Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/231273422

A Premixed Flamelet−PDF Model for Biomass Combustion in a Grate Furnace

Article  in  Energy & Fuels · May 2008


DOI: 10.1021/ef7007562

CITATIONS READS

23 926

5 authors, including:

R. J. M. Bastiaans Jeroen A. van Oijen


Eindhoven University of Technology Eindhoven University of Technology
114 PUBLICATIONS   1,658 CITATIONS    188 PUBLICATIONS   3,639 CITATIONS   

SEE PROFILE SEE PROFILE

Philip de Goey
Eindhoven University of Technology
425 PUBLICATIONS   8,355 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Turbulent combustion with FGM View project

Biomass combustion modeling View project

All content following this page was uploaded by R. J. M. Bastiaans on 19 May 2018.

The user has requested enhancement of the downloaded file.


A premixed flamelet-PDF model for biomass combustion
in a grate furnace
Albrecht, B.A.; Zahirovic, S.; Bastiaans, R.J.M.; van Oijen, J.A.; de Goey, L.P.H.

Published in:
Energy & Fuels

DOI:
10.1021/ef7007562

Published: 01/01/2008

Document Version
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the author’s version of the article upon submission and before peer-review. There can be important differences
between the submitted version and the official published version of record. People interested in the research are advised to contact the
author for the final version of the publication, or visit the DOI to the publisher’s website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page numbers.
Link to publication

Citation for published version (APA):


Albrecht, B. A., Zahirovic, S., Bastiaans, R. J. M., Oijen, van, J. A., & Goey, de, L. P. H. (2008). A premixed
flamelet-PDF model for biomass combustion in a grate furnace. Energy & Fuels, 22(3), 1570-1580. DOI:
10.1021/ef7007562

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
• You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal ?
Take down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.

Download date: 23. Mar. 2017


1570 Energy & Fuels 2008, 22, 1570–1580

A Premixed Flamelet-PDF Model for Biomass Combustion in a


Grate Furnace
B. A. Albrecht†,§, S. Zahirovic‡, R. J. M. Bastiaans*,†, J. A. van Oijen†, and L. P. H. de Goey†
Combustion Technology, Department of Mechanical Engineering, Technische UniVersiteit EindhoVen, Den
Dolech 2, 5600 MB EindhoVen, The Netherlands, and Institute for Resource Efficient and Sustainable
Systems, Graz UniVersity of Technology, Infeldgasse 21B, 8010 Graz, Austria.

ReceiVed December 13, 2007. ReVised Manuscript ReceiVed March 10, 2008

In this paper, a new modeling approach for biomass furnaces is presented. A flamelet model is applied for
the prediction of combustion and NOx emissions in a pilot-scale low-NOx biomass grate furnace. The model
describes the combustion chemistry using premixed flamelets. The chemical system is mapped on three control
variablessthe mixture fraction, enthalpy, and a reaction progress variablesby means of the flamelet-generated
manifold chemical reduction technique (FGM). The density and source terms are tabulated as functions of the
control variables in a preprocessing step to speed up the numerical calculations. The turbulence-chemistry
interaction is described by an assumed shape probability density function (PDF) approach. Generally, transport
equations are solved for mean and variance of mixture fraction, mean progress variable, and the mean enthalpy.
The FGM-PDF model is first validated for a diffusion flame: Sandia Flame D. Good agreement of predictions
with measurements is found. The model is then applied to a 2D cross section of a biomass combustion grate
furnace. FGM-PDF is compared with the standard eddy dissipation concept model (EDC). The predictions of
the two models are similar. The FGM-PDF model reduces the calculation time with GRI-Mech2.11 reaction
mechanism from weeks to hours, when compared to EDC. Furthermore, in FGM-PDF more physics can be
taken into account thanks to the integration of flamelets in the turbulence-chemistry interaction.

1. Introduction being determined by the turbulent mixing rate. Therefore,


accurate predictions of intermediates and pollutants cannot be
The use of grate furnaces is regarded as a popular combustion achieved. An improved version of EDM is the eddy dissipation
technology used for thermal conversion of solid biomass fuels. concept (EDC).3,4 This model can handle complex reaction
The combustion of solid biomass in these furnaces can lead to mechanisms but uses an implementation of turbulence-chemistry
a considerable exhaust of NOx, originating from the nitrogen interaction based on homogeneous reaction. Furthermore, it
present in biomass fuels. The levels of NOx emitted depend on requires a very high computational time.
the amount of the nitrogen in the fuel as well as on the design The model proposed here for the modeling of biomass grace
and operation parameters of the furnace. As NOx compounds furnaces is based on a (premixed) flamelet-PDF (probability
are harmful to the environment, governments continue to lower density function) approach implemented in RANS. The com-
their emission limits. Therefore, research is done into the bustion system is described with detailed chemistry by premixed
reduction of NOx emissions. To optimize the design of biomass flamelets, the advance of chemical reaction being monitored
combustion furnaces in general and for NOx emission reduction by a reaction progress variable. The turbulence-chemistry
by primary measures in particular, numerical models can be interaction is accounted for with an assumed shape PDF
used. The requirements for these models are twofold: the model approach. Therefore, this combustion model has a low calcula-
has to be sufficiently detailed to predict the NOx production tion time (comparable to EDM) and is thought to improve the
accurately, and the calculation time of the model must be within accuracy of gas-phase species and temperature predictions due
reasonable limits to be able to calculate the emission levels for to the more physical turbulence-chemistry interaction compared
different design parameters and operating conditions. to EDC.
Combustion modeling for industrial applications like biomass
furnaces relies usually on the Reynolds averaged Navier-Stokes 2. The FGM-PDF Combustion Model
(RANS) simulation approach. Within RANS, the most com-
monly used combustion model is the computationally inexpen- In this paper, a flamelet model is used for the modeling of
sive eddy dissipation model (EDM) coupled with a global combustion and NOx emissions in biomass grate furnaces. The
reaction scheme.1,2 The problem with EDM is that it does not model was reported in detail elsewhere;5,6 here only a brief
take into account detailed chemical kinetics, the reaction rates description of the model is given. The model contains the

* To whom correspondence should be addressed. E-mail: r.j.m.bastiaans@ (1) Scharler, R.; Fleckl, T.; Obernberger, I. Prog. Comput. Fluid Dyn.
tue.nl. 2003, 3, 102–111.
† Technische Universiteit Eindhoven. (2) Scharler, R. Ph.D. Thesis, Graz University of Technology, Austria,
‡ Graz University of Technology. 2001.
§ Present address: Advanced Engineering Engines, DAF Trucks N.V., (3) Fluent 6.1 User’s Guide; Fluent Inc.: Lebanon, NH, 2003.
Hugo van der Goeslaan 1, P.O. Box 90065, 5600 PT Eindhoven, The (4) Gran, I.; Magnussen, B. Combust. Sci. Technol. 1996, 119, 191–
Netherlands. 217.

