Professional Documents
Culture Documents
Advanced Design Optimization of Combustion Equipment For Biomass
Advanced Design Optimization of Combustion Equipment For Biomass
Renewable Energy
journal homepage: www.elsevier.com/locate/renene
a r t i c l e i n f o a b s t r a c t
Article history: Design: of engineered combustion equipment normally involves laborious “build and try” designs to
Received 18 February 2019 identify the best possible configuration. The number of design iterations can be reduced with engi-
Received in revised form neering experience of what might work. The expensive cut-and-try approach can be improved using
24 June 2019
computational aided engineering tools coupled with optimization techniques to find the optimal design.
Accepted 15 July 2019
Available online 16 July 2019
For example, the “best” air duct configuration with the lowest pressure loss and smallest fan size for an
air-fed biomass gasifier may take several weeks using the standard computational fluid dynamics (CFD)
“cut and try” approach. Alternatively, coupling an efficient design optimization algorithm with an
Keywords:
Biomass combustion
existing CFD model can reduce the time to find the best design by more than 50% and can allow the
Computational fluid dynamics engineer to examine more design options than possible using the “cut-and-try” approach. Combining an
Reduced design cycle time efficient optimization algorithm with an existing CFD model of a biomass gasifier to find the “optimal”
Advanced burner design design is the focus of this work. Shape optimization has been performed by combining the optimization
Design optimization tool Sculptor® with the commercial CFD code STARCCMþ. This work illustrates how the “linked”
approach is used to examine design factors to optimize an entrained flow biomass gasifier to improve
overall system performance in a methodical comprehensive fashion.
© 2019 Elsevier Ltd. All rights reserved.
https://doi.org/10.1016/j.renene.2019.07.074
0960-1481/© 2019 Elsevier Ltd. All rights reserved.
1598 J.D. Smith et al. / Renewable Energy 145 (2020) 1597e1607
Fig. 1. Typical Biomass/coal burners: a) Low-Emission Scroll-type Biomass/Coal burner (Coen Sales brochure), b) BWE Bio-dust burner equipped with a gas lance and/or an oil lance
(BWE Sales brochure).
the standard “model-build-and-try” approach commonly used devices. To support these goals, computational and numerical
with CFD analysis. techniques have been applied to optimize a biomass fired burner.
The use of detailed chemistry schemes for gas phase reactions is
well established for the combustion of hydrocarbons such as
2. Biomass combustion
methane, acetylene, ethane, propane and butane. The most widely
used detailed mechanism for the combustion of methane and
Coal has been a primary choice amongst fossil fuels, which are a
methane hydrogen mixtures is GRI-Mech 3.0 (developed by the Gas
primary source of copious amounts of energy. Owing to combus-
Research Institute) consisting of 53 species and 325 chemical
tible content and a high calorific value (HHV: 25e30 MJ/kg) [5], coal
equations showing the formation and destruction of intermediate
fired power plants have previously provided a reliable energy
radicals/species [9]. Mechanisms such as the LLNL (Lawrence Liv-
supply to society and industry. With depleting coal and other fossil
ermore National Lab) developed C1eC4 oxidation/combustion
fuel reserves, it is imperative to switch to renewable energy sources
mechanism also includes a general NOx formation mechanism.
that may continue to provide energy of a similar magnitude by
These and other such detailed mechanisms are often available for
combustion processes. Biomass is a renewable fuel with compara-
use in ChemKin/CANTERA format which commercial CFD codes can
ble calorific value to coal (HHV: 10e20 MJ/kg) [5] which is generally
readily use. However, the inclusion of a large numbers of species
available and represents a prime alternative to coal as feedstock.
and chemical reactions makes their usage “as-is” prohibitively
With coal-like compositions, the goal of this work is to design and
expensive from a computation point of view for industrial com-
optimize systems for biomass firing to reduce society's dependence
bustion applications. A common workaround for this issue applied
on fossil fuels for power production.
to gas phase flows is the use of chemistry agglomeration or ISAT
Considering this goal, co-fired units have become more popular.
