlouhy, T., Vypocty kotlu a spalinovych vymeniku, CTU Prague, 1999.