10.1021/ef7007562 CCC: $40.75  2008 American Chemical Society


Published on Web 05/03/2008
Flamelet-PDF Model for Biomass Combustion Energy & Fuels, Vol. 22, No. 3, 2008 1571

following elements. A laminar database is constructed using used as a progress variable is not unique. The requirement
the flamelet generated manifold (FGM) chemical reduction imposed on such a combination is the monotonical variation of
method.7 This method tabulates all the thermochemical variables its mass fraction between reactants and products with a
(species mass fractions, temperature, density, etc.) made avail- reasonable resolution for the entire range. Therefore, the mass
able as solutions of freely propagating, premixed flame simula- fraction of CO2 was chosen as the progress variable
tions. This is tabulated as a function of the mixture fraction,
Y ) YCO2 (2)
enthalpy, and a progress variable, referred to as control variables.
A turbulent database is consequently calculated by using an In our FGM-PDF combustion model, we will use the scaled
assumed shape PDF-averaging approach for the control vari- form of the progress variable, denoted by c (between 0 and 1,
ables. This results in an extension of the control variables with like the mixture fraction), for the PDF closure and the unscaled
the variances of the original (laminar) control variables. In this form Y in the transport equations. The motivation of this
paper, the enthalpy is considered as distributed according to a approach is given in the following. In the present model, all
delta function and thus the variance of enthalpy is not taken the thermochemical variables (species mass fractions, density,
into account. The obtained database serves as a look-up table temperature, etc.) are functions of the mixture fraction and the
and is calculated in a preprocessing step to speed up the turbulent progress variable. It follows that the turbulence-chemistry
flame simulations. During a simulation, turbulent transport interaction is described by the joint PDF of these two control
equations for the control variables are solved. The simulation variables. In the presumed shape PDF approach used here, the
makes use of the density and progress variable source term given shape of the single-variable PDF is assumed. The joint PDF
by the table look-up. At the end of the simulation, the species can be constructed from the single-variable PDF if the variables
mass fractions and the temperature corresponding to the values are statistically independent. Because the fluctuations of an
of the control variables in every point of the computational unscaled progress variable Y are strongly related to the fluctua-
domain are retrieved from the tables and thus the complete tions of the mixture fraction Z, the joint PDF of Z and Y cannot
structure of the combustion system is made available. Due to be decomposed in independent factors. To remove the correla-
the fact that the model uses the FGM technique to reduce the tion (or at least to reduce it significantly), a scaled progress
chemical kinetics and a PDF approach for the turbulence-chemistry variable c is used instead of Y.11,12 The scaled progress variable
interaction, the model will be referred to as the FGM-PDF c is defined as
model.
Y - Yu
c) (3)
3. Turbulent Transport Equations for Control Variables Yb - Yu
3.1. Laminar Control Variables. In the FGM-PDF model, In eq 3, Yu is the unscaled progress variable in the reactants
transport equations for control variables are solved together with (unburnt mixture) and Yb in burnt products at equilibrium. Thus,
the averaged equations of continuity and momentum. In this upon entering the turbulent lookup table the progress variable
section we will present the equations for the control variables. is first scaled to retrieve data. This way of modeling is preferred
Therefore, we first have to mention the definition of the control above the use of c directly in the transport equations because
variables that we apply. For the mixture fraction, the Bilger that would introduce additional terms in these equations. These
formulation8 modified to exclude the oxygen terms was em- additional terms are related to the dependency of c on Z in the
ployed. This is similar to that used in the experimental work denominator in eq 3.12 The modeling of these terms is complex
on Sandia Flame D, which is the flame chosen for the validation and approximations have to be used to close them.11
of the FGM-PDF model9,10 In principle, the combination of a mixture fraction and a
progress variable determines the chemical state of the system

(
1 ZH
2 MH
+2
ZC
MC
- )( 1 ZH
2 MH
+2
ZC
MC ox ) if it were adiabatic. In the application of a biomass grate furnace,
however, use is made of recirculation of flue gases and also