(In-Situ Adaptive Tabulation) methods to reduce the effective
When appropriate amounts of biomass and coal are mixed and co-
number of chemical species accounted for in the detailed com-
fed to an optimized burner, SOx, NOx, and greenhouse gas emis-
bustion model. Another option is the use of steady and unsteady
sions decrease [6]. Using demonstrated co-firing technology re-
laminar flamelet models and flamelet generated manifold models,
ported elsewhere [7], significant cost savings in power generation
which solve for species in the mixture fraction space to simulate gas
using biomass range from $80,000/year to $400,000/year units [8].
phase combustion. The current model incorporates heterogeneous
In addition, energy security is enhanced using a sustainable energy
(particle surface) reactions as well as gas phase reactions, and
of renewable fuel sources such as biomass to operate combustion
J.D. Smith et al. / Renewable Energy 145 (2020) 1597e1607 1599
Table 1 biomass volatiles typically include H2O, H2, CO, and N2. Since
Proximate and ultimate analyses of the woody biomass (pulverized wood). biomass has a higher oxygen content compared to coal, more CO is
Proximate analysis (wt%) generally produced in the gas phase during biomass devolatiliza-
Moisture 5.06
tion. Biomass char oxidation also generates more CO and H2 than
Volatile Matter 82.95 coal char does which provides more energy, which tends to sustain
Fixed Carbon 16.15 biomass devolatilization and consequently affects near combustion
Ash 0.90 near the burner. The homogeneous gas phase reaction mechanism
Ultimate analysis (wt%) with associated chemical kinetic parameters together with the
Carbon 48.6 heterogenous particle devolatilization and char oxidation reactions
Hydrogen 6.4 are now presented.
Oxygen 44.0
Nitrogen 0.1
Sulfur 0.9 2.1. The combustion model
Low heating value (MJ/kg) 18.6
⇔
LES- RANS models. For this study aimed at optimizing industrial
combustion equipment, the steady-state RANS technique was used CO þ H2 O CO2 þ H2 (4)
to reduce the computational effort required to run several design
Reaction rate constants for the first three homogeneous gas-
cases. This approach was justified because most industrial com-
phase reactions are denoted by k1, k2 , and k3 . The water gas shift
bustion equipment operate in a “quasi steady” condition plus
(WGS) reaction includes both a forward step (k4f ) and a reverse step
typical experimental data of species compositions, velocity profiles,
(k4r ). Heterogenous combustion includes particle devolatilization
and temperature profiles are collected as time averaged values.
and char oxidation described next.
Thus, this assumption is valid as long as the combustion equipment
The particle devolatilization model used in this work is based on
does not operate in a highly non-steady condition such as flame lift-
the Kobayashi et al. [11] “competing-rate” model. Overall weight
off, extinction, re-attachment or ignition. Since the present work
loss is determined by:
considers optimizing industrial biomass fired burners operating in
a “quasi-steady” condition, the RANS based models are justified. ðt
Comparing numerous coal and biomass compositions shows dmv
¼ mp;0 ma ða1 R1 þ a2 R2 Þexp ðR1 þ R2 Þdt (5)
that coal generally has a higher C content and a lower O content dt
0
than biomass [5,7]. Accordingly, the contribution to the combustion
process from volatile matter is much higher for biomass combus- where R1 ¼ A1 expðE1 =RTp Þ and R2 ¼ A2 expðE2 =RTp Þ are the two
tion. The mass fraction of volatile content plays a pivotal role in the competing rates that control the devolatilization over different
devolatilization reactions, char oxidation reactions and gas phase temperature ranges. The yield factors a1 and a2 represent devola-
reactions. Devolatilization dominates the reaction process in the tilization at low and high temperatures, respectively. The yield
near burner region, where volatiles evolve from the biomass par- factors are feed specific and are set based on the fuel's proximate
ticles and react in the gas phase while the resulting char particles analysis (see Table 1).
oxidize away from the burner. Based on the fuel composition, The heterogeneous reactions included in the combustion model
Table 2
Reaction rates for bio-char oxidation and gas combustion reactions.
