COMBUSTION MODELING AND
OPTIMIZATION
Application of Computational Fluid Dynamics and Systematic
Optimization in Emission Reduction of Biomass Fluidized Bed Boiler
Ari Saario, Antti Oksanen
Tampere University of Technology, Finland
Matti Ylitalo, Juha Roppo
Metso Power Oy, Finland
Acknowledgments:
Combustion Modeling and Optimization
Combustion Modeling and Optimization
Fluidized Bed Boiler Modeling
•
Bubbling fluidized bed boiler burning biomass studied using computational fluid dynamics (CFD)
•
More stringent nitric oxide emission limits
⇒ new and improved design tools required
•
Nitric oxide reduction by ammonia injection (SNCR process): two-step global reaction mechanism
NH
3
+ O
2
→
NO + H
2
O +
1
2
H
2
NH
3
+ NO
→
N
2
+ H
2
O +
1
2
H
2 too high temperature
→ more NO produced too low temperature
→
NH
3 passes unreacted
(1)
(2)
Combustion Modeling and Optimization
Combustion Modeling and Optimization
Unstructured (tetrahedral) grids
0.2
0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
Unstructured grid I (73 994 cells)
Unstructured grid II (74 306 cells)
Unstructured grid III (74 526 cells)
Unstructured grid IV (74 006 cells)
Dense unstructured grid (2 067 513 cells)
Very dense structured grid (4 264 650 cells)
0.05
0.1
0.15
r / x
0.2
0.25
0.3
Combustion Modeling and Optimization
Structured grids - local grid refinement
4 x 10
−3
3
Baseline grid (102 524 cells)
Once refined grid (105 968 cells)
Twice refined grid (109 412 cells)
Three times refined grid (112 856 cells)
Dense grid (2 067 424 cells)
Very dense grid (4 264 650 cells)
2
1
0 0.05
0.1
0.15
r / x
0.2
0.25
0.3
Combustion Modeling and Optimization
Temperature distribution
Side
Left C
Side
Left B
Side
Left A
6
5
1100
1150
1200
1250
1100
4
1100
1200
1300
3
1350 1200
1150
2
1250 1300
1350
1150
1100
1000
1
1200
1300
1100
8 7
1350
6
1300
1200 1100
5 4
Width (m)
3
Front Left
1000
2
Front Right
1
Combustion Modeling and Optimization
Side
Right C
Side
Right B
Side
Right A
Combustion Modeling and Optimization
Optimization problem
Find x = [x
1
, x
2
, . . . , x n
] T which minimizes f ( x ) = [f
1
( x ), f
2
( x ), . . . , f k
( x )] T subject to g j
( x ) ≤ 0 j = 1, 2, . . . , p h j
( x ) = 0, j = 1, 2, . . . , q x is design variable vector f ( x ) is objective function vector g j
( x ) and h j
( x ) are inequality and equality constraints
Combustion Modeling and Optimization
Optimization - CFD interaction
OPTIMIZATION
ALGORITHM
Design
Variables
@
@
R
CFD
@
@
Objective
Function
Combustion Modeling and Optimization
Why optimize ?
Exhaustive search over the whole design space:
Assume nine design variables (n = 9) and
64 design points per each variable (q i
= 64)
Number of required function evaluations, N:
N =
Q n i=1 q i
= 1.8e+16
Assume further that one CFD evaluation takes one hour −→ we need 2.1e+12 years to complete the optimization !
Combustion Modeling and Optimization
Optimization methods
A great number of possibilities exist
Genetic algorithm:
• population-based method
• survival of the fittest: selection - crossover - mutation
• no need for gradients
• not dependent on starting point (global method)
• requires many function evaluations
Powell’s method:
• point by point iterative method
• no need for gradients
• dependent on starting point (local method)
Combustion Modeling and Optimization
Genetic algorithm principle
START g = 0
Generate initial population P(g) with s individuals
[s = 10]
Evaluate fitness of each individual in P(g)
[CFD]
Replace j worst
individuals in P(g) by j best individuals in
P(g-1) [j = 1]
Apply mutation with probability p
m
Apply crossover with probability p
c
[uniform crossover,
STOP
YES
Stopping criterion met ?
[g > 48]
NO g = g + 1
Apply selection
[tournament of two individuals]
Combustion Modeling and Optimization
Minimum NO emission
85
← ←
Baseline injection case
80
Pop. 10 (a)
Pop. 10 (b)
Pop. 10 (c)
Pop. 10 (d)
Pop. 10 (e)
Pop. 50
75
70
65
60
0 100 200 300 400 500 600 700
CFD evaluations
Combustion Modeling and Optimization
Pareto points in objective space
110
100
90
80
70
60
50
0 10 20 30 40 50
Ammonia concentration (ppm vol
)
Baseline case
SOO (NO)
SOO (NH3)
MOO
Combustion Modeling and Optimization
Optimum design variables
5
4
Side
Left C
3
2
1
Rear
Front
Left
Side
Left A
Side
Left B
Injection
Front
Right
Side
Right A
Side
Right B
Side
Right C
Combustion Modeling and Optimization
Combustion Modeling and Optimization
Conclusions
• Careful CFD modeling study may be able to give information about trends in boiler emissions
• Fast-growing computing power → use of systematic optimization with CFD offers great possibilities in near future.
• Combination of CFD and optimization gives high quality results in efficient manner.
Combustion Modeling and Optimization
Conclusions - CFD Modeling
Challenges / problems in fluidized bed boiler
CFD modeling
• Description of bubbling bed and fuel feed
• Uncertainties in CFD model boundary conditions
• Coarse computational grids
• Shortage of experimental data for model validation
• Nitrogen chemistry and turbulence - chemistry interaction modeling
• Radiative heat transfer modeling
Combustion Modeling and Optimization
Conclusions - Optimization
Challenges / problems in optimization
• Great number of CFD evaluations required
• Definition of objectives and design variables
• Finding best optimization method and its parameters for specific case
• Multi-objective optimization
• Constrained optimization
Combustion Modeling and Optimization
Selected publications
SAARIO, A., OKSANEN, A. YLITALO, M. (2007). Application of
Computational Fluid Dynamics and Multi-Objective Optimization in Design of
Low Emission Combustion Equipment. 15 th IFRF Member’s Conference, Pisa,
Italy.
to appear .
SAARIO, A., OKSANEN, A. (2007). Detailed Study on the Effect of
Computational Grid in Industrial-Scale Boiler Modeling.
Submitted .
SAARIO, A., OKSANEN, A., YLITALO, M. (2006). Combination of Genetic
Algorithm and Computational Fluid Dynamics in Combustion Process Emission
Minimisation.
Combustion Theory and Modelling , Vol. 10, pages 1037 - 1047.
SAARIO, A., OKSANEN, A., YLITALO, M. (2006). Nitric Oxide Emission
Modeling in Bubbling Fluidized Bed Furnace for Biomass.
International Journal on Energy for a Clean Environment , Vol. 7, pages 1 - 22.
SAARIO, A., OKSANEN, A., MAKIRANTA, R., YLITALO, M. (2005).
Optimization of Selective Non Catalytic Reduction Process Using Computational
Fluid Dynamics and Genetic Algorithms. Swedish - Finnish Flame Days, Boras,
Sweden.
Combustion Modeling and Optimization