( ) ( )
Z) (1) secondary air is introduced in the furnace. Therefore, the scaled
1 ZH ZC Z
1 H ZC enthalpy, h, is needed to include these effects
+2 - +2
2 MH MC fu 2 MH MC ox
H - Hmin
In eq 1, ZH and ZC are the element mass fractions for hydrogen h) (4)
Hmax - Hmin
(H) and carbon (C), respectively, and MH, MC are their molar
masses. Following eq 1, Z varies between 0 in the oxidizer where H is the actual (total) enthalpy, Hmax is the adiabatic
stream (denoted with “ox”) and 1 in the fuel stream (denoted enthalpy for a specific Z, and Hmin is the enthalpy of the burnt
with “fu”). products corresponding to Z, when brought to the ambient
For the progress variable, it is natural to take a combination temperature Tmin ) 298 K. The scaled enthalpy, h, was used as
of reaction products, but even intermediates and reactants can control variable in the turbulent database while in the transport
be taken into account. The choice of the combination of species equations solved by the FGM-PDF model the unscaled enthalpy
H was involved. This approach is similar to that used for the
(5) Ramaekers, W. J. S. M.Sc. Thesis, Eindhoven University of progress variable and is motivated by the fact that solving a
Technology, Eindhoven, The Netherlands, 2005.
(6) Albrecht, B. A.; Bastiaans, R. J. M.; van Oijen, J. A.; de Goey, L.
transport equation for a scaled variable (like h and c) leads to
P. H. Submitted for publication in Flow, Turbulence Combust., 2008. extra source terms. These are caused by the dependence of the
(7) van Oijen, J. “Flamelet-Generated Manifolds: development and denominator of eq 4 on Z.
application to premixed laminar flames”; Ph.D. Thesis, Eindhoven University 3.2. Turbulent Control Variables: PDF Averaging. In the
of Technology, Eindhoven, The Netherlands, 2002.
(8) Bilger, R. W.; Starner, S. H.; Kee, R. J. Combust. Flame 1990, 80, present study, the averaging procedure for any turbulent
135–149.
(9) Barlow, R. S.; Frank, J. H. Piloted CH4/air flames C, D, E and F - (11) Louis J. J. J. “On Turbulent Combustion of Coal Gas”; Ph.D. Thesis,
release 2.0, Sandia National Laboratories, Livermore, 2003; http://www- University of Twente, 1997.
.ca.sandia.gov/TNF. (12) Bray, K.; Domingo, P.; Vervisch, L. Combust. Flame 2005, 141,
(10) Barlow, R. S.; Frank, J. H. Proc. Combust. Inst. 1998, 27, 1087. 431–437.
1572 Energy & Fuels, Vol. 22, No. 3, 2008 Albrecht et al.

(( ) )
fluctuating quantity φ, which, in terms of the control variables ∂FũjZ̃′′2 µt ∂Z̃′′2 µt ∂Z̃ ∂Z̃
∂ ε̃
is defined as φ(Z,c,h), is based on Favre averaging, given by ) FD + + C1 - C2F Z̃′′2
∂xj ∂xj Sct ∂xj Sct ∂xi ∂xi k̃
φ̃ ) ∫10∫10∫10φ(Z,c,h)P̃(Z,c,h) dZ dc dh (5) (15)

(( ) )
with P̃(Z,c,h) ) FP(Z,c,h)/Fj. Because of the introduction of the
∂FũjỸ′′2 ∂ µt ∂Ỹ′′2 µt ∂Ỹ ∂Ỹ ε̃
scaled progress variable, we removed correlation with the ) FD + +2 - C2F Ỹ′′2 +
mixture fraction. This allows us to compute the joint PDF as a ∂xj ∂xj Sct ∂xj Sct ∂xi ∂xi k̃
product of marginal PDF’s of Z, c, and h 2Y ′′ ωY (16)
P(Z,c,h) ) P(Z) P(c)P( h) (6) In the equations, Fj is the mean density, ũj is the mean velocity
for the enthalpy this is trivial because it is taken as a delta vector, D is the molecular diffusion coefficient, µt is the turbulent
function distribution. For the marginal PDF’s of Z and c, we viscosity, Sct is the turbulent Schmidt number, ω j Y is the mean
follow many other investigators, e.g.,refs 13, 14, by assuming chemical source term of the progress variable, C1 and C2 are
a β-function shape PDF. These PDF’s depend on the mean and modeling constants, k̃ is the mean turbulent kinetic energy, and
variance of the control variables ε̃ is the mean dissipation rate. In the equations given above,
the gradient transport assumption15 coupled with unity Lewis
P(Z) ) P(Z,Z̃,Z̃′′2) (7) numbers for all variables have been used to model the turbulent
˜ diffusion terms.
P(c) ) P(c,c̃,c′′2) (8) Besides these equations, the Favre-averaged Navier-Stokes
equations and equations for k̃ and ε̃, given by the turbulence
resulting in
model, are solved during a simulation. Furthermore, the values
∫∫ of constants Sct, C1, and C2 are 0.85, 2.43, and 2.0, respectively.
1 1
φ̃ ) ˜2)P̃(c,c̃,c′′
φ(Z,c,h)P̃(Z,Z̃,Z′′ ˜2) dZ dc (9)
0 0 These values are recommended in ref 3 for the mixture fraction.
This introduces another complication because we need an In this work, they have been also adopted for the mean and
averaged scaled mean and variance of the progress variable. In variance of the progress variable.
other words, we need expressions for c̃ and in terms of Ỹ and. To conclude, the turbulent database is constructed a priori to
The complete derivation of these explicit relations is given in speed up the numerical simulations. This is basically a five-
ref 6 and the result is dimensional look-up table where the density, source term of
progress variable, temperature, and species mass fractions are
Ỹ - Ỹu given as functions of Z̃, Z′′ ˜2 , c̃, ˜ c′′2 , and h̃. Transport
c̃ ) (10)
Ỹb - Ỹu equations are solved for Z̃, Z̃′′2, Ỹ, Ỹ′′2, and H̃. Then c̃, c̃′′2 are
calculated with eqs 10 and 11 from Ỹ and Ỹ′′2. The scaled
and enthalpy, h̃, is determined with eq 4 from H̃. Then, Z̃, Z̃′′2, c̃,
˜2 + Ỹ2 - Y
˜ ˜ ˜ c̃′′2, and h̃ are used to retrieve the density and the chemical
˜ Y′′ u2
- 2Y (Y - Yu)c̃ - (Y
u b b
- Yu)2c̃2
c′′ 2 ) (11) source terms of Ỹ and Ỹ′′2 from the turbulent database. The
˜
(Y - Y )
b u density and the source terms are then used back in the transport
equations. The other variables from the database (temperature
To close the system, we only have to additionally tabulate the
and species mass fractions) are retrieved at the end of the
terms consisting of Yu and Yb as functions of Z̃ and in the
simulation for post-processing.
database.
3.3. Transport Equations. The laminar transport of Z, Y,
and H is given by standard convection-diffusion transport 4. Validation of the FGM-PDF Combustion Model
equations, which differ from each other only through the The FGM-PDF model was validated for a diffusion turbulent
presence of a source term in case of Y (the mixture fraction flame, the Sandia Flame D. This flame was selected for model
and the enthalpy have no source terms, they are conserved validation because it is very well documented9,10,16 and it
scalars). The turbulent transport equations for the mean of Z, received considerable attention from different investigators.17–22
Y, and H are derived by Favre averaging their laminar In this paper, only a selection of the results obtained with the
counterparts. The variance equations can be derived from the
scalar laminar transport equations by multiplying both sides with
(15) Peters, N. Turbulent combustion; Cambridge University Press:
the fluctuation of the scalar followed by Favre averaging (for London, 2000.
derivation, see, for example, ref 15). The steady-state transport (16) Schneider, Ch.; Dreizler, A.; Janicka, J.; Hassel, E. P. Combust.
equations for the control variables read Flame 2003, 135, 185–190.
(17) Ihme, M.; Pitsch, H. LES of a non-premixed flame using an