Ar (s1) Q 0 00
Reaction Rate Er nr ½C j;r ðhj;r þhj;r Þ A B Source
(J/kmol) j
k1 4.4 x 1011 1.25 x 108 0 [CH4]0.5[O2]1.25 11.5 2.75 Jones & Lindstedt [12]
k2 2.5 x 1016 1.68 x 108 1 [H2]0.5[O2]2.25[H2O]1 3.1 0.75 Jones & Lindstedt [12]
k3 3.16 x 1012 1.67 x 108 0 [CO]1.5[O2]0.25 2.1 0.53 Wu et al. [13]
k4f 5 x 1012 2.83 x 108 0 [CO]0.5[H2O]1 4 0.5 Callaghan [14]
k4r 9.5 x 1010 2.39 x 108 0 [CO2] [H2]0.5 4 0.5 Callaghan [14]
k5 0.052 6.1 x 107 0 e e e Chen et al. [15]
k6 0.0782 1.15 x 108 0 e e e Chen et al. [15]
k7 0.0732 1.125 x 108 0 e e e Chen et al. [15]
k8 1.2 x 105 7.53 x 107 0 e e e Govind & Shah [16]
1600 J.D. Smith et al. / Renewable Energy 145 (2020) 1597e1607
Biomass / (1-Yh) * CharhþYh * Volatile (High Temperature) (7) 2.3. Burner geometry
Table 4
Optimization parameters for initial designed experiment. Fig. 4. Mesh of a sphere embedded in an ASD grid with control points that can move
and deform the ASD grid deforming the sphere mesh.
Parameter Design Change Considered
Fig. 5. A 4 x 10 x 6 cylindrical ASD volume with control points arranged to change Fig. 7. Burner Step height modified by moving control points at outer edge in radial
burner step height, quarl slope, and quarl shape. direction by 20% resulting in a larger step height (original height ¼ 0.00783 m ¼>
modified height ¼ 0.009396 m). For a fixed reactor inlet diameter increasing the
burner step height also decreases the burner quarl slope.
The goal of the design optimization was to find the ‘best’ design
that would improve burner operation. In this case, the key is
identifying how the design changes affected the flare shape and
size of the Internal Recirculation Zone (IRZ). Ultimately, the opti-
mization will directly impact NOx reduction from the burner.
Table 6
Base case input data.
used for the base case conditions. These conditions were also used which supports early devolatilization compared to the other cases.
to analyze biomass combustion for the three cases with modified A closer look at Fig. 10d reveals a ‘second flame’, closer to the flow
geometries. Solutions for each of the cases were obtained on the axis, originating from the region where the high-temperature
same mesh for each case. devolatilization occurs, based on plots of volatile mass fraction.
CFD predictions for gas phase temperature, particle temperature This phenomenon shows only approximately a 5% reduction in
and velocity profiles in the reactor based on the input data listed in peak flame temperature compared to Fig. 10b and c. Also, the bulk
Table 6 are presented in Fig. 10, Fig. 11 and Fig. 12. gas temperature in Fig. 10d appears to be lower than 1400 K,
compared to the design other cases, which show bulk gas tem-
perature above 1600 K.
4.1. Analysis of optimized burner design Secondly, based on the streamlines of particles colored by
temperature from Fig. 11, it is seen in Fig. 11d that the recirculating
Examining Fig. 10 shows that the flame temperatures for all particle paths arise primarily from backflow along the wall from top
cases are similar. The base case peak flame temperature is to bottom. In comparison with the unmodified geometry in Fig. 11a,
approximately 400 K lower than the peak flame temperature for all there is a drastic reduction in the stagnancy of particle flow caused
other cases. Further, it is observed that the base case shown in by internal recirculation zones (IRZs). It is observed in both Fig. 11b
Fig. 10a and d have similar and shorter flame lengths compared to and c, that the IRZ seen in Fig. 11a on the right of the flame, gets
Fig. 10b and c, which appear to be longer with similar flame lengths. elongated and tilts towards the wall, at the top. This may be
Based on these observations, it is concluded that the 20% change in attributed to the gas temperature profiles seen in Fig. 10b and c,
step height produces a more stable flame anchored near the burner,
Fig. 10. Temperature profiles of biomass combustion: (a) Base case (b) Curved quarl case (c) Increased quarl slope case (d) Increased step height case.
J.D. Smith et al. / Renewable Energy 145 (2020) 1597e1607 1605
Fig. 11. Particle temperature streamlines: (a) Base case (b) Curved quarl case (c) Increased quarl slope case (d) Increased step height case.
where the flame is anchored further away from the burner than 5. Conclusions and recommendations
those in Fig. 10a and d. This is due to relatively inefficient particle-
air mixing, which delays particle heating and subsequent particle Combustion of biomass in the 300 kW BERL setup was modeled
devolatilization. using the eddy break-up partially-premixed combustion model.
