modeling and process control of grate furnaces · modeling and process control of grate furnaces...

Workshop on

Modeling and process control of grate furnaces

Arranged by:

Arranged by:

Jaap Koppejan, TNO Science and Industry, Netherlands Sjaak van Loo, Chess, Netherlands

September 28, 2005 Hilton Hotel Innsbruck, Austria

2

ThermalNet/IEA Bioenergy/Opticomb workshop Modeling and process control of grate furnaces

September 28, Innsbruck, Austria

Table of contents

Programme........................................................................................................... 3

Report of the workshop ...................................................................................... 4

Annexes

Annex 1. Introduction, Sjaak van Loo

Annex 2. Combustion on a grate: dynamic modelling, process identification and process control Robert van Kessel, TNO, the Netherlands

Annex 3. Characterisation of N-release from a biomass fuel layer by pot furnace experiments and derivation of release functions Selma Zahirovic, Graz University of Technology, Austria

Annex 4. CFD modelling of NOx formation in biomass grate furnaces with detailed chemistry Selma Zahirovic, Graz University of Technology, Austriai

Annex 5. Biomass combustion on grates and NOx formation mechanisms Claes Tullin, SP, Sweden

3

Programme

30 September 2005, 08:30 – 11:00 Hilton Hotel Innsbruck, Austria

From Topic

8:30 Welcome and introduction Sjaak van Loo, CombNet Coordinator

8:40 Combustion on a grate: dynamic modelling, process identification and process control Robert van Kessel, TNO, the Netherlands

9:10 Characterisation of N-release from a biomass fuel layer by pot furnace experiments and derivation of release functions Selma Zahirovic, Graz University of Technology, Austria

9.40 CFD modelling of NOx formation in biomass grate furnaces with detailed chemistry Selma Zahirovic, Graz University of Technology, Austria

10:10 Biomass combustion on grates and NOx formation mechanisms Claes Tullin, SP, Sweden

10:40 Discussion

11:00 Closure

4

Report of the workshop Introduction, Sjaak van Loo Sjaak van Loo, coordinator of the combustion technology section of ThermalNet (CombNet), welcomed all participants (approx. 25) and speakers to the workshop. In this workshop, recent developments in the modelling and process control of grate furnaces are presented. This workshop was organised with key inputs from the EU-OptiComb project and IEA Bioenergy Task 32 (Biomass Combustion and Cofiring). The coordinator of OptiComb (Robert van Kessel) is also active as expert in CombNet. Over the whole project duration (2005-2007), CombNet will faciliatate and co-organise at least three workshops, as shown below: Organisers Topic Date, venue ThermalNet Opticomb IEA Bioenergy Task 32

Modelling and process control of grate furnaces

September 28, 2005, Innsbruck, Austria

ThermalNet IEA Bioenergy Task 32

Small combustion systems October 21, 2005, Paris, France

ThermalNet IEA Bioenergy Task 32

Biomass/coal co-firing Autumn 2006, Glasgow, UK

By far the largest share of all combustion installations for biomass and/or waste are equipped with a grate furnace. Grate furnaces are appropriate for biomass fuels with a high moisture content, varying particle sizes (with a downward limitation concerning the amount of fine particles in the fuel mixture), and high ash content. In practice the variability of the fuel may however result in fluctuations in combustion conditions, which may in return lead to ash related problems and fluctuations in steam production. In order to further lower emissions and costs while increasing combustion efficiency and stability of the combustion process, it is important that the combustion process is understood in detail. Recently detailed static and dynamic combustion models have been developed that describe the combustion of the fuel layer on the grate, as well as the reactions in the gas phase. Using this knowledge it is possible to design advanced combustion control mechanisms that significantly improve the combustion process. Combustion on a grate: dynamic modelling, process identification and process control Robert van Kessel, TNO, the Netherlands Robert van Kessel (R&D manager at TNO Science and Industry, Netherlands) presented the work done in the European OPTICOMB project, which provided significant inputs to this workshop, and then focused on work done at TNO. The overhead sheets presented are included in Annex 2.

5

The EU OptiComb project aims at improving the design of grate furnaces, in order to improve efficiency, lower emissions and improving controllability of the combustion process. TNO coordinates this project, in which 7 partners participate, including an equipment manufacturer Vyncke and an actual combustion unit in the Netherlands. The majority of the work that is presented in this ThermalNet workshop is derived from this EU project. TNO’s role in OptiComb is related to the development, validation and application of a dynamic model for grate systems. TNO has an extensive background and experience on this topic, particularly in the area of incinerators for municipal solid waste. Having available a reliable and accurate dynamic model for grate furnaces makes it possible to design more accurate control systems, leading to stabilized combustion conditions and steam production. An interesting spin-off of the work done is the development of an on-line calorific value soft sensor, which can be applied to evaluate the calorific value of the fuel instantaneously as it is burning on the grate. While conventional control systems are based on the steam production, having data on the heating value available earlier makes it possible to anticipate future process variations and effectively interact with the process to further stabilize the process. Characterisation of N-release from a biomass fuel layer by pot furnace experiments and derivation of release functions Selma Zahirovic, Graz University of Technology, Austria Selma Zahirovic presented the results of experimental work performed on a pot furnace, in order to derive relations of nitrogen release as a function of different parameters such as process conditions and fuel composition. This work was done using a pot furnace, to simulate what is actually happening in a (packed-bed) grate furnace. The work aimed at obtaining information about the flue gas composition above the fuel layer, and quantifying the rate of production of flue gas species dependent on variation of bed parameters with special attention on the release of NOx precursors. The overhead sheets presented are included in Annex 3. In the experiments, NH3 was found to be the main NOx precursor when MDF board and bark were used as fuel. In case of sawdust, NH3 and HCN were found to be the main precursors. Good quality experimental data was obtained that enabled the correlation of release of NOx precursors as a function of fuel and bed parameters. The empirical N-release functions that were obtained were of great value to develop both CFD models to describe the gas phase, as well as the fuel layer models of TU Graz and TNO. CFD modelling of NOx formation in biomass grate furnaces with detailed chemistry Selma Zahirovic, Graz University of Technology, Austria In her second presentation, Selma Zahirovic presented a 3D CFD NOx postprocessing model which was developed particularly for biomass grate furnaces. Initially an existing empirical model for fixed beds was extended by describing release of N species which are relevant for NOx formation, based on pot furnace experiments. The CFD model that describes the gas phase formation of NOx in a postprocessing calculation module was based on the Eddy Dissipation Concept and includes the Kilpinen 92 mechanism.