((
) )
extended flamelet/progress variable model, 43rd AIAA Aerospace Sciences
∂F̃ũjZ̃ ∂ µt ∂Z̃ Meeting and Exhibit, 10-13 Jan 2005, Reno, NV; Paper 2005-0558.
) F̃D + (12) (18) van Kuijk, H A.J.A.; Bastiaans, R. J. M.; van Oijen, J. A.; de Goey,
∂xj ∂xj Sct ∂xj
L. P. H. Modelling NOx-formation for application in a biomass combustion

(( ) )
furnace. In Proceedings of the European Combustion Meeting 2005;
∂FũjỸ ∂ µt ∂Ỹ Vandooren, J., Ed.; Combustion Institute: Louvain-la-Neuve, Belgium, 2005;
) FD + + ωY (13) Paper 160.
∂xj ∂xj Sct ∂xj
(19) Zahirovic, S.; Scharler, R.; Obernberger, I. Advanced gas phase

(( ) )
combustion models: valiadation for biogases by means of LES and
∂FũjH̃ ∂ µt ∂H̃ experiments as well as application to biomass furnaces. In Proceedings of
) FD + (14) The 7th European Conference on Industrial Furnaces and Boilers; Porto,
∂xj ∂xj Sct ∂xj
18-21 April 2006.
(20) Pitsch, H.; Steiner, H. Phys. Fluids 2000, 12:10, 2541–2554.
(13) Correa, S. M.; Shyy, W. Prog. Energy Combust. Sci. 1987, 13, (21) Xu, J.; Pope, S. B. Combust. Flame 2000, 123, 281–307.
249–292. (22) Roomina, M. R.; Bilger, R. W. Combust. Flame 2001, 125, 1176–
(14) Jones, W. P.; Whitelaw, J. H. Combust. Flame 1982, 48, 1–26. 1195.
Flamelet-PDF Model for Biomass Combustion Energy & Fuels, Vol. 22, No. 3, 2008 1573

