Pre-drying of Lignite to increase Efficiency of Brown Coal Fired Power Plants

Transcription

1 INFUB 6th European Conference on -5 April, Estoril-Lisbon, Portugal Pre-drying of Lignite to increase Efficiency of Brown Coal Fired Power Plants Prof. Dr.-Ing. K. Görner, Dipl.-Ing. M. Liebetruth LUAT Lehrstuhl für Umweltverfahrenstechni und Anlagentechni (Tel.: ) (Institute for Environmental Process Engineering and Plant Design) ( Germany, Leimugelstr., 454 Essen ( Content Introduction.... Pre-dried Brown Coal Fired Power Plant Test ig... Mathematical Modelling.. Basic Equations.. Turbulence Modelling. 4. Pyrolysis Modelling. 5.4 Char Combustion Modelling Gas Phase eactions System Analysis General Furnace Performance Influence of Coal Humidity Coal and Process Model Parameters Conclusions. 6 eferences.. Abstract The use of pre-dried brown coals is one possibility of increasing the efficiencies of brown coal fired utility boilers. As a mean value an increase of -4 %-points can be achieved [], []. On the other hand the usage of such fuels leads to a modified combustion behaviour. Especially to loo at this point a mathematical model including all model parameters was developed and tested. For this tas the commercial software Fluent was used. In the study different models and model parameters were tested, especially the influence of the inetics for pyrolysis and char burnout, particle tracing procedures, radiation models and properties and others. For validation purposes experimental data was available by WE trials, done in their industrial combustion chamber in Niederaußem. Detailed axial and radial profiles of temperature and species concentrations for different brown coals and a large variety of process parameters were used. The comparison between experimental and numerical data forms the main part of this paper. Thus, the validity of the developed model can be demonstrated.

2 Introduction Brown coal fired power plants are fueled by raw coal with a water content of typically 5-55 wt-%. The coal drying process is realised in the coal mills by hot flue gases, where typical flue gas temperatures are in the range of. to. C. For that reason the exergy losses are very high and in consequence the over all efficiency is relatively low. By drying the coal outside the main process at a lower temperature level or by mechanical measures these losses can be reduced significantly []. Using such an external drying process the coal water mass fraction is decreased down to approx. wt-%. In this case the fuel is called pre-dried lignite. The drying process can be undertaen in a fluidised bed dryer heated by steam or by a mechanical device in combination with steam. The related overall process is shown in fig.. BC BC: aw Brown Coal DBC: Pre-dried Brown Coal BC DBC condensate Fig. : Basic schemes of the raw (left part) and pre-dried (right part) brown coal fired combustion process The reduced water content results in a completely different combustion behaviour. Flame temperatures increase faster and reach a higher absolute level. Because of the second fact the NO x emission is expected to be higher and the slagging and fouling behaviour must be taen into account very carefully. Looing in more detail into the combustion process leads to modified pyrolysis and char burnout characteristics. This was studied by mathematical modelling and numerical simulation. Different models and parameter sets were implemented and tested for this tas. Pre-dried Brown Coal Fired Power Plant Test ig WE in Germany is running a test rig for brown coal combustion. Both raw brown coal and pre-dried brown coal can be tested in the rig by simply changing the burner. The combustion chamber itself consists of a vertical cylindrical path fired from the top (see fig. ). After a horizontal path the convective heat exchangers are arranged. At different axial positions (related to the flue gas stream) a relatively high number of windows is positioned for measurement purposes. So axial and radial profiles can be achieved for temperatures and species. For model validation special trials were used where temperature profiles are compared for demonstrating model accuracy.