6

The resulting computer model describes both release of NOx precursors from the fuel bed as well as NOx formation in the gas phase. Validation of the model using FTIR measurements in a 440 kWth pilot scale furnace with horizontal boiler passes showed very good agreement of measured and calculated NOx emissions at the boiler outlet for different primary air ratios. Validation in a 7.2 MWth industrial scale plant with vertical boiler passes showed that measured NOx emissions are lower than calculated NOx emissions. An anticipated reason is be calculation errors in the Eddy Dissipation Model for the primary combustion zone, resulting in wrong prediction of hot spots which cause thermal and/or prompt NOx. Still, model prediction showed better results than literature data. It was concluded that the newly developed NOx postprocessing calculation unit gives results which are in good qualitative agreement with measurements under different operation conditions. To shorten calculation time, a reduced NOx mechanism is currently being developed. Biomass combustion on grates and NOx formation mechanisms Claes Tullin, SP, Sweden Whereas most of the work on NOx formation in fixed bed furnaces sofar has focused on the gas phase, Claes Tullin (SP, Sweden) presented recent work done on formation of NOx inside the fuel bed. An experimental rig was used to describe the properties of the propagating ignition front in terms of temperature and gas composition inside the bed. Concentrations of different gas components (both major species as well as nitrogen compounds) were obtained using a suction probe inside the bed. These measured concentrations were confirmed by mass balance calculations, assuming that hydrogen, nitrogen and tar concentrations (which were not measured) close the mass balance. The measurements concluded that fuel nitrogen is the major source for NOx formation, with NH3 as major precursor. This observation was also made in the work of TUG. At the temperatures measured inside the bed, thermal and prompt NOx formation mechanisms are much less relevant.

Annex 1. Introduction, Sjaak van Loo

Modeling and process control of grate furnaces

ThermalNet Workshop

September 28, Innsbruck, Austria

CombNet, September 27-30, Innsbruck, Austria

Introduction

Deliverables:•Network•Publications•Technical reports•Technology reviews

CombNet program:• Joint ThermalNet/IEA workshop

Small combustion systemsOctober 21, Paris, France

• Joint ThermalNet/IEA meeting/workshopBiomass/coal co-firingAutumn 2006, Glasgow, UK

• Joint TN/Obticomb/IEA workshopModeling and process control of grate furnacesSeptember 28, Innsbruck, Austria


Modeling and process control of grate furnacesTechnical focus:• Great Grate Combustion of biomass:

Largest share of biomass combustion installationsHigh fuel flexibility: moisture content

ash contentparticle size

• Decrease in emissions and costs• Increasing in combustion efficiency and stability of the combustion process� Design of advanced combustion control mechanisms

In this workshop, recent developments in the modeling and process control of grate furnaces are presented

ThermalNet: Non-ThermalNet:• Science and modeling EU: OptiComb• Environment, health and safety IEA Task 32• Gas treatment


Agenda8:40 Introduction OptiComb

Robert van Kessel, TNO, The Netherlands

8:50 Biomass combustion on grates and NOx formationmechanismsClaes Tullin, SP, Sweden

9:20 Characterization of N-release from a biomass fuel layerby pot furnace experiments and derivation of release functions Selma Zahirovic, Graz University of Technology, Austria

9:50 CFD modeling of NOx formation in biomass grate furnaceswith detailed chemistryRobbert Scharler, Graz University of Technology, Austria

10:20 Combustion on a grate: dynamic modeling, process identification and process controlRobert van Kessel, TNO, The Netherlands

Discussion

Annex 2. Combustion on a grate: dynamic modelling, process identification and process control Robert van Kessel, TNO, the Netherlands

TNO Science and Industry

Combustion on a grate: dynamicmodelling, process identificationand process control

ThermalNet/OPTICOMB/IEA workshopThermalNet MeetingInnsbruckSeptember 30, 2005

Robert van Kessel

Grate combustion 2

Contents

• OPTICOMB• Dynamic model for grate systems• Validation of dynamic models• On-line calorific value sensor• Application dynamic model• Conclusions

Grate combustion 3

Background OPTICOMB• Combustion of biomass play important role in sustainable

energy• At present in grate systems a limited range of fuels can be

used. More vast range of fuels result in a lower availability,due to limited flexibility of grate systems and controlconcepts.

• Improving grate, furnace and control concept design willimprove performance of biomass combustion grate systems

Grate combustion 4

Objectives

• Development and demonstration of advanced control conceptsfor biomass combustion grate systems.

• The development of guidelines, including demonstration, tominimise the important emissions of NOx and CO.

• Improvement of the efficiency (technical and economical) ofbiomass combustion plants.

• Design rules for biomass combustion systems and processcontrol systems.