This observation is supported from the velocity vector seen in The geometry was subjected to shape deformations to develop
Fig. 12. The base case in Fig. 12a shows an elliptical IRZ centered to three unique burner designs. Of the three, it appears the increase in
the right of the end of the flame. The geometry changes in Case b step height produced the most significant impact on the flame due
and Case c deform flow profiles as seen in the vector plots Fig. 12b to a reduced size and location of the IRZ. The flame also appeared to
and c which lead to elongation of these IRZs. As mentioned earlier, be anchored closer to the burner for the step change case which
the changes in IRZ shape and location result in inefficient mixing, impacted the region where the first biomass particle devolatiliza-
causing the flames (Fig. 12b and c) to be anchored higher than those tion occurs. This was attributed to increased mixing efficiency
in Fig. 12a and d, due to a delayed onset in particle devolatilization. created by the swirling air facilitated by a larger step size creating a
However, the velocity profile seen in vector plot Fig. 11d indicates a narrower primary air-particle stream. The longer step appeared to
breakdown of the velocity component tangential to the flame, create a smaller area for mixing particle laden air with the swirling
contributing to more uniform mixing in regions with large tem- secondary air. Although, devolatilization kinetics control the rate of
perature gradients, leading to a more homogeneous temperature volatile release which also impacts local flame temperature near
distribution. It is believed that this explains the uniform bulk the burner, the increased mixing leads to earlier devolatilization
temperature near the flame shown in Fig. 10d. which also anchors the flame closer to the burner. Consequently,
Based on these observations, it is suggested that the increased the dissipation of energy into the bulk gas produces a more uniform
step height before ignition provides a longer draft with a smaller temperature distribution near the flame.
cone angle, allowing the swirling air from the annular inlet to mix The observation of most important factor effects is based on
with a narrower stream of particle laden air. The particle visual inspections of the CFD predictions shown in the figures
streamlines from this region appear to anchor the flame in above. Specific factor effects can be quantified using traditional
Fig. 11d, producing a narrower flame than seen in Fig. 11b and c, statistical design of experiments (DOE) techniques. The purpose of
which is anchored closer to the burner and has a more uniform this paper has been to demonstrate the power of the linked
temperature distribution near the flame and a smaller IRZ in the ASD þ CFD methodology using Scupltor and StarCCM þ to examine
bulk fluid. the impact of burner design on mixing and combustion
1606 J.D. Smith et al. / Renewable Energy 145 (2020) 1597e1607
Fig. 12. Vectors of velocity showing recirculation zone changes to the right of the flame in biomass combustion from (a) base case (b) curved quarl (c) quarl slope increase from 200
to 250 (d) 20% increase in step height.
performance. It is recommended that additional work be Bowman, R. Hanson, S. Song, W. J. Gardiner, V. Lissianski and Z. Qin. [Online].
Available: http://www.me.berkeley.edu/gri_mec. [Accessed 15 June 2019].
completed using this linked approach in a traditional DOE to
[10] B. Franzelli, E. Riber, L. Gicquel, T. Poinsot, Large Eddy Simulation of com-
quantify factor effect. bustion instabilities in a lean partially premixed swirled flame, Combust.
Flame 159 (2) (2012) 621e637.
[11] H. Kobayashi, J. Howard, A. Sarofim, Coal devolatilization at high tempera-
Acknowledgment
tures, in: Sixteenth Symposium (International) on Combustion - the Com-
bustion Institute, The Combustion Institute, Pittsburgh, 1976, pp. 411e425.
Financial support for this work was provided by the Wayne and [12] W.P. Jones, R.P. Lindstedt, Global reaction schemes for hydrocarbon com-
Gayle Laufer Foundation and by Elevated Analytics. bustion, Combust. Flame 73 (1988) 233e249.
[13] Y. Wu, P. Smith, J. Zhang, J. Thornock, G. Yue, Effects of turbulent mixing and
controlling mechanisms in an entrained flow coal gasifier, Energy Fuel. 24 (2)
References (2010) 1170e1175.
