Modeling and process control of grate furnaces

Transcription

1 Workshop on Modeling and process control of grate furnaces Arranged by: Jaap Koppejan, TNO Science and Industry, Netherlands Sjaak van Loo, Chess, Netherlands September 28, 2005 Hilton Hotel Innsbruck, Austria Organised by:

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. Annex 2. Annex 3. Annex 4. Annex 5. Introduction, Sjaak van Loo Combustion on a grate: dynamic modelling, process identification and process control Robert van Kessel, TNO, the Netherlands 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 CFD modelling of NOx formation in biomass grate furnaces with detailed chemistry Selma Zahirovic, Graz University of Technology, Austriai Biomass combustion on grates and NOx formation mechanisms Claes Tullin, SP, Sweden 2

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 3

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 ( ), CombNet will faciliatate and co-organise at least three workshops, as shown below: Organisers Topic Date, venue ThermalNet Opticomb Modelling and process control of grate furnaces September 28, 2005, Innsbruck, Austria IEA Bioenergy Task 32 ThermalNet Small combustion systems October 21, 2005, Paris, IEA Bioenergy Task 32 ThermalNet IEA Bioenergy Task 32 Biomass/coal co-firing France 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. 4

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, NH 3 was found to be the main NOx precursor when MDF board and bark were used as fuel. In case of sawdust, NH 3 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 NO x 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. 5

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 NO x formation, with NH 3 as major precursor. This observation was also made in the work of TUG. At the temperatures measured inside the bed, thermal and prompt NO x formation mechanisms are much less relevant. 6

7 Annex 1. Introduction, Sjaak van Loo Modeling and process control of grate furnaces ThermalNet Workshop September 28, Innsbruck, Austria Introduction Deliverables: Network Publications Technical reports Technology reviews CombNet program: Joint ThermalNet/IEA workshop Small combustion systems October 21, Paris, France Joint ThermalNet/IEA meeting/workshop Biomass/coal co-firing Autumn 2006, Glasgow, UK Joint TN/Obticomb/IEA workshop Modeling and process control of grate furnaces September 28, Innsbruck, Austria CombNet, September 27-30, Innsbruck, Austria

8 Modeling and process control of grate furnaces Technical focus: Great Grate Combustion of biomass: Largest share of biomass combustion installations High fuel flexibility: moisture content ash content particle 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 CombNet, September 27-30, Innsbruck, Austria Agenda 8:40 Introduction OptiComb Robert van Kessel, TNO, The Netherlands 8:50 Biomass combustion on grates and NO x formation mechanisms Claes Tullin, SP, Sweden 9:20 Characterization of N-release from a biomass fuel layer by pot furnace experiments and derivation of release functions Selma Zahirovic, Graz University of Technology, Austria 9:50 CFD modeling of NO x formation in biomass grate furnaces with detailed chemistry Robbert Scharler, Graz University of Technology, Austria 10:20 Combustion on a grate: dynamic modeling, process identification and process control Robert van Kessel, TNO, The Netherlands Discussion CombNet, September 27-30, Innsbruck, Austria

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

10 Combustion on a grate: dynamic modelling, process identification and process control TNO Science and Industry Robert van Kessel ThermalNet/OPTICOMB/IEA workshop ThermalNet Meeting Innsbruck September 30, 2005

11 Contents OPTICOMB Dynamic model for grate systems Validation of dynamic models On-line calorific value sensor Application dynamic model Conclusions Grate combustion 2

12 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 control concepts. Improving grate, furnace and control concept design will improve performance of biomass combustion grate systems Grate combustion 3

13 Objectives Development and demonstration of advanced control concepts for biomass combustion grate systems. The development of guidelines, including demonstration, to minimise the important emissions of NOx and CO. Improvement of the efficiency (technical and economical) of biomass combustion plants. Design rules for biomass combustion systems and process control systems. The design and testing of a new grate Grate combustion 4

14 Project structure Started , End date 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 5

15 Description of Work Main research points NOx formation mechanisms CFD modelling Fuel layer modelling Dynamic modelling Controller design Grate combustion 6

16 Description of Work Experiments 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 concept All experiments with different fuels Grate combustion 7

17 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 8

18 Dynamic model for grate stoker systems Waste input (disturbances) Controller system Combustion process Model structure Grate combustion 9

19 Modelling of combustion process Comprises 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 10

20 Application of fuel layer model 1) 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 be used as boundary condition for CFD calculations (in cooperation with TU Graz) Grate combustion 11