• The design and testing of a new grate

Grate combustion 5

Project structure• Started 1-1-2003, End date 1-7-2006• Partners

• TNO-Science and Industry (NL), co-ordinator• VT-TUG (A) Selma Zahirovic• TU/e (NL)• Vyncke (B)• IST (P)• SP (Sweden) Claes Tullin• BES (NL)

Grate combustion 6

Description of WorkMain research points• NOx formation mechanisms• CFD modelling• Fuel layer modelling• Dynamic modelling• Controller design

Grate combustion 7

Description of WorkExperiments in 7.5 MWth Biomass combustion plant at

Schijndel (NL)• on-line calorific value sensor• system identification experiments to reveal plant dynamics• testing of control conceptAll experiments with different fuels

Grate combustion 8

Expected results OPTICOMB

• Innovative control concepts for biomass combustion.• Furnace concept for a new multi fuel biomass combustion

plant• Reduction of CO and NOx by 20-50%• Increased energy efficiency and availability• A new multi fuel grate system• A 3D-CFD combustion model for biomass fuels

Grate combustion 9

Dynamic model for grate stoker systems

Model structure

Controllersystem

Waste input(disturbances)

Combustionprocess

Grate combustion 10

Modelling of combustion processComprises 3 models:• Fuel layer model (dynamic)• Gas phase model (stationary)• Steam system model (dynamic)

Modelling of fuel layer model

Two different treatments• Simplified model for Model Predictive Control applications• Detailed modelling of fuel layer (1-D, 2-D)

Grate combustion 11

Application of fuel layer model1) Dynamic fuel layer model forms the basis for dynamic

model of a grate combustion process

2) Stationary model, which is part of the dynamic model can beused as boundary condition for CFD calculations (in co-operation with TU Graz)

Grate combustion 12

evaporation front

cold

reaction front

char burn out

preheated primary air

complete combustion

cold

reaction front

char burn out

primary air

cold

reaction front

char burn out

preheated primary air

evaporation front

complete combustion

ignition induced by grate movement

A: Combustion with no preheated primary air

B: Combustion with preheated primary air and

no effect of grate movements

C: Combustion with preheated primary air

including effect of grate movements

Grate combustion 14

Combustion process: steam system• Model components

• Superheater: flue gas stationary / steam stationary• Drum: flue gas stationary / steam dynamic• Economiser: flue gas stationary / steam stationary

Grate combustion 15

Validation of dynamic models (1)

• How to validate dynamic models?• Step response method• System identification

• System identification:• Experimental modeling resulting in dynamic input-output relations

without any physical meaning (black-box modeling)• Can be used for MIMO systems and for closed-loop systems.

Grate combustion 16

White-box model

Black-box model

t

u(t)

t

y1(t)

t

y2(t)

Comparison of y1(t) and y2(t):

Enough Resemblence ?

Yes STOP

No

Adapt parameters white-box model and obtain new

response(s) y2(t)

Grate combustion 17


Schematic of the controlled process, with:G incineration process C controlleru output controller u* process inputy output signal v process disturbanceex excitation signal r reference signal, setpoint

C Gr y

ex

u u*

H

v

e

Grate combustion 18


Waste input as a function of time.Comparison physical simulation model (-) and real plant data (- -).

Grate combustion 19


0 20 40 60 80 100 120 140 160 180 200−0.2

0

0.2

0.4

0.6

0.8

1

1.2Step applied to waste inlet flow of 10 [% controller scale]

time [min]

steam

pro

ducti

on [k

g/s]

Comparison dynamic model and plant results

Grate combustion 20

Grate stoker system

Structure Solid fuel combustion process

Controllersystem

Waste input(disturbances)

Combustionprocess

Grate combustion 21

0 100 200 300 400 500 600 700

30

40

50

60

70

time [min]

Dos

age

[%]

modelled controllermeasured

0 100 200 300 400 500 600 700

30

40

50

60

70

time [min]

Dos

age

[%]

modelled controller & process

Validation of controller model

Grate combustion 22

Grate stoker system

Structure MSWC process

Controllersystem

W aste input(disturbances)

Com bustionprocess

Grate combustion 23

On-line calorific value sensor (1)

Changing calorific value of the fuel is one of the main problemsin solid fuel (biomass, waste) combustion:

Development of an on-line calorific value sensor

Requirements:• No energy balance, and• No mass flows

The patented sensor is based upon a model and thefollowing measurements:• H2O, O2 en CO2-concentrations (with IR)• Relative humidity of the ambient air

Grate combustion 24

Calorific value waste, moisture fraction waste andcalorific value combustible part as function of time

0 10 20 30 40 50 60 70 808

9

10

11

12

13

tijd [dag]

Ha

fva

l [M

J/kg

]

0 10 20 30 40 50 60 70 800.25

0.3

0.35

0.4

tijd [dag]

Xw

ate

r [k

g/k

g]

0 10 20 30 40 50 60 70 8020

25

30

35

40

tijd [dag]

Hb

ran

d [M

J/kg

]


Grate combustion 25


0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 165

70

75

80

85

90

95

100

105

tijd [dag]

PH

Isto

om [t

/h]

berekendgemeten

Calculated and measured steam production as a function of time.

Grate combustion 26


Possible applications:• Calorific value sensor as a diagnostic tool• Continuous determination on-line mass- and energy

balances• Source of additional information for operators• Integration of the sensor in the control concept in order to

reduce fluctuations

Grate combustion 27

Application of dynamic combustion model• Process Analysis• Simulator• Optimisation of control concepts

Grate combustion 28

Optimisation of control concept (1)

Different possibilities for optimisation of control conceptby using validated model• Tuning present control concept• Testing new classical control concepts• Development of new advanced control concepts

e.g. Model Predictive Control

Grate combustion 29

Optimisation of control concept (2)

AVR plant, optimisation by tuning of control parameters

−40 −30 −20 −10 0 10 20 30 400

0.02

0.04

0.06

0.08

0.1

0.12

P(x

)

Φsteam,actual

−Φsteam,set

[t/h]

previous σ = 5.45tuned σ = 3.63

Grate combustion 30

Model Predictive ControlMPC: based upon measurements from the past, a model of the

plant and the control objectives it predicts the plantbehavior in the near future with respect to the constraintsand boundary conditions of the system.

Based upon the control objectives it calculates at every sampletime t, the most optimal control actions for the near future.At every time sample t this is repeated.