3 Fig. : Test rig at WE in Niederaußem utility Mathematical Modelling. Basic Equations The balance of momentum, species and enthalpy (heat) is realised by transport equations which have to be solved numerically. Momentum transfer is described by Navier-Stoes-Equations of the form: t i ( ρu i) + ( ρuu i ) = µ + µ δi + ρgi u i u Similar equations are used for enthalpy (heat) and species such as C x H y, O, CO, CO, H O and other gas phase species. Special emphasis has to be drawn on the heterogeneous combustion of coal, where a particle path and a gas/fluid path has to be described in some detail (fig. ). energy particle p rim ary p yrol ysi s particle p c l e - path trac r est part. ta r CO prim a r y volatile comp C x H y H gas/fluid g - / fluid path - p a th p art. s econd a r y volatile comp. C x H y ta r u Ash Part. CO H O CO H O soot CO H O CO soot H O p i Fig. : Principle scheme of the coal combustion model

4 Additionally the aerodynamic behaviour of the discrete phase is described in the Lagrangean way. This gives the possibility of coupling both framewors, Eulerean and Lagrangean (see fig. 4). Also the turbulent interaction (momentum transfer) between the particulate and the continuous phase is modelled applying Monte-Carlo- Models. The coal transformation itself can not be described as an isolated process but must be modelled as a very complex coupled process (see fig. 4) as a consecutive process along the particle trac. coal transformation g as phase Lagrange frame Euler frame dry brown coal heat up vapourization vaporization momentum, e nerg y water carbonhydroxides flow Navier - Stoes t urbulen ce NG - SM p yroly si s char burnout ash C x H y O z O CO r ea c tion Eddy - Dissipation radiation P - m odel Dis c rete - o rdinate - m odel Fig. 4: Interaction between continuous and discrete phase In addition, the complete model for the numerical simulation of the combustion of predried pulverised lignite was developed considering model and parameter accuracy, numerical robustness and applicability for large domains lie furnaces of steam generators.. Turbulence Modelling Turbulence modelling plays a very important role. For isotropic turbulence structures the --model of turbulence is convenient. The describing transport equations and the parameter set are as follows: t t µ ( ) ( ) eff ρ + ρu = + P ρ ( ρ ) + ( ρu ) = σ µ σ,eff eff,eff + c, P c, constant c µ σ,eff σ,eff c, c, value,9,,,44,9 ρ Transport equation for Transport equation for Model constants 4

5 Especially for swirling flows and the spreading rate of ets a better turbulence description can be achieved by using higher turbulence models lie the NG-- model. In this model the analytical derivation from the Navier-Stoes equations by using the "renormalization group" (NG) method results in constants that differ from those in the standard - model, and in additional terms and functions in the transport equations for and [4]. In nonisotropic turbulent flow situations lie swirled flows the eynolds-stress-model leads to more realistic results, normally. ûiû t c + c S,p S,4 + ( uûiû ) = cs, ûûl ( ûiû ) cs,p( P i, P δi ) u u i ( P i, P i ) cs,pûlû δ + l + cs, ( û iû δi ) l u ( u i ûiû δi ) + c S,5 ( Pi, Pi, ) û û + ûiû + δi i x N constant c S, c S, c S, c S,4 c S,5 values,5,4,5,5,5 All these turbulence models have been tested with respect to accuracy of complete model and model parameters sets.. Pyrolysis Modelling A large variety of different coal pyrolysis models are available. In the present wor the Kobayashi model [5] was used. The volatile matter release m V in this case is a function of two concurrent rate processes, and, which are dominant at different temperature levels. mv m m p, a t = ( α +α )exp( ( + E Tp = B e t E Tp = B e )dt)dt With: T P particle temperature, E, E activation energy, universal gas constant, α, α weighting factors Detailed model parameter tests had been carried out for the evaluation of the right set related to IFF measurements [6], carried out in a plug flow reactor. Pre-dried coal will show a different pyrolysis behaviour, because the water diffusion mass flow in the particle pore system and the boundary layer is significantly reduced. 5

