Fuel combustion behaviour analysis combining furnace measurements - - PowerPoint PPT Presentation

Fuel combustion behaviour analysis combining furnace measurements with CFD modelling Jyväskylä, 23rd September 2014

Sirpa Kallio VTT Technical Research Centre of Finland

Contributors: Perttu Jukola, Marko Huttunen, Juho Peltola, Timo Niemi, Lars Kjäldman

2 23/09/2014 2

Contents

• Background on CFD approaches for FB’s
• CFD modeling activities at VTT
• CFD modeling of bubbling bed combustion at VTT
• CFD modeling of CFB combustion at VTT
• CFD modeling in combination with furnace

measurements

3

Background: CFD modeling of gas-solid flow

4 23/09/2014 4

What is CFD and what is it used for?

• Computational fluid dynamics (CFD) aims at

describing the processes in 2D or 3D geometries, either as a function of time or by means of a time- averaged/steady-state description

• Several methods available aimed at different

applications

• CFD is useful when there are large local variations

in the process, mixing is important and/or the geometry plays a central role

5 23/09/2014 5

What do we mean by a CFD simulation

Typically 3D Finite Volume Method (FVM) simulations:

1. Create a CAD model of the geometry 2. Split the 3D fluid volume into to small control volumes 3. Solve balance equations for desired properties in each of these control volumes

• Momentum, mass, energy….

4. May include tracking of particles, droplets, bubbles etc. 5. Computational time: hours, days, weeks, months…

6 23/09/2014 6

What do we get from a CFD simulation

Full 3D fields of:

1. Gas and solids velocities 2. Pressure 3. Volume fraction 4. Temperature 5. Turbulence properties (mixing) 6. Species concentrations 7. Reactions rates 8. Local size distribution 9. Fluctuations of all the above

• 10. Information on fouling, erosion etc. with specific submodels

7 23/09/2014 7

Alternative approaches to CFD modelling of Fluidization

1. DEM, Discrete Element Method 2. MP-PIC, Multi-Phase Particle-In-Cell 3. Transient Euler-Euler 4. Time-averaged Euler-Euler

Small scale Large scale

8 23/09/2014 8

Alternative approaches: DEM

• Individual particles are

tracked and collision between the particles are calculated explicitly

• Soft- or hard-sphere models

for the collisions + Accurate, simple closure models

• Computationally very, very demanding
• Large, soft and light particles => less effort
• Small, hard and heavy particles => more effort
• Suitable for studies of small details (valves, injectors, guide

vanes) and gas-solid-solid interaction models

9 23/09/2014 9

Alternative approaches: MP-PIC

• Solid phase is modelled as parcels of particles that are

tracked.

• Collisions are not calculated explicitly, but modelled

with a statistical model + Usable in reasonably large applications (pilot scale) + Easy description of solids diameter distribution and its evolution

• Reasonably fine mesh resolution or sub-grid-closures are

needed for accurate results

• Solid-solid interaction: Currently available tools do not

produce good results in dense conditions (BFB)

10 23/09/2014 10

Alternative approaches: Transient Euler-Euler

• The most commonly used approach in fluidization simulations.
• Both phases are modelled as continuous fluids
• Can be coupled with Lagrangian fuel particles
• Statistical interaction models

+ Most mature approach: interaction models, chemistry etc. + Usable in reasonably large applications (pilot scale) + Good results in dense conditions

• Difficult to describe a solids diameter distribution
• Reasonably fine mesh resolution or sub-grid-closures are

needed for accurate results

11 23/09/2014 11

Alternative approaches: Time-averaged Euler- Euler

• TASIF: A time-averaged version of the commonly used Euler-

Euler model

• The required closure models have been developed at VTT
• Eulerian gas and bed mass phases + Lagrangian fuel

particles + Drastically reduced the computational effort in large applications

+ Suitable for industrial scale applications + Allows for more complex process models

• More reliant on the quality of the closure models
• Fluctuations are modelled, not resolved => their effects on the process also

have to be modelled.