Mathematically: an optimization problem

Grate combustion 31

Conclusions• Complex processes like solid fuel grate combustion can be

better understood by modelling• Validation is very important• On-line calorific value sensor is available• New control concepts can be tested easily with a process

model• Will be applied next year in OPTICOMB project at a Dutch

plant

Annex 3. Characterisation of N-release from a biomass fuel layer by pot furnace experiments and derivation of release functions Selma Zahirovic, Graz University of Technology, Austria

International workshop

Modelling and process control of grate furnaces30 September 2005Innsbruck, Austria

Institute for Resource Efficient andSustainable Systems Graz University of Technology

CharacterisationCharacterisation of Nof N--releasereleasefrom a biomass fuel layer by pot furnace experimentsfrom a biomass fuel layer by pot furnace experiments

and derivation of Nand derivation of N--release functionsrelease functionsEmil Widmann, Selma Zahirovic, Robert Scharler, Ingwald Obernberger

Institute for Resource Efficient andSustainable Systems Graz University of Technology OverviewOverview

� Scope of work

� Description of the experimental set-up

� Experimental results for fuels tested

� Derivation of the N-release functions

� Summary and conclusions

Institute for Resource Efficient andSustainable Systems Graz University of Technology Scope of workScope of work

� Experimental investigation of the combustion properties of a packed bed (fuel-layer) for three biomass fuels in order to

� Obtain information about the flue gas composition above the fuel layer,

� Quantify the rate of production of flue gas species dependent on variation of bed parameters with

� Special attention on the release of NOx precursors

� Derivation of N-release functions based on experimental data for the purpose of modelling


Pot furnace experiments vs. Pot furnace experiments vs. combustion of fuel on the grate combustion of fuel on the grate

Experimental installation was designed in a way to represent combustion conditions of a biomass fuel layer on a grate as close as possible

� Allows to control combustion parameters

� Allows access for measurements

hgrate

Aslice

t on grate

q radiation

Φ air flow

Biomass

hreactor

Areactor

t burnout

q radiation

Φ air flow

Biomass

Institute for Resource Efficient andSustainable Systems Graz University of Technology Experimental setExperimental set--up up

Experimental set-up (left) and scheme (right) of the pot furnace reactor

Explanations: A...SiC reactor core; B...heater elements; C...heated filter; D...dilution unit; E...extractive FT-IR; F...in-situ FT-IR; G...primary air supply; H…sample holder

B1

B2

B3

B4B590

50

10

25 25

in-situ FT-IR(NH3, CO, CO2,CH4 and H2O)

thermocouples

fuel bed(with 6 thermocouples)

air flowoil sealing

insulating firebrick

heater elementssection 1

extractive FT-IR

weight balance

heater elementssection 2


Fuel analysis Fuel analysis –– ultimate analysisultimate analysisand particle size distributionand particle size distribution

Bark particle size mass fractionC 49.50 [wt% d.b.] 0.47 [%] < 2 [mm] 0.08 [-]H 5.60 [wt% d.b.] 0.09 [%] 2 [mm] - 4 [mm] 0.13 [-]N 0.27 [wt% d.b.] 0.01 [%] 4 [mm] - 8 [mm] 0.30 [-]O (calc.) 40.10 [wt% d.b.] 0.57 [%] 8 [mm] - 12.5 [mm] 0.26 [-]ash 4.50 [wt% d.b.] 0.01 [%] 12.5 [mm] - 16 [mm] 0.07 [-]water 7.40 [wt% w.b.] 0.16 [%] > 16 [mm] 0.16 [-]

rms absaverage value

MDF particle size mass fractionC 46.20 [wt% d.b.] 0.9 [%] < 2 [mm] 0.06 [-]H 6.60 [wt% d.b.] 0.5 [%] 2 [mm] - 4 [mm] 0.06 [-]N 6.90 [wt% d.b.] 0.2 [%] 4 [mm] - 10 [mm] 0.25 [-]O (calc.) 38.40 [wt% d.b.] 0.6 [%] 10 [mm] - 16 [mm] 0.29 [-]ash 1.90 [wt% d.b.] 0.1 [%] 16 [mm] - 40 [mm] 0.26 [-]water 7.50 [wt% w.b.] 0.1 [%] > 40 [mm] 0.09 [-]


Sawdust particle size mass fractionC 49.10 [wt% d.b.] 1.1 [%] < 0.4 [mm] 0.06 [-]H 6.60 [wt% d.b.] 0.5 [%] 0.4 [mm] - 0.63 [mm] 0.13 [-]N 0.06 [wt% d.b.] 0.004 [%] 0.63 [mm] - 1.0 [mm] 0.26 [-]O (calc.) 44.20 [wt% d.b.] 0.6 [%] 1.0 [mm] - 1.6 [mm] 0.30 [-]ash 0.20 [wt% d.b.] 0.0 [%] 1.6 [mm] - 2.5 [mm] 0.15 [-]water < 0.10 [wt% w.b.] - [%] > 2.5 [mm] 0.09 [-]



Comparison with FID Comparison with FID measurementsmeasurements

0.00E+00

5.00E-04

1.00E-03

1.50E-03

2.00E-03

2.50E-03

3.00E-03

3.50E-03

0 200 400 600 800 1000 1200

time [s]

Cre

leas

ein

CxH

yOz

[mol

/s]

FID: Carbon release inCxHyOz

FTIR: Carbon release inCxHyOz

Release of hydrocarbons measured with extractive FT-IR (for different fuels and combustion conditions) was cross-checked with measurements performed with FID equipment:


Elemental recovery rates Elemental recovery rates for reference experimentsfor reference experiments

Elemental recovery rate rj relates the measured (flue gas concentration) yield of each element to the total amount of the element in the experiment (fuel analysis):

0.0

20.0

40.0

60.0

80.0

100.0

120.0

140.0

r-tot r-C r-H r-O

reco

very

rate

s[w

t%]

sawdust bark MDF


Results Results ––reference experiment barkreference experiment bark

0

200

400

600

800

1000

1200

1400

0 100 200 300 400 500 600 700 800 900 1000 1100 1200

time [s]

tem

pera

tur[

°C]