6 .4 Char Combustion Modelling Char burnout is a function of diffusion and inetic controlled processes. Pyrolysis and char burnout tae place simultaneously and not as a consecutive process. Therefore the normal parameter set has to be evaluated. In the applied surface-reactionmodel [7] a diffusion rate and a inetic rate are calculated and weighted to give a resulting char combustion rate: [( T + T )/ ].75 p = C oxygen diffusion Dp ( E / ) C exp Tp = rate of reaction dm dt p = πdppo resulting rate of burnout + with D P, T P, m P as diameter, temperature and mass of particle, p O partial pressure of oxygen in the particle surrounding gas..5 Gas Phase eactions Gas phase reactions are modelled using a modified Magnussen-Hertager-Approach [8], the Eddy Dissipation Concept. In this model a rate of reaction is related to the rate of dissipation of the reactant- and product-containing eddies. i, In addition an Arrhenius reaction rate can be calculated and be accounted for. The slowest rate controls the rate of the reaction. The model parameters, especially A, were taen for special consideration. For the gas phase reactions, a standard - equation gas phase reaction approach was applied, assuming a very fast transformation from long chain hydrocarbons to lighter hydrocarbon radicals (C x H y O z ): CO-oxidation = ν i, m Mi Aρ ν M C x H y O z + (.5 y +.5 x - z) O.5 y H O + x CO CO +.5 O CO, i: species ID, ν: stoichiometric factors : reaction ID, : reactants A: mixing constant is also a rate limiting reaction and must also be modelled in this sense. 6

7 4 System Analysis 4. General Furnace Performance Fig. 5 shows the modelled combustion chamber with details of the chamber cooling, the low NO x -burner and the additional burnout (overfire) air inection. burner t erti a r y air Prim a r y air and coal s e ec o nd a r y air combustion chamber water cooled serie : VIP Q th :,8 MW m DBC : 5 g/h λ:,8 /, deflector additional burnout air burn out air measuring plane steam cooled Fig. 5: Details of the combustion chamber consisting of a cylindrical furnace with a top burner (primary, secondary and tertiary air) and an overfire air inlet The overall flow situation is shown in fig. 6. It is the typical shape of a et flame. ecirculation zones are present between the flame an the furnace walls and behind the additional air inection (overfire air). axial velocity m/s -.8 axial velocity m/s -.8 Fig. 6: Predicted velocity field left: full flow field right: recirculation zones 7

8 4. Influence of Coal Humidity Fig. 7 shows the influence of the coal water content on the axial temperature profile. Vaporization enthalpy leads to a later increase of the temperature and a smaller temperature gradient. The maximum axial temperature is slightly below of that of the water free case. Temperatur temperature [ C] mit with Wasser water without ohne Wasser water burner Brennerabstand distance [m] [m] Fig. 7: Influence of the coal humidity 5* -7 [g/s] 8 cm Fig. 8: Characteristic of water release Water release is assumed to be an thermal controlled process. With a reduced water content of the coal more energy is available for the pyrolysis and the pyrolysis rate is increased slightly. The water mass transfer is shown in fig. 8. It shows that the particles begin to release water immediately after inection, when hot secondary air comes in contact with the particle stream. The water release of larger particles becomes more intensive when the first volatiles start to react in the gasphase. 4. Coal and Process Model Parameters Turbulence model influence is shown in fig. 9. On the left hand side the results for the NG--turbulence-model and on the right hand side that of the eynolds-stress-model are shown. The discrepancies are very small. This is a nown effect for very high volume expanding processes, as flames are typically. The isotropic volume expansion leads to a rather isotropic turbulence structure. In the present case the et flame turbulence structure is relatively isotropic from the beginning. The use of the SM-model with its strong increase of numerical effort leads not to a significant improvement of accuracy. Therefore, especially under the aspect of the applicability to large computational domains the NG---turbulence model is recommended. 9.5 [m/s] -.7 Fig. 9: Comparison of the flow field as a function of different turbulence models left: NG-- Model right: eynolds- Stress-Model 8

9 The influence of the EDC parameters on the temperature field has been tested and is shown in fig.. Magnussen and Hertager proposed a mixing constant of A = 4 while the IFF suggested.6. The axial temperature profile with A= shows a slightly better correspondence with the measured profile. The overall agreement of all the results with measurements is the best using the original value of A = axiale temperature Temperatur in in C Standardwert default value A=4 8 A= 6 IFF-Vorschlag 4 IFF-proposal A=.6 Messung Measurment burner Brennerabstand distance in m Fig. : Comparison of different EDC parameters in relation to measured and predicted axial temperature distribution The influence on the temperature field is demonstrated in fig.. It shows that with a mixing constant less than 4 it is not possible to compute the real shape of the et flame because the reaction rate is unphysically low. standard parameter 8 T < K T < K standard T > 4 parameter K A= T < K A= T > 4 K IFF-proposal T > 4 K temperature [K] e mperature [K] Fig. : Comparison of the influence of different EDC-parameter on the temperature field As shown in sec.. of this paper the pyrolysis inetic for pre-dried coal is different from that of raw coals. Kobayashi model parameters were optimised under these aspects (see fig. ). The basic case of combustion modelling was done with inetic 6% weight loss [wt-%] 4% % measured inetic fitted in linear model A B C D E % particle temperature [K] measured inetic fitted in Kobayashi model Fig. : Comparison of different pyrolysis model parameters 9