Avtivities of the CFD modeling team at VTT

13 23/09/2014 13

Computational Fluid Dynamics (CFD) at VTT

• Applied since 1982
• CFD Modelling Team: 15 researchers
• Research & application areas
• Combustion & emissions, gasification, fast

pyrolysis

• Multiphase flows, rotating machinery, fluid-

structure interactions, nuclear safety analysis, molecular modelling

• Model development and testing
• Applied research
• Investigation of practical application cases
• Utilize simulation to understand process

behaviour and as a design tool

• Co-operation within VTT and with several

companies and universities

14 23/09/2014 14

Combustion & emissions

• CFD application areas…
• Pulverized fuel combustion
• Bubbling and circulating fluidised bed

combustion (BFB & CFB)

• Grate fired combustion
• inc. spreader stoker fired units
• Recovery boilers & lime kilns
• Gaseous & liquid fuel flames
• Common issues in CFD studies
• NOx emissions (e.g. IED) & emission

reduction (low NOx techniques + thermal DeNOx / SNCR)

• Heat transfer
• Furnace availability
• Slagging, fouling, corrosion (tendencies)
• Co-firing of coal / biomass / peat / etc.
• Fluidization issues (CFB)

15

CFD modeling of bubbling fluidized bed combustion at VTT

16 23/09/2014 16

Methods for Bubbling Fluidized Bed Boilers

• Dense bottom bed described by

simple empirical balance models

• CFD used for the freeboard
• Lagrangian particle tracking for

fuel particles

• CFD modeling for the bottom

bed possible but nor currently linked to boiler simulations

17 23/09/2014 17

Example 1: An Air Distribution Study of a Bubbling Fluidized Bed Boiler

Jukola, Huttunen, Dernjatin & Heikkilä: New methods for NOx emission reduction in fluidized bed combustion: CFD modelling results from a stationary fluidised bed boiler, VGB Powertech Journal, 11 / 2013

• Furnace capacity: 175 Mwfuel
• fuel mixture: peat 30 %,

biomass 70 %

• Main topic: NOx formation
• effect of 2’ry air elevation
• low furnace air distribution
• Other topics
• burnout (indicator: CO at

furnace exit)

• upper furnace fouling

tendency (indicator: furnace exit gas temperature, FEGT)

18 23/09/2014 18

An Air Distribution Study of a Bubbling Fluidized Bed Boiler, cont.

Simultaneous decrease in exit gas T, CO and NOx when “optimizing” Stoichiometric Ratio in Zone I (SR1)

19 23/09/2014 19

REDUCE FURNACE SLAGGING Particle concentration near walls Old New

Example 2: Modelling of Combustion in BFB Furnaces: Slagging & Fouling Issues

REDUCE SUPERHEATER /REHEATER SLAGGING Combustion rate of char in freeboard Old New

(178 MWf; Bio/Peat = 45% / 55 % of Fuel Power) (294 MWf; Bio/Peat = 30% / 70 % of Fuel Power)

20

CFD modeling of circulating fluidized bed combustion at VTT

21 23/09/2014 21

Method developed at VTT: Steady-state CFD modelling

 

q q q q s qs s g gs q M q q q q q q q q q q q

p K p p

qs

" " ) 1 ( =

) 1 (

u u u u τ τ g U U           



                     



=

q q q

U    

Fast simulation method for large geometries. Time-averaged form of the continuity and momentum equations used in the transient simulations for phase q are: phase-weighted average velocity

q q

  / u U 

q q q

U u u   "

velocity fluctuation u instantaneous velocity

22 23/09/2014 22

Description of the CFD model: hydrodynamics

Gas and solids momentum

• Closures for drag, Reynolds stresses, volume fraction-pressure

correlation and solids pressure terms.

Turbulence, Reynolds stresses

• Transport equations are solved for the solids normal

components of velocity correlations.

• Gas phase stresses are calculated with algebraic correlations

from the solid phase stresses.

• Fluctuation time scales are obtained from algebraic correlations.
• Small scale and dilute suspension turbulence: dispersed k-ε model

with a modified turbulent viscosity

23 23/09/2014 23

Description of the CFD model: energy and material balance equations

Energy: Specific enthalphy equations for both phases

• Unisotropic diffusion coefficients approximated from Reynolds

stresses and time scales

• Phase interaction (Gunn, 1978), wall heat transfer based on: Vijay&

Reddy (2005)

Species equations for gas components

• Isotropic diffusion coefficient approximated from Reynolds stresses

and time scales.

• Gas reactions are assumed to be limited by mixing.
• Species: O2, N2, CO2, CO, H2O, H2 and CHxOy.
• Reactions:

1. CHxOy+(x/2+1-y)/2 O2 → CO + x/2 H2O 2. CO +0.5 O2 → CO2 3. H2 + 0.5 O2 → H2O

24 23/09/2014 24

Description of the CFD model: solid fuel

Fuel particles: modelled with a Lagrangian approach

• Coupled to velocities and velocity fluctuations of both phases.
• Recirculation of fuel particles.
• Closures:

i. Heat transfer: Palchonok (1998) ii. Drying: constant particle size iii. Devolatilization: Ross et al. (2000), single volatile species (CHxOy), constant particle size iv. Oxidation: shrinking particle model, internal and external mass transfer, chemical kinetic inside the char particle. Kinetic parameters based on the results of Konttinen et al. (2002).