TC - flue gas (averaged) TC - Bed1 (h = 90 mm) TC - Bed2 (h = 50 mm)TC - Bed3 (h= 10 mm) TC - Bed4 (h = 50 mm) TC - Bed5 (h = 50 mm)


Results Results ––reference experiment MDFreference experiment MDF

0

20

40

60

80

100

120

140

160

180

0 200 400 600 800 1000 1200 1400 1600 1800 2000

time [s]

sam

ple

mas

s[g

]

0

200

400

600

800

1000

1200

0 200 400 600 800 1000 1200 1400 1600 1800

time [s]

Tem

pera

tur[

°C]

TC - flue gas (averaged) TC - Bed1 (h = 90 mm) TC - Bed5 (h=50 mm)TC - Bed4 (h = 50 mm) TC - Bed2 (h = 50 mm) TC - Bed3 (h = 10 mm )


Results Results ––reference experiment sawdustreference experiment sawdust

0

20

40

60

80

100

120

140

0 500 1000 1500 2000 2500

time [s]

[g]

0

100

200

300

400

500

600

700

800

900

0 200 400 600 800 1000 1200 1400 1600 1800 2000

time [s]

tem

pera

tur[

°C]

TC - flue gas (averaged) TC Bed1 (h = 90 mm) TC Bed5 (h = 50 mm)TC Bed3 (h = 10 mm) TC Bed2 (h = 50 mm) TC Bed4 (h = 50 mm)


Conversion rates Conversion rates for different fuelsfor different fuels

Conversion rate ui relates the yield of each nitrogen species (with exception of N2) to the total amount of nitrogen in the fuel:

2.73.5 3.2 1.74.30.1 0.4 0.4

18.3

6.7

15.7

43.3

69.4

8.2

30.9

0

10

20

30

40

50

60

70

80

90

u NO u NH3 u HCN u NO2 u N2O

nitr

ogen

conv

ersi

onra

te[%

]

reference experiment sawdustreference experiment barkreference experiment MDF board


Influence of the fuel N content Influence of the fuel N content on the total conversion rate on the total conversion rate

for different fuels for different fuels

0

20

40

60

80

100

120

0 1 2 3 4 5 6 7 8

nitrogen content [wt% d.b.]

conv

ersi

onra

teTF

N[%

] waste wood

bark

sawdust

MDF board

fibreboard

Conversion rate uTFN relates the yield of all nitrogen species (with exception of N2) to the total amount of nitrogen in the fuel:


ModelledModelled releaserelease of N of N speciesspeciesbasedbased on experimental on experimental datadata ––

sawdustsawdust

0

20

40

60

80

100

120

140

0 0.2 0.4 0.6 0.8 1normalised length on grate [-]

conc

entr

atio

n[p

pmV

d.b.

]

HCN modelledHCN experiments

0

50

100

150

200

250


conc

entr

atio

n[p

pmV

d.b.

]

NH3 modelledNH3 experiments

0

20

40

60

80

100

120

140

160


conc

entr

atio

n[p

pmV

d.b.

]

NO modelledNO experiments

iii dku += λ

experiment

model

Institute for Resource Efficient andSustainable Systems Graz University of Technology Summary and conclusions ISummary and conclusions I

� Fuel analysis was performed for bark, MDF and sawdust.

� Good quality of the experiments performed at the pot furnace wasachieved: high elemental recovery rates and good agreement of results of two different measurement systems for the detection of hydrocarbons.

� Species release rates were determined for all fuels under different combustion conditions.

� NH3 was found to be the main NOx precursor for MDF board and bark.

� NH3 and HCN were found to be the main precursors for sawdust.

� Total conversion rates drop with increasing content of fuel N.

Institute for Resource Efficient andSustainable Systems Graz University of Technology Summary and conclusions II Summary and conclusions II

� Experimental data was applied for the derivation of empirical N-release functions for different fuels as a function of stoichiometric ratio.

� The empirical N-release functions have been implemented in an empirical fuel layer model of TUG and are currently being implemented in the fuel layer model of TNO.

� The model validation in both cases is based on the data gained from the pot furnace experiments.

� The fuel layer models developed provide a valuable basis for CFDsimulations of gas phase combustion and NOx formation.

Annex 4. CFD modelling of NOx formation in biomass grate furnaces with detailed chemistry Selma Zahirovic, Graz University of Technology, Austriai

International workshop

Modelling and process control of grate furnaces30 September 2005Innsbruck, Austria


CFD modelling of CFD modelling of NONOxx formation in formation in biomass grate furnaces with detailed chemistrybiomass grate furnaces with detailed chemistry

Robert Scharler, Selma Zahirovic, Emil Widmann, Ingwald Obernberger

Institute for Resource Efficient andSustainable Systems Graz University of Technology OverviewOverview

� Scope of work

� Modelling

� Empirical fixed bed modelling

� Modelling of turbulent reactive flow – basic combustion modelling

� CFD NOx postprocessing

� Test of the CFD NOx postprocessor – methodology and discussion of results

� Simulation of a 440 kWth pilot-scale plant (fibre board as fuel)

� Simulation of a 7.2 MWth industrial-scale plant (waste wood as fuel)

� Summary and conclusions

Institute for Resource Efficient andSustainable Systems Graz University of Technology Scope of workScope of work

� Development of a 3D CFD NOx formation model (postprocessor) including detailed reaction kinetics for biomass grate furnaces

� must be applicable to engineering problems

� with reasonable accuracy

� with reasonable calculation time

� Test of the CFD NOx postprocessor

� Simulation of a pilot-scale biomass grate furnace and comparison with measurement data taken during two test runs with fibre board as fuel

� Simulation of an industrial-scale biomass grate furnace and comparison with measurement data taken during normal boiler operation with waste wood as fuel


Empirical fixed bed model Empirical fixed bed model ––basic versionbasic version

� Definition of profiles for the distribution of primary air and recirculated flue gas as well as drying and thermal decomposition of the solid biomass (C, H, O) along the grate on the basis of test runs

� Definition of conversion parameters for CH4, H2, CO, CO2, H2O, and O2 in the flue gas released based on literature data and lab-scale experiments

� Stepwise balancing of mass, species and energy

0

400

800

1200

1600

0 0.5 1 1.5Length on grate [m]

Tem

pera

ture

[K]

0

4

8

12

16

wt%

H2O

(w.b

.)