10 values measured by the IFF [6]. Using these, the volatile release occurs too late, leading to an unstable flame ignition far from burner (see fig. ). The optimisation was done by evaluating new parameter sets as a function of particle temperature-time-history with respect to the pyrolysis gas release. An example of the influence of the parameters is given by fig.. The overall flame performance is significantly increased by this fact, as can be shown by fig. and 4 in comparison temperature [K] volatile [mass.-%] temperature [K] volatile [mass.-%] Fig. : Pyrolysis model parameters experimentally Fig. 4: Pyrolysis model parameters with set B * -7 [g/s] The char burnout rate is indicated in fig 5. After the ignition of particles the oxygen in the et is consumed relatively quicly, so further burnout occurs mainly at the ets boundary. The applied inetic/diffusion model is not able to account for a reduction of char reactivity caused by an increasing mass fraction of ash. By reducing the inetic rate in a reasonable range it was investigated if this model disadvantage has a significant influence on the computation results. Fig. 5: Char burnout rate ([g/s]) 5* -7 As expected the locations of the maximum burnout rates are a function of the oxygen transport to the particles and are therefore controlled by mixing processes. Only a strong reduction of the inetic rate leads to significant differences as shown in fig. 6. On the left hand side the maximum burnout rates for a reduction of the pre-exponential factor C down to % in comparison to the measured value on the right hand side. Consequently the model disadvantage can be neglected. Fig. 6: Maximum char burnout rates at different inetic rates ([g/s], right: reduced char reactivity)

11 Taing all these models and model parameters into account the comparison for the axial temperature distribution is very satisfying (fig. 7). 4 Temperatur temperature in [ C] Brennerabstand burner distance in [m] m ABL simulation Simulation measurment Messung Fig. 7: Comparison of measured and predicted axial temperature profiles A final question exists on the general validity of the complete model and model parameter set. This could be confirmed calculating and comparing other coals (brown coals) without varying these qualities. The optimisation was carried out with brown coal with a provenience Hambach. Also very good results could be achieved with the coal Garzweiler without any further model or parameter adaptation. So a general validity for henish brown coal can be confirmed. 5 Conclusions For pre-dried brown coals a mathematical model had been developed and evaluated. For this purpose the main sub-models lie turbulence model, pyrolysis model, char burnout model, gas phase reaction model and the adherent parameters had been investigated and optimised using experimental data from a.8 MW test rig. The general validity can be shown by predicting other coals without any additional parameter fits. The scale-up will be done for utility boiler furnaces partially or fully fired by pre-dried brown coal. 6 eferences [] Wic, W., Kallmeyer, D., VGB Kraftwerstechni, 77(997)4, pp - [] Pollac, M., Heitmüller,.J., VDI Berichte 8, (996), pp -48 [] Klutz, H.-J., Holzenamp, M., VDI Berichte 8 (996), pp 9-5 [4] FLUENT.INC: Fluent User Guide, Vol. (996) [5] Kobayashi, H., Howard, J.B., Sarofim, A.F., 6th Symp. (Int.) on Combustion, The Comb. Inst., (976), pp 4-45 [6] IFF Doc. No. L9/y/6, IFF esearch Station B.V., User Manual to IFF User Defined Subroutines to Fluent 4.5, 999 [7] Field, M.A., Combust. & Flame, (969), No., pp 7-5 [8] Magnussen, B.F., Hertager, B.H., 6th Symp. (Int.) on Combustion, The Comb. Inst. (976)