FGM-PDF model for Sandia Flame D is presented. For more centerline are compared with measurements. The prediction of
results and a comprehensive analysis of the simulations, the the flow field (Figure 1a) shows that the axial velocity is well
reader is referred to ref 6. predicted in the first and last part of the domain, but it is
4.1. Computational Mesh, Boundary Conditions, and overpredicted around x/D ) 25. On the other hand, the turbulent
Flamelet Generation. Sandia Flame D is a piloted jet flame, kinetic energy is underpredicted in the interval x/D ) 10-30.
the fuel consisting of a mixture of CH4 (25 vol % ) and air (75 In Figure 1b, the mean and variance of the mixture fraction
vol % ). The fuel nozzle has a diameter of D ) 7.2 mm and it are presented. Both quantities are in reasonable agreement with
is surrounded by a pilot nozzle with a diameter of Dp ) 2.53D the measurements in spite of the deviations in the flow field. It
and an air coflow. The pilot produces the same equilibrium can be noticed that the shape and the magnitude of the variance
composition as a mixture of the main fuel and oxidizer with a of the mixture fraction are well predicted, but the maximum
mixture fraction of Z ) 0.27, which is slightly lean compared value is slightly shifted to the left.
with the stoichiometric mixture fraction of Z ) 0.35. The bulk Panels c-f of Figure 1 present predictions of species and
velocities of the fuel jet, pilot and air coflow are 49.6, 11.4, temperature, given by the combustion model. In general, all main
and 0.9 m/s, respectively. The corresponding fuel stream based species (CH4, O2, CO2, and H2O) and the temperature show
Reynolds number is Re ) 22 400. good agreement between simulations and experimental data. As
Due to the symmetry of the flame, a 2D axisymmetric for the intermediates (CO and H2), it can be seen that these are
modeling domain was employed. The structured numerical grid overpredicted for x/D < 40. The difference in the accuracy of
(125D × 42D) consists of 359 × 113 nodes. The central fuel the predictions for CO2 and CO (see Figure 1d) could be
jet (half of it) contains 5 cells on the radius with a minimum explained by the fact that probably the formation of CO does
cell size of 0.7 mm. Stretching is applied toward the outer lateral not follow closely the progress of CO2 (which is the progress
boundary. The minimum axial cell size is 1.2 mm and stretching variable). It follows that, to improve the prediction of CO, an
is applied toward the outlet. At the inlet boundary all required extra transport equation for the CO mass fraction might be
quantities are specified according to the experimental data given needed. Nevertheless, this step was not taken in the present
in ref 16 for the fuel jet, pilot and coflowing air stream. investigation. It is also worth mentioning that modest predictions
The flamelets used for the creation of the database have been of CO and H2 were also reported by other investigators (see
computed with the laminar flame code CHEM1D.23 The for example ref 20), in spite of the fact that in ref 20 a much
chemistry was described with the GRI-Mech3.0 reaction mech- more elaborate model was used which combines the Lagrangian
anism.24 The turbulence model was the realizable k-ε model.3,25 flamelet model and unsteady flamelets with LES.
The laminar database was constructed from 100 premixed In Figure 1f, besides the temperature, the mass fraction of
flamelets with mixture fractions covering the flammability range NO is also plotted. It can be observed that the predicted profile
from Z ) 0.18 to Z ) 0.56. These flammability limits of NO is close to the measurements only in the central part of
correspond to the values reported in ref 26. Outside this range, the domain. Thus, the magnitude of the maximum value of NO
all the variables have been interpolated between the values at is well predicted though its position is shifted to the left. In the
the flammability limits and at the inlets (Z ) 0 and Z ) 1). The rest of the domain, however, the predicted NO is practically
reason for taking premixed flames, even in this nonpremixed zero, while the experimental data show that NO is present. This
test case, is in the generality of the method. Parametrization on shows that the transport of NO is not well represented in the
the basis of mixture fraction and a progress variable was already model and it can be improved by solving an extra transport
introduced by, e.g., ref 27, but they used nonpremixed flamelets equation for the NO mass fraction.
as a basis. Here we will use premixed flamelets because then In conclusion, the FGM-PDF model gives a reasonable
the progress is well-defined. In this way, the use of a scalar- agreement for predictions of the experimental data for Sandia
dissipation rate is circumvented which is associated with Flame D. The overprediction of CO and H2 in the rich domain
additional assumptions. If the chemistry is close to equilibrium, indicates that these intermediates are in reality determined by
it does not matter what flamelets are used. Moreover, in the the turbulent transport (diffusion) rather than by the chemical
grate furnace, gradients of progress variable are assumed to be composition corresponding to the local combination of mixture
larger than the gradients of mixture fraction, the latter being fraction and progress variable. A similar finding was also
associated with the size of the grate. reported in ref 20. This can be resolved by solving extra
As in most investigations, in this test case heat loss is not transport equations for these species.
taken into account, so the equation for enthalpy was not used. In the next section, the FGM-PDF model is used for the
The turbulent database contains 51 points for mean mixture simulation of a biomass grate furnace. Because of the fact that
fraction, 40 for the mean progress variable, and 10 for each CO and H2 are not intermediates but main species in this case
variance. The points for the mean mixture fraction and progress (they are components of the fuel), it is expected that the
variable are equidistant while those for their variances are prediction of CO and H2 will be improved with respect to the
clustered in the region of small values. Sandia Flame D. This will be confirmed by the results presented
4.2. Simulation Results for Sandia Flame D. In Figure 1, in the following section.
the predictions given by the FGM-PDF model along the
(23) CHEM1D, A one-dimensional laminar flame code, Eindhoven 5. Application of FGM-PDF to a Biomass Grate Furnace
University of Technology; http://www.combustion.tue.nl/chem1d.
(24) Bowman, G.; Frenklach, M.; Gardiner, B.; Smith, G.; Serauskas, In cooperation with Graz University of Technology (TU-
B. GRI-Mech, Gas Research Institute, Chicago, IL; http://www.me.berke- Graz), the FGM-PDF model was applied to a pilot-scale low-
ley.edu/gri-mech/index.html.
(25) Jones, W. P.; Launder, B. E. Int. J. Heat Mass Transfer 1972, 15,
NOx biomass grate furnace for the prediction of combustion
301–314. and NOx emissions. In Figure 2 the general layout of such a
(26) Cashdollar, K. L.; Zlochower, I. A.; Green, G. M.; Thomas, R. A.; furnace is shown. Low-NOx grate furnaces are divided into a
Hertzberg, M. J. J. Loss PreVention Process Ind. 2000, 13, 327–340. primary and a secondary combustion zone by staged supply of
(27) Pierce, C. D.; Moin, P. J. J. Fluid Mech. 2004, 504, 73–97.
(28) van Kuijk, H. A. J. A.; van Oijen, J. A.; Bastiaans, R. J. M.; de the combustion air. The design of the primary combustion zone
Goey, L. P. H. Combust. Flame, in press. ensures sufficient residence time in an oxygen-lean atmosphere
1574 Energy & Fuels, Vol. 22, No. 3, 2008 Albrecht et al.

Figure 1. Axial profiles of (a) mean axial velocity and turbulent kinetic energy, (b) mean mixture fraction and mixture fraction variance, (c) mean
mass fractions of CH4 and O2, (d) mean mass fractions of CO2 and CO, (e) mean mass fractions of H2O and H2 and (f) mean temperature and mean
mass fraction of NO (lines, predictions; symbols, measurements).

zone is designed as an oxygen-rich zone to provide for flue gas


burnout. The staged flue gas recirculation is introduced for
temperature control in the furnace in order to prevent ash
slagging or sintering as well as for enhancement of turbulent
mixing, resulting in a faster CO burnout and more efficient NOx
reduction. The biomass fed into the furnace is transported by
the grate from left to right. The primary air as well as a part of
the total amount of the recirculated flue gas is blown from
below. The fuel dries and undergoes gasification/combustion
reactions on the grate. As a result, a gaseous mixture of
gasification/combustion products and air is formed just above
the grate. This mixture starts to burn in the primary combustion
zone and is then mixed with the rest of the recirculated flue gas
and secondary air (see Figure 2) injected through the
nozzles. After leaving the secondary combustion zone, the com-
bustion products exit the furnace and than pass through the
boiler.
Figure 2. Schematic of a biomass grate furnace. Courtesy of TU-Graz. In contrast to the simulation of Sandia Flame D, here the
enthalpy is required. The combustion inside the furnace is
in order to reduce nitrogen oxides formed during the conversion considered adiabatic (no heat loss to the walls), and therefore
of the solid biomass fuel on the grate. The secondary combustion the transport equation of the enthalpy has no source term.
Flamelet-PDF Model for Biomass Combustion Energy & Fuels, Vol. 22, No. 3, 2008 1575