Temperaturewt% H2O (w. b.)

Example: Calculated profiles of temperature and H2O concentration in the flue gas along the grate


Extension of the fixed bed model Extension of the fixed bed model ––release of N speciesrelease of N species

� The empirical fuel bed combustion model was extended in order to describe the release of N species (NO and NH3 as well as HCN) which are relevant for the formation of fuel NOx in biomass grate furnaces (fibre board, waste wood, bark)

� Conversion functions (as a function of local λ) were defined for the investigated fuels based on lab-scale pot furnace (batch) reactor experiments; NH3 showed to be the predominant NOx precursor, HCN was found only in very low concentrations

Example: calculated profiles of NH3, HCN and NO concentration in the flue gas along the grate(left...pilot-scale plant; fuel: fibre board; right...industrial-scale plant; fuel: waste wood)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.0 0.5 1.0 1.5 2.0 2.5length on grate [m]

[wt%

NH

3,H

CN

-wet

flue

gas]

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

[wt%

NO

-wet

flue

gas]

NH3HCNNO

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0 1 2 3 4 5length on grate [m]

[wt%

NH

3,H

CN

-wet

flue

gas]

0.00

0.01

0.02

0.03

0.04

0.05

[wt%

NO

-wet

flue

gas]

NH3HCNNO

[wt%

NO

-wet

flue

gas]


Modelling of turbulent reactive flow – basic combustion simulation

� Turbulence Realizable k-ε model

� Gas phase combustion Eddy Dissipation model (Amag = 0.6, Bmag = 0.5) /global methane 3-step mechanism (CH4, CO, CO2, H2, H2O und O2)

� Radiation Discrete Ordinates model

Modelling of NOx formation – postprocessing mode

� Eddy Dissipation Concept (EDC)

� Kilpinen 92 mechanism (50 species, 253 reactions)

� ISAT (In-Situ Adaptive Tabulation) algorithm for reaction kinetics

CFD modelsCFD models


Eddy Dissipation Concept (EDC) – implementation in Fluent 6.1 based on Gran and Magnussen (1996)

Net production rate Ri [kg/m3s]

Eddy Dissipation ConceptEddy Dissipation Concept

−

−

⋅= iYiYiR ~*31*

2

γτ

γρ

ρ… time averaged (-) density [kg/m3]τ*… residence time fine structures [s] = f(tk) = f(ε, ν) modelledγ… length scale of fine structure regions [-] = f(k, ε, ν) modelledYi… Favre-averaged (~) and fine structure values (*) of species mass fraction Yi [-] of species i [-]

� Empirical expression; reactions occur mainly in the smallest length scales of the turbulent energy cascade (fine structures) where turbulent energy is dissipated

� EDC assumes that the fluid state is determined by the fine structure state (*), the surrounding state (~) and the fractions of the fine structure (γ3)

� Fine structures are treated as ideal reactors (in FLUENT… plug flow reactor) =>integration of reaction kinetics / closure of equation system


Test runs Test runs ––440 440 kWkWthth pilotpilot--scale plantscale plant

� Conventional flue gas analysis at boiler outlet (NOx, CO, O2, CO2)

� In-situ FT-IR measurement ports I - III (CH4, CO, CO2, H2O, NH3) –case A: port III; case B: port II

� Temperature measurements (thermocouples T1 – T3)� Additionally: data from literature, experience and lab-

scale pot furnace experiments concerning relevant species concentrations (NO, NO2, HCN, NH3)

CFD model boundary/furnace outlet

PCZ…primary combustion zoneSCZ…secondary combustion zone

operation data case A case Bfuel fibre board fibre boardwater content 10.60 10.60 wt% d.b.nitrogen content 3.06 3.06 wt% d.b.fuel power related to NCV 456 448 kWth

lambda fuel bed eff 0.78 1.50 -lambda primary eff 0.97 1.63 -total air ratio 1.41 1.61 -flue gas recirculation ratio 0.49 0.46 -adiabatic flame temperature 933 888 °Cmeasured NOx emissions 264 303 ppmv


Simulated NHSimulated NH33 profiles profiles ––440 440 kWkWthth pilotpilot--scale plant scale plant

NH3 mole fraction [-] profiles in the symmetry plane of the pilot-scale biomass grate furnace

� NH3 is consumed not immediately above the fuel bed but somewhere in the furnace depending on the stoichiometry (earlier for higher λ) =>confirmation by in-situ FT-IR measurements and pot furnace experiments

� NH3 and HCN concentrations at furnace outlet are very low => confirmation by literature data and experience (with extractive FT-IR measurements at outlet of various boilers)

Case A…λprim < 1 Case B…λprim > 1


Simulated NO profiles Simulated NO profiles ––440 440 kWkWthth pilotpilot--scale plant scale plant

NO mole fraction [-] profiles in the symmetry plane of the pilot-scale biomass grate furnace

� Simulated NOx emissions consisted mainly of NO; NO2 concentrations were very low (between 5 and 10 ppmv) => confirmed by experience (with conventional flue gas analysis and extractive FT-IR measurements at the outlet of various boilers) and literature

Case A…λprim < 1 Case B…λprim > 1


NONOxx emissions emissions ––440 440 kWkWthth pilotpilot--scale plant scale plant

Explanations: Case A…λprim < 1; case B…λprim > 1;literature data…NH3 and HCN in concentrations with same order of magnitude;experimental data TU Graz…NH3 predominant species, HCN is negligible

� Very good agreement of measured NOx emissions at boiler outlet and simulations

� Simulated NOx emissions at furnace outlet are lower for case A =>confirmed by conventional flue gas analysis at boiler outlet