25 23/09/2014 25

Example 1: Chalmers 12 MW boiler

• Based on Åmand et al. (1997)
• Coal 2: 8 MW, Polish coal.
• Flue gas recirculation with the

primary air.

• Only the riser section and short
• utlet and solids return sections

are in the computational domain.

• Recirculation of fuel and solids is

modelled with boundary conditions.

26 23/09/2014 26

Results: Volume fraction and mass weighted velocities

• Typical CFB velocity

and volume fraction fields.

• Volume fraction is

unrealistically low at the very top

27 23/09/2014 27

Results: Fuel particles, volatile release rate and char content

• A significant portion of

fuel recirculates: returned immediately in the simulation.

• Fuel particle residence

times up to one minute

• n one pass, or even

longer for the largest particles.

• Most of the particles

remain at the bottom, as evidenced char and volatiles release patterns.

28 23/09/2014 28

Results: O2, CO and H2O mole fractions, temperature

• High O2 concentration

near gas inlets, secondary air mixing is quite weak.

• H2O is mostly released

near the fuel inlet: Drying is a fast process compared to solids mixing.

• CO concentration is

high at the bottom.

• The unrealistically low

solid volume fraction at the top is reflected in the temperature field as too low temperatures at the very top.

29 23/09/2014 29

Results: Comparison to the experiment

Åmand et al. Simulated Solids circulation 29 kg/s 32 kg/s Char content at bottom 2.4-4.3 % 0.3 – 2.0 % Char content at outlet 0.5 % Char content in cyclone leg 1.0 %

30 23/09/2014 30

Example 2: A 135 MWe CFB boiler in Ruzhou, China

31 23/09/2014 31

Ruzhou CFB: solids content and gas velocitry

120 MWe load, asymmetric air & fuel distribution:

32 23/09/2014 32

Ruzhou CFB: temperature and oxygen distribution

120 MWe load, asymmetric air & fuel distribution:

33 23/09/2014 33

Ruzhou CFB: comparison to measurements

120 MWe load, asymmetric air & fuel distribution:

34 23/09/2014 34

Ruzhou CFB at two loads

120 Mwe load case: 80 Mwe load case:

Measured flue gas oxygen concentration 8.9%!

35

CFD modeling in combination with furnace measurements

36 23/09/2014 36

Why to combine CFD and experimental analyses

• Testing of new fuels and process conditions can become

expensive in large scale CFB units due to damages or unintended shutdowns.

• For that reason fuel tests are often first carried out in laboratory

and pilot scales.

• Extending the results directly to the furnace scale can be risky

since no simple scaling laws can be applied

• To reduce the uncertainties, CFD modeling can be used as a

tool to evaluate the scaling effects.

• CFD models need as inputs information on fuel

characteristics.

37 23/09/2014 37

Example of extending pilot scale test results to large scale - Pilot scale experiments at two loads

Fuel analysis: Inflow rates of fuel and gases, temperatures of gas flows and bed inventory in the two experiments:

• CFB pilot at VTT, Jyväskylä
• Cylindrical 8.3 m high riser
• Diameter 0.167 m.
• Secondary/tertiary air at 0.675 m, 1 m

and 1.37 m heights.

• Fuel feed at 0.365 m height.

38 23/09/2014 38

Example of extending pilot scale test results to large scale - Simulation of the CFB pilot

Case 2 low load: Case 1 high load:

39 23/09/2014 39

Example of extending pilot scale test results to large scale - Simulation of a larger CFB

Geometry of Chalmers 12 MW boiler: height 14 m, cross-section 1.4 m X 1.6 m

40 23/09/2014 40

Example of extending pilot scale test results to large scale - Results at the higher load

41 23/09/2014 41

Example of extending pilot scale test results to large scale - Results at the lower load

42 23/09/2014 42

Example of extending pilot scale test results to large scale - Comparison

Load: high low high low

O2 CO

43 23/09/2014 43

Conclusions

• Fuel characterization is needed to produce input data for CFD
• To extend the results from pilot scale combustion experiments to

the conditions in a larger boiler the geometrical effects need to be accounted for.

• CFD simulation can assist in this evaluation.
• Many of the characteristics of the combustion process that are
• bserved in the small scale are also seen in the industrial boiler,

but in the larger geometry there is a larger lateral variation of combustion conditions.

• Furnace measurements help to evaluate the accuracy of the

models.

TECHNOLOGY FOR BUSINESS