Table 1. Fuel Composition and Operating Parameters for the


Furnace
quantity value operating parameter value
thermal load (kWth) 500 recirculation ratioa 0.37
water content (wt % db) 8 primary air ratiob 0.60
C content (wt % db) 49.5 grate effective air ratioc 0.78
H content (wt % db) 5.6 effective air ratiod 0.95
O content (wt % db) 40.1 total air ratioe 1.60
N content (wt % db) 0.27
ash content (wt % db) 4.53
LHV (MJ/kg wb) 16.73
a Mass of flue gas recirculated/mass of total flue gas in the furnace.
b Based on the total amount of primary air. c Based on the total amount
of primary air and recirculated flue gas supplied below the grate.
d Based on the total amount of primary air and total amount of

recirculated flue gas. e Based on the total amount of combustion air Figure 3. Comparison of new (solid line) and original (dashed line)
injected. CO mass fraction profiles on the grate: x is the distance along the grate
and L is the total length of the grate (L ) 2.84 m).
Table 2. Composition of Fuel and Oxidizer Streams
Table 3. Boundary Conditions for Gas Recirculation and
species (mass fractions) fuel oxidizer Secondary Air Inlets
CO 0.100 - EDC FGM-PDF
H2 0.005 -
CH4 0.010 - recirculation gases
CO2 0.200 - YO2 0.079 -
H2O 0.095 - YCO2 0.174 0.174
NO 5 × 10-5 - YH 2 O 0.057 -
NH3 3 × 10-4 - YN2 0.690 -
N2 0.590 0.767 temperature (K) 473 -
O2 - 0.233 velocity (m/s) 18.814 18.814
temperature (K) 1386 361 mixture fraction - 0.456
stoichiometric mixture fraction 0.63 enthalpy (kJ/kg) - -2131a
turbulence intensity (%) 10 10
hydraulic diameter (m) 0.034 0.034
However, the recirculation and addition of secondary air results secondary air
in changing enthalpy boundary conditions at the inflows. Thus, YO2 0.233 -
the transport equation of H resembles the mixture fraction, but YCO2 0 0
YH 2 O 0 0
it has different boundary conditions. It should be noted that the YN 2 0.767 -
combustion inside the furnace is in fact not adiabatic, heat being temperature (K) 298 -
lost through radiation. Here radiation was not considered because velocity (m/s) 18.783 18.783
the aim of the investigation was to compare the FGM-PDF mixture fraction - 0
flamelet model with the EDC model rather than to reproduce enthalpy (kJ/kg) - 0
turbulence intensity (%) 10 10
the exact conditions in the furnace. To keep the number of hydraulic diameter (m) 0.034 0.034
control variables to a minimum, the variance of the progress a The negative value of the total enthalpy is caused by the negative
variable was removed. This is justified by the expected modest value of the enthalpy of formation of products.
influence of the variance of the progress variable, as observed
in case of Sandia Flame D (as shown in ref 6).
N. Therefore, in principle, only four transport equations are
5.1. Mixture Fraction Based Grate Profiles. Grate profiles
needed to describe the mixing process. Additionally, since the
of species, temperature, and velocity at the top of the biomass
element mass fractions sum up to unity, three equations are
layer are used as boundary conditions of the gas phase turbulent
required. In the present method, in fact we assume that there is
combustion problem. The measurements are founded on realistic
correlation in the fuel and in the oxidizer; i.e., ZH is coupled to
solid biomass fuel composition and operation parameters (see
ZC, and ZO is coupled with ZN. This removes two equations, so
Table 1) and have been measured by TU-Graz. An empirical
that only a single mixture fraction would be sufficient.
model for fixed bed combustion of solid biomass fuels developed
Now, new approximated grate profiles of the species can be
in ref 2 and extended in order to allow for release of the NOx
generated, assuming that the species Yi present on the grate result
precursors (NO and NH3 as well as HCN in small amounts1)
from the mixing of the new defined fuel and oxidizer, in terms
was applied for the calculation of the composition, temperature,
of Z.
and velocity of the flue gas released as a function of the position
on the surface of the fuel bed. ) Yox
Ynew i + Z(Yi - Yi )
fu ox
i (17)
Based on the species mass fractions along the grate, provided
by TU-Graz, the element mass fractions Zj are calculated. From where Yi stands for CO, H2, CH4, CO2, H2 O, NO, and NH3.
that the mixture fraction Z is calculated with eq 1. Z ) 1 (fuel) As an example, the comparison of the original and new profile
is defined at the position on the grate where the combined of CO mass fraction is given in Figure 3. It can be noticed that
element mass fraction (first term in the numerator of eq 1) is the new profile does not match always with the original one.
maximum. The oxidizer is defined as air (Z ) 0). The This is caused by the fact that the Zj present on the fuel bed
composition and temperature of the fuel (Z ) 1) and oxidizer have slightly different profiles from each other and as a result
(Z ) 0) as well as the stoichiometric mixture fraction are shown they cannot exactly be described by a single mixture fraction,
in Table 2. but the correspondence is good enough. It should be noted that
From a fundamental point of view in combustion of hydro- it is not the scope of the present investigation to achieve a very
carbon with air there are four elements present, C, H, O, and accurate description of the species released from the biomass
1576 Energy & Fuels, Vol. 22, No. 3, 2008 Albrecht et al.

Figure 4. Comparison of 2D biomass grate furnace profiles predictions with FGM-PDF model (left) and EDC model (right).

bed, but rather to compare the FGM-PDF model with the EDC boundary conditions for the turbulence quantities were 30% for
model, for a specific choice of boundary conditions. The new the turbulence intensity and 1.5 m for the hydraulic diameter.
profiles based on the single mixture fraction approach are The boundary conditions used in EDC and FGM-PDF
considered a good choice, since Yi ≈ Ynew i . A comparison of simulations for the gas recirculation and secondary air inlets
the results achieved with the EDC model for the original and are given in Table 3. In case of the FGM-PDF calculations,
new profiles is presented at the end of section 5.4. The choice boundary conditions for the variance of the mixture fraction
of the EDC as reference for comparison with FGM-PDF is have also been set. Thus, at the grate, a profile was specified
justified by the fact that EDC is the only turbulent combustion for the variance of the mixture fraction. This profile has the
model standard available which can handle variable fuel inlet shape of the maximum possible value of the variance of the
composition profiles together with detailed reaction mechanisms. mixture fraction
5.2. Boundary Conditions. Based on the geometry of the ˜2 ) Z̃(1 - Z̃)
low-NOx pilot-scale furnace depicted in Figure 2, a 2D Z′′ max (18)
computational domain was constructed. The unstructured mesh Due to the fact that the values given by eq 18 are unrealistically
has 120 000 cells with a minimum edge size of 1 mm around large, the magnitude of the variance of Z was taken equal to
the inlets and a maximum edge size of 20 mm. At the grate, ˜
20% of Z′′ 2
max . This magnitude ensures that at the grate the
the boundary conditions for the EDC simulations consisted of creation and destruction of source terms in the transport equation
profiles for species mass fractions, temperature, and velocity. of the variance of mixture fraction (see eq 15) are balancing
The species and temperature have been calculated by using eq each other. As for the other two inlets (gas recirculation and
17, while the velocity has been taken directly from the data secondary air), the variance of the mixture fraction was set to
provided by TU-Graz. As for FGM-PDF simulations, the profiles zero.
needed are the mixture fraction, progress variable, enthalpy 5.3. Manifold Construction. As in the case of Sandia Flame
variable, and velocity. The Z profile is calculated with eq 1, D, a flow solver with the realizable k-ε turbulence model was
while the profiles of Y and H are calculated with eq 17. employed for a biomass grate furnace simulations. The chem-
Furthermore, for both EDC and FGM-PDF simulations, the istry was described by the GRI-Mech2.11 detailed mechanism.24
Flamelet-PDF Model for Biomass Combustion Energy & Fuels, Vol. 22, No. 3, 2008 1577