� Case A: simulated NOx emissions based on the release of NH3 from the fuel bed as predominant NOx precursor (lab-scale experiments) are closer to the NOx measurements at boiler outlet than based on a release of NH3 and HCN in similar concentrations (literature data)

case A case B

measured 264 303

calculated TU Graz 287 332literature 335 -

source data empirical fixed bed model

NOx emissions [ppmv]


Test runs Test runs ––7.2 7.2 MWMWthth industrialindustrial--scale plantscale plant

� Conventional flue gas analysis at boiler outlet (NOx, CO, O2, CO2)

� Additionally: data from literature, experience and lab-scale pot furnace experiments concerning relevant species concentrations (NO, NO2, HCN, NH3)

PCZ…primary combustion zoneSCZ…secondary combustion zone

flue gas recirculation below the grate

21 43

secondary air

primary air

biomass fuel bed

flue gas recirculation

above the grate

SCZ

PCZ -cooled walls

PCZ

SCZ

PCZ

CFD model boundary/furnace outlet

operation datafuel waste woodwater content 17.70 wt% d.b.nitrogen content 1.20 wt% d.b.fuel power related to NCV 7,570 kWth

lambda fuel bed eff 1.12 -lambda primary eff 1.29 -total air ratio 1.75 -flue gas recirculation ratio 0.18 -adiabatic flame temperature 1,120 °Cmeasured NOx emissions 140 ppmv


NONOxx emissions emissions ––7.2 7.2 MWMWthth industrialindustrial--scale plant scale plant

Explanations: Literature data…NH3 and HCN in concentrations with same order of magnitude;experimental data TU Graz…NH3 predominant species, HCN in low concentrations;lowered temperature peaks...peak values of mean flue gas temperature in the primarycombustion zone were lowered with a damping function

� Larger deviations between measured and simulated NOx emissions than for the pilot-scale plant

� Simulated NOx emissions based on a release of NH3 from the fuel bed as predominant NOxprecursor (lab-scale experiments) are closer to the NOx measurements at boiler outlet than based on a release of NH3 and HCN in similar concentrations (literature data)

� Simulated NOx emissions decline with reduced temperatures in the primary combustion zone

measured 140

calculated TU Graz 233literature 293

TU Graz lowered temperature peaks 213

source data empirical fixed bed model

NOx emissions [ppmv]

note


Simulated Simulated temperature and NO profiles temperature and NO profiles ––7.2 7.2 MWMWthth industrialindustrial--scale plant scale plant

Profiles of fine scale temperature [°C] (left) and NO mole fraction [-] (right) in the symmetry plane of the industrial-scale biomass grate furnace

� Very high fine scale temperatures => possible errors of fixed bed modelling and basic combustion simulation with the EDM

� Simulated NOx emissions decrease with reduced temperatures in the primary combustion zone => predicted “hot spots” may cause thermal and prompt NOx

Institute for Resource Efficient andSustainable Systems Graz University of Technology Summary ISummary I

� Lab-scale pot-furnace experiments revealed that NH3 is the dominating species released

from the fuel bed for fibre board, waste wood and bark =>

the results are an important basis for CFD NOx postprocessing

� 3D simulations of biomass grate furnaces with the new CFD NOx post-processor including

detailed chemistry were performed for the first time

� Simulation time: between 1 and 3 weeks; a reduction by parallel processing and a recently

improved ISAT algorithm is expected

� Both furnaces: good qualitative agreement of simulation results concerning relevant

species concentrations (NO, NO2, HCN, NH3) with measurements under different operating

conditions as well as with data from lab-scale experiments, experience and literature

Institute for Resource Efficient andSustainable Systems Graz University of Technology Summary IISummary II

� Pilot-scale plant: very good agreement of NOx measurements after boiler outlet and simulation results for air lean and air rich conditions in the primary combustion zone (deviation about +10 % in both cases)

� The effect of air staging was correctly reproduced in the simulations

� Industrial-scale plant: reasonable agreement of NOx measurements and simulation results, but larger deviations than for the pilot-scale plant (+50% to +65%)

� Failings of the empirical fixed bed model and the basic combustion simulation with the EDM are responsible for the larger deviations; e.g. calculated NOx formation rates above the fuel bed were too high due to over-predicted flue gas temperatures (“hot spots”)

Institute for Resource Efficient andSustainable Systems Graz University of Technology ConclusionsConclusions

� The newly developed NOx postprocessor has been successfully tested

� The NOx postprocessor for biomass grate furnaces is a powerful tool for the design and optimisation of furnace geometries and process control

� Further comparisons with measurements are necessary in order to improve and validate the model

� Improvements concerning fixed bed modelling and combustion modelling (test of advanced models) are in progress

� A reduced NOx mechanism is being developed in order to reduce calculation time for engineering applications and to overcome failings of basic combustion simulation with a coupled simulation of the combustion process and NOx formation

Annex 5. Biomass combustion on grates and NOx formation mechanisms Claes Tullin, SP, Sweden

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Biomass Combustion on Gratesand

NOx-formation mechanisms

Claes TullinMarie Rönnbäck

Jessica Samuelsson

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Outline

• Introduction• What goes on in a fixed biomass bed?• N-conversion in a fixed biomass bed

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Do we know enough? Available data and models

Boundary conditions Nitrogen chemistry

Fair knowledge

Reasonableknowledge

Limitedknowledge

Very limitedknowledge

Combustion processes:

Drying

Devolatilisation and gas phase combustion

Char combustion

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Issues in grate combustion

• Fuel homogenity and feed control• Evenly distributed fuel bed• Fuel transportation control• Air distribution control (∆p over grate)• Air stoichiometry• Secondary combustion• …..

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Part 1What goes on in a biomass fuel bed?