[14] C. Callaghan, Kinetics and Catalysis of the Water-Gas-Shift Reaction: A
[1] P. Smith, W. Sowa, P. Hedman, Furnace design using comprehensive com- Microkinetic and Graph Theoretic Approach, Worcester Polytechnic Institute,
bustion models, Combust. Flame 79 (1990) 111e121. U.S.A., Worcester, Massachusetts, March 31, 2006.
[2] A.P. Horsman, Design Optimization of a Porous Radiant Burner, University of [15] C. Chen, M. Horio, T. Kojima, Numerical simulation of entrained flow coal
Waterloo, Waterloo, 2010. gasifiers Part I: modeling of coal gasification in an entrained flow gasifier,
[3] T. Goel, D. Dorney, R. Haftka, W. Shyy, Improving the hydrodynamic perfor- Chem. Eng. Sci. 55 (18) (2000) 3861e3874. September.
mance of diffuser vanes via shape optimization, Comput. Fluid 37 (2008) [16] R. Govind, J. Shah, Modeling and simulation of an entrained flow coal gasifier,
705e723. AIChE J. 30 (1) (1984) 79e92.
[4] S. Hajitaheri, Design Optimization and Combustion Simulation of Two Gaseous [17] A. Sayre, N. Lallemant, J. Dugue, R. Weber, Scaling Characteristics of the
and Liquid-Fired Combustors, University of Waterloo, Waterloo, Ontario, Aerodynamics and Low NOx Properties of Industrial Natural Gas Burners
Canada, 2012. Scaling 400 Study. Part 4. 300 kW BERL Test Results, Topical Report, October
[5] M. Tamura, W.L. Van de Kamp, Interim Progress Report on the Combustion 1992eDecember 1993, 1994, https://ntrl.ntis.gov/NTRL/dashboard/
Characterization of Substitute Fuels for Co-firing with Pulvverized Coals searchResults/titleDetail/PB95148524.xhtml.
during the Period 2000-2001," IFRF Doc N.F106/y/1, 2001. September. [18] C. Penny, STAR-CCM New Features List for Version 13.02, 2018 [Online].
[6] B. Moulton, Principles of burner design for biomass Co-firing, in: 34th Inter- Available: http://www.ata-plmsoftware.com/wp-content/uploads/2018/03/
national Technical Conference on Coal Utilization and Fuel Systems, Clear- STARCCM_New_Features_List-v13.02.EN_smaller.pdf. (Accessed 21 February
water, Florida, 2009. May 31. 2019).
[7] L. Baxter, Biomass-coal Co-combustion: opportunity for affordable renewable [19] J. Smith, V. Screedharan, A. Suo-Antilla, 30 Smith, J.D., V. Sreedharan, V. Rao,
energy, Fuel 84 (2005) 1295e1302. M. Landon, Z.P. Smith, Advanced design optimization of combustion equip-
[8] National Renewable Energy Laboratory, DOE/EE-0288: Federal Energy Man- ment using Sculptor® with CFD tools, in: American Flame Research Com-
agement Program: Biomass Cofiring in Coal-Fired Boilers, Department of En- mittees 2014 e Industrial Combustion Symposium, Hyatt Regency Hotel,
ergy, June, 2004. Houston, Texas, 2014. September 7-10.
[9] G. Smith, D. Golden, M. Frenklach, N. Moriarty, B. Eiteneer, M. Goldenberg, C. [20] J. Park, Optimal Latin-hypercube designs for computer experiments, J. Stat.
J.D. Smith et al. / Renewable Energy 145 (2020) 1597e1607 1607
Plan. Inference 39 (1) (April 1994) 95e111. phenomena among raw biomass, torrefied biomass and coal in an entrained-
[22] Wikipedia, Latin hypercube sampling [Online]. Available: https://en. flow reactor, Appl. Energy 112 (2013) 421e430. December.
wikipedia.org/wiki/Latin_hypercube_sampling. (Accessed 18 February 2019). [23] C. Chen, M. Horio, T. Kojima, Numerical simulation of entrained flow coal
gasifiers. Part I: modeling of coal gasification in an entrained flow gasifier,
Chem. Eng. Sci. 55 (2000) 3861e3874.
Further readings