Figure 5. Comparison of 2D biomass grate furnace profiles predictions with FGM-PDF model (left) and EDC model (right).

This mechanism was selected because it is the most detailed and the secondary air enters the furnace through 4 tiny holes in
reaction mechanism for hydrocarbon combustion with less than a single row, after the first turn of the gases in the furnace.
50 species (the maximum number of species that can be used In Figure 4a, at the bottom of the furnace, the distribution of
by the EDC model in the flow solver). The laminar database the mixture fraction on the grate can be seen. The mixture
was constructed from 41 × 41 (41 Z values for each of the 41 fraction is 1 somewhere in the middle of the grate, where only
h levels) premixed flamelets with mixture fractions covering gaseous fuel is present, and 0 at the sides of the grate, where
the flammability range from Z ) 0.4 to Z ) 0.8 and scaled only air is present. In between, there are mixtures of fuel and
enthalpy variable from 0 (products at ambient temperature) to air with mixture fractions between 0 and 1. This distribution of
1 (adiabatic). The flamelets and their flammability limits were the mixture fraction on the grate corresponds to the different
calculated with the laminar flame code CHEM1D.23 Outside degrees of biomass conversion to gasification/combustion
the flammability range, all variables have been interpolated products along the grate when blowing air from below. Later
between the values at the flammability limits and at the inlets on, the streams with different mixture fractions coming from
(Z ) 0 and Z ) 1). The turbulent database contains 41 × 41 × the grate mix with each other and start to burn, leading to a
21 × 11 points corresponding to the mean mixture fraction, more homogeneous mixture with rising temperature (see also
mean enthalpy, mean progress variable, and mixture fraction Figure 4b). The recirculated gases are than mixed with the main
variance, respectively. The points for the mean mixture fraction, flow. As a result, the temperature decreases (locally) and the
enthalpy, and progress variable are equidistant while those for mixture is diluted. After the first turn in the furnace, the main
the mixture fraction variance are clustered in the region of small flow meets the secondary air flows. Thus, the rest of the fuel
values. present in the mixture is combusted and the mixture fraction
5.4. Results and Discussion. In Figures 4 and 5, a com- decreases, the mixture changing from slightly rich to lean (Zst
parison of the results obtained with the FGM-PDF model and ) 0.63). In Figures 4c and 5a, the mass fractions of CO2 and
the EDC model is given. It can be noticed that the simulation CO are shown. They represent the main product and fuel,
does not reproduce exactly the position of the flows entering respectively. The predictions given by FGM-PDF and EDC are
the furnace from figure 2. Thus, the recirculated gases are generally in good agreement, but the flame predicted by the
introduced through a small orifice placed in the vertical left wall FGM-PDF model seems to start sooner (see Figure 4, b and c).
1578 Energy & Fuels, Vol. 22, No. 3, 2008 Albrecht et al.

Figure 6. Comparison of 2D biomass grate furnace predictions with EDC model for the new (left) and original (right) fuel-bed profiles.

The higher concentration of CO2 in the flame formed around It should be mentioned that in Figures 4 and 5 the same color
the main fuel stream indicates that the FGM-PDF model predicts scale was used for EDC and FGM-PDF results corresponding
conditions closer to equilibrium in that region. to a certain quantity. The only noticeable differences between
Compared to EDC, conversion is faster in FGM because the maximum values predicted by the two models have been
source terms are overpredicted due to the low variance of observed for temperature and NO mass fraction. Hence, the
mixture fraction. It appears that the production of the variance maximum value of temperature was 1910 K for FGM-PDF as
of mixture fraction at the inflow boundary is relative small, compared to 1870 K for EDC (see Figure 4b) while the
due to mixture fraction gradients that are small and given by maximum value of NO mass fraction was 3.63 × 10 -4 for
the size of the grate. Furthermore, the composition is already FGM-PDF as compared to 2.22 × 10 -4 for EDC (see Figure
close to equilibrium at the grate. In FGM the small variances 5b).
are directly taken into account whereas in EDC it takes a little In conclusion, the predictions given by FGM-PDF model
longer due to the integration of the large consumption rate. Also agree generally well with the predictions of EDC model, while
the turbulence model is not ideal. the computational time is decreased from weeks (EDC) to hours
As for CO, the very good agreement between the two models (FGM-PDF). The only important difference between the models
proves that CO is indeed much better predicted by FGM-PDF is noticed for NO (63%), which is very sensitive to temperature
in the case of the biomass grate furnace than for Sandia Flame differences. This is caused by the fact that NO is a slow species
D, as argued before. In Figure 5, b and c, NO and NH3 mass and therefore it cannot be well predicted by databases based on
fraction are presented, respectively. In the conditions present steady-state flamelets. To improve the prediction of NO, an extra
in the furnace, it can be assumed that two mechanisms of NO transport equation should be solved. However, in this investiga-
formation are active: thermal (T > 1800 K) and fuel NO (from tion this step was not taken.
NH3). Thus, Figure 5, b and c, shows a correlation between the In order to evaluate the influence of the approximation made
consumption of NH3 and the production of NO, while Figures when constructing the new profiles (based on the mixture
5b and 4b indicate that the peak value of NO correspond to the fraction) on the predictions, a comparison with the original
top temperature locations. profiles is given in Figures 6 and 7. The same EDC model with
Flamelet-PDF Model for Biomass Combustion Energy & Fuels, Vol. 22, No. 3, 2008 1579

Figure 7. Comparison of 2D biomass grate furnace predictions with EDC model for the new (left) and original (right) fuel-bed profiles.