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Videoanalysis of a fuel bed in a 12 MW boiler

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Video

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Gas composition in a fixed bed of biofuel

- measurements in and above a downward propagating ignition front

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Propagation of ignition front

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Experimental rig and fuel

Fuel: pellets of compressed sawdust

diameter 8 mm

moisture 11 %

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Purpose

To describe the properties of the ignition front in terms of gas composition

To confirm the measured gas composition by closing the mass balance

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Experimental rig –ignition front counter-current to the air flow

Grate: 0.35 m x 0.35 m, height 0.7 m

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Measurement set-up

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Results: Measured concentrations for a batchSuperficial velocity 0.14 m/s, pellet with moisture 11 %

50 60 70 80 90 100 110 120 1300

5

10

15

20

25

30

Con

c.(V

ol-%

,wet

gas)

Time (min)

H2

H2 H2

O2

CH4THC

COCO2

H2O

60 80 100 1200

10

20

30

40

50N2→

N2→N2→

Nitr

ogen

(Vol

-%,w

etga

s)N

itrog

en(V

ol-%

,wet

gas)

Con

c.(V

ol-%

,wet

gas)

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Results: Measured concentrations in the frontSuperficial gas velocity 0.14 m/s, pellet with moisture 11 %

55 60 650

5

10

15

20

25

Con

c.(V

ol-%

,wet

gas)

Time (min)

H2

O2

CH4THC

CO

CO2

H2O

Position in conversion front (mm)0 15 30 45 60

Con

c.(V

ol-%

,wet

gas)

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Nitrogen and hydrogenComparison exp data and mass balance calculations

50 60 70 80 90 1000

20

40

60

80

Time (min)

H2,

N2

(Vol

-%,w

etga

s)

H2

N2

Thin lines: results from mass balance

Thick lines: measured with bag sampling

H2,

N2

(Vol

-%,w

etga

s)

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Results: Tar

50 60 70 80 90 1000

0.05

0.1

0.15

0.2

Time (min)

Tar(

kg/k

g)

Tar (kg tar/kg devolatilized fuel)All hydrocarbons that condense > 190 °C)

Tar(

kg/k

gs)

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Part 2NOx-formation mechanisms

�NOx mechanisms

�Primary NOx-reduction methods

�Secondary NOx-reduction methods

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

NOx-mechanisms

• Fuel-N oxidation• Thermal NOx• Prompt NOx

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Conversion of Fuel-N to NOx

NHi NO

N2

O2

NO

NHi

Char-N

Vol-N∼80 %

~20 %

Fuel-N

� Fuel-N is the major source for NOx duringbiomass combustion

Important parameters:

- Fuel-N content

- Temperature

- O2

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

500 700 900 1100 1300 1500 1700 190010-1

100

101

102

103

104

ppm

Temperature [ οC]

21 % O2101

Oxidation of N2 in air – Equilibrium

NONO ↔+ 221

221

ppm

NO

Equilibrium concentrations of NO in a gasmixture of O2 and N2.

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Thermal NOx – Oxidation of air N2

1300 1500 17000

100

200

300

400

500

600

700

800

ppm

/s

Temperature [οC]

21% O2101

� Formation of thermal NOx negligeable at T < 1400 °C.

Extended Zeldovich mechanism

NNOON +→+2

ONOON +→+ 2

HNOOHN +→+

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Prompt NOx – Oxidation of N2 in air

What is prompt NOx?

(3. Thermal NOx at supercritical equilibrium concentrations of radicals)

2. Reaction via N2OMONMON +↔++ 22

NONOOON +↔+2

1. Oxidation of nitrogen in air involving hydrocarbon radicalsNHCNNCH +↔+ 2

T < 1400 °C � Negligeable formation of prompt NOx

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Fuel-N conversion to NOx– a very complex process

How is the nitrogen bound in the fuel?In biomass - N bound mainly in proteins

Fate of N in the fuel during pyrolysis/devolatilisation and char combustion?

Solid phase reactionsProtein depolymerisationChar formationEmitted from fuel particles as NH3, HCN, HNCO, NO … or N2Heterogeneous (char catalysed) reactionsInfluence of inorganic material…..

Complex homogenous gas phase chemistry

NOx emissions = NOx formed – NOx destroyed

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Methods for NOx-reduction

Primary methods�Combustion control�Air staging�Fuel staging�Flue gas recirculation

Secondary methods�SCR – Selective Catalytic Reduction�SNCR – Selective Non-Catalytic Reduction

NO

N2

O2

NONHi

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Heated sampling line

Dry air

Mass flowcontroller

FilterTar trapSuction probe

Cooler

CO2 CO/CH4

O2

Heated filter

bag

Cooler

NO

FTIR

THC

Absorption

Measurements in burning fuel bed

�Major species: O2, H2O, CO2, CO, H2,

N2, THC

�Nitrogen compounds:NH3, HCN, NO, NO2,N2O

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

-2 0 2 4 6 8 100

200

400

600

800

1000

1200

ο C

min

0.070.140.210.310.35

”Low” T´s indicate that thermal NOx is not important

Temp in reaction front at different air flows (m/s)

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Gas composition in a fuel bed

0100200300

ppm

NH3

NOHCN

0 10 20 30 40 50 60 7005

101520

min

%

O2

CO2

H2O

THC

CO

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Conclusions- Processes within a biomass fuel bed

Gas concentrations of all major species in and above an ignition front propagating counter-current to the air were successfully measured

The measured composition was confirmed by closing the mass balance

Concentrations of hydrogen, nitrogen and tar, that are commonly not measured, can be calculated for combustion of biofuel for this combustion case

SPSw

edis

hN

atio

nalT

estin

gan

dR

esea

rch

Inst

itute

Conclusions- Nitrogen chemistry in biomass fuel beds

Fuel nitrogen is the major source for NOx-formation on grates

Thermal and/or prompt NOx only forms at high temperatures

NH3 a major precursor for NOx

Nature is on our side, i.e. NOx emissions can be decreased by primary measures

Well known methods available for secondary NOx reduction

modeling and process control of grate furnaces · modeling and process control of grate furnaces...

Documents