GRI-Mech2.11 mechanism was used in this comparison. All only one mixture fraction will not suffice anymore and more
the quantities are plotted using the same scale for the new and have to be added. For more complex fuels, a more extended
original profiles. The results presented in Figures 6 and 7 show kinetic scheme has to be employed to generate the manifold.
a reasonable agreement between the two simulations. Quanti- Besides this, attention should be given to the physical validation
tatively, however, significant differences are calculated for with measurements. Moreover, the correctness of the predictions
temperature and CO 2. Thus, the top temperature is 1870 K for associated with turbulence modeling is an important item for
the new profiles as compared with 1770 K for the original ones future research. Another important aspect is the conversion in
and the maximum mass fraction of CO2 is 0.237 for the new the fuel bed and the coupling with the present gas phase
profiles and 0.260 for the original profiles. These differences combustion model. At the moment our efforts are dedicated to
are caused by the fact that, when constructing the new profiles, the last issue (e.g., ref 28), although the other issues need
element and enthalpy conservation was not enforced separately
attention in the future as well.
on each of them but rather globally through the mixture fraction.
In the present study, radiation was not taken into account,
A better approximation of the profiles could be achieved by
although it is important for the coupling of the gas-phase
introducing an additional degree of freedom. This can be realized
turbulent combustion with the grate. For this, the prediction of
by adding a second mixture fraction. In this investigation,
however, we did not aim to a very accurate description of the radiant species is important. These include water and carbon
species released from the biomass bed, but rather to compare dioxide, which are easily predicted by both methods, and soot.
the FGM-PDF model with the EDC model, for a specific choice The application of FGM enables us to include soot quite easily
of boundary conditions. as well. In this case, the manifold should be created with a
In reality the “fuel” leaving the grate is not of constant detailed scheme including higher hydrocarbons. This is possible
composition and will contain higher hydrocarbons and HCN. with marginal additional computational costs, which have to
It is shown that the calculation results are sensitive to the fuel be invested in the creation of the manifold. However, the present
composition profile at the grate so one may expect the effect of EDC calculations in a 2D domain were already very expensive.
reducing the equation set to only one mixture fraction to be To take a large chemical mechanism into account with EDC is
significant. In order to handle nonconstant fuel composition, almost impossible. The coupling of the grate with the gas phase
1580 Energy & Fuels, Vol. 22, No. 3, 2008 Albrecht et al.

turbulent combustion for the radiant energy fluxes will be perature and CO2 predictions in which the spatial structure
implemented in new projects. remains the same. Furthermore, there is a relative large influence
of the amount of primary air, recirculation, and secondary air
6. Conclusions in the results. On top of this, the applied turbulence model will
not be able to provide perfect quantitative results. Thus, the
A new FGM-PDF procedure was applied to model combus- present approach is a good approach to guide the design,
tion in a grate furnace. This FGM-PDF combustion model was recirculation, and staging strategy with respect to emissions.
validated for a piloted diffusion turbulent flame, the Sandia In conclusion, this paper presents the use of a flamelet model
Flame D. The comparison of the predictions for the validation (FGM-PDF) for a very complex industrial application: a biomass
case leads to the following conclusions. The FGM-PDF model grate furnace. The FGM-PDF model produces comparable
gives generally good agreement of predictions with measure- results with EDC, using a detailed chemical description of the
ments. The overprediction of CO and H2 in the rich domain combustion system (as in the EDC) but at much lower
indicates that these intermediates are in reality determined by calculation time and memory usage. Thus, the calculation time
the transport (diffusion) rather than by the chemical composition with GRI-Mech2.11 was reduced from weeks (EDC) to hours
corresponding to the local combination of mixture fraction and (FGM-PDF). Furthermore, FGM-PDF is based on more physical
progress variable. A similar finding was also reported in ref grounds than EDC with respect to the modeling of turbulent-
20. Therefore, the predictions of CO and H2 can be improved chemistry interaction thanks to the flamelet structure integrated
by solving extra transport equations for these species. in its PDF approach.
The FGM-PDF model was successfully applied to a biomass For more quantitative predictions, reproducible measurements
combustion grate furnace. Realistic species profiles on the grate must be used to validate the method. For the simulation of fuels
were assumed. An enthalpy variable had to be added to take with changing composition, more mixture fractions should be
into account the influence of the recirculation gases and taken into account. Turbulence modeling can be improved by
secondary air on the combustion. The results of FGM-PDF have switching to Reynolds stress models in ensemble averaging or
been compared with the EDC model. Overall good agreement by introducing large eddy simulations. It is questionable whether
was found between the two models. The main difference is that these improvements pay off in the design process. At the
FGM-PDF predicts a faster conversion of the fuel bed species moment, research is conducted in the description of conversion
into products. Furthermore, slightly higher temperatures ac- in the fuel bed (e.g., ref 28) and coupling of the bed with the
companied by higher NO levels are predicted by FGM-PDF. present gas phase combustion model.
From the paper, it is clear that the usability of the method is
very high due to high accuracy and small computational times. Acknowledgment. The present research was funded by EU
This is true if the fuel has a constant composition. When the within the European project OPTICOMB, NNE5-2001-000639. The
fuel has local changes in composition then more (element authors thank Robert Scharler and Ingwald Obernberger from Graz
fractions) conservation equations have to be applied. It is shown University of Technology for their valuable contribution to this
work.
that deviations of constant composition change the result.
However, the largest absolute differences appear in the tem- EF7007562

View publication stats

You might also like