21 cold reaction front char burn out cold reaction front complete combustion char burn out cold reaction front ignition induced by grate movement complete combustion char burn out primary air evaporation front preheated primary air evaporation front preheated primary air 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 12

22 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 14

23 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 15

24 u(t) t y1(t) Black-box model y2(t) t Comparison of y1(t) and y2(t): Enough Resemblence? Yes STOP White-box model t No Adapt parameters white-box model and obtain new response(s) y2(t) Grate combustion 16

25 Validation of dynamic models (3) e ex H r C u u* G v y Schematic of the controlled process, with: G incineration process C controller u output controller u* process input y output signal v process disturbance ex excitation signal r reference signal, setpoint Grate combustion 17

26 Validation of dynamic models (4) Waste input as a function of time. Comparison physical simulation model (-) and real plant data (- -). Grate combustion 18

27 Validation of dynamic models (5) 1.2 Step applied to waste inlet flow of 10 [% controller scale] steam production [kg/s] time [min] Comparison dynamic model and plant results Grate combustion 19

28 Grate stoker system Waste input (disturbances) Controller system Combustion process Structure Solid fuel combustion process Grate combustion 20

29 Validation of controller model modelled controller measured Dosage [%] time [min] 70 modelled controller & process 60 Dosage [%] time [min] Grate combustion 21

30 Grate stoker system W aste input (d isturban ces) Controller system Com bustion process Structure MSWC process Grate combustion 22

31 On-line calorific value sensor (1) Changing calorific value of the fuel is one of the main problems in 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 the following measurements: H 2 O, O 2 en CO 2 -concentrations (with IR) Relative humidity of the ambient air Grate combustion 23

32 On-line calorific value sensor (2) Hafval [MJ/kg] tijd [dag] 0.4 Xwater [kg/kg] tijd [dag] 40 Hbrand [MJ/kg] tijd [dag] Calorific value waste, moisture fraction waste and calorific value combustible part as function of time Grate combustion 24

33 On-line calorific value sensor (3) 105 berekend gemeten PHIstoom [t/h] tijd [dag] Calculated and measured steam production as a function of time. Grate combustion 25

34 On-line calorific value sensor (4) 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 26

35 Application of dynamic combustion model Process Analysis Simulator Optimisation of control concepts Grate combustion 27

36 Optimisation of control concept (1) Different possibilities for optimisation of control concept by 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 28

37 Optimisation of control concept (2) 0.12 previous σ = 5.45 tuned σ = P(x) Φ Φ [t/h] steam,actual steam,set AVR plant, optimisation by tuning of control parameters Grate combustion 29

38 Model Predictive Control MPC: based upon measurements from the past, a model of the plant and the control objectives it predicts the plant behavior in the near future with respect to the constraints and boundary conditions of the system. Based upon the control objectives it calculates at every sample time t, the most optimal control actions for the near future. At every time sample t this is repeated. Mathematically: an optimization problem Grate combustion 30

39 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 Grate combustion 31

40 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 Characterisation of N-releaseN from a biomass fuel layer by pot furnace experiments and derivation of N-release N functions Emil Widmann, Selma Zahirovic, Robert Scharler, Ingwald Obernberger Institute for Resource Efficient and Sustainable Systems Graz University of Technology International workshop Modelling and process control of grate furnaces 30 September 2005 Innsbruck, Austria Institute for Resource Efficient and Sustainable Systems Graz University of Technology Overview Scope of work Description of the experimental set-up Experimental results for fuels tested Derivation of the N-release functions Summary and conclusions

41 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Scope 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 NO x precursors Derivation of N-release functions based on experimental data for the purpose of modelling Biomass Institute for Resource Efficient and Sustainable Systems Graz University of Technology q radiation Pot furnace experiments vs. combustion of fuel on the grate q radiation t on grate t burnout h reactor h grate A reactor A slice Biomass air flow air flow 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

42 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Experimental set-up extractive FT-IR thermocouples heater elements section 1 in-situ FT-IR (NH3, CO, CO2, CH4 and H2O) heater elements section 2 oil sealing weight balance insulating firebrick fuel bed (with 6 thermocouples) air flow 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 B5 B1 B2 B B Institute for Resource Efficient and Sustainable Systems Graz University of Technology Fuel analysis ultimate analysis and particle size distribution Bark average value rms abs particle size mass fraction C [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.) [wt% d.b.] 0.57 [%] 8 [mm] [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 [-] MDF average value rms abs particle size mass fraction C [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.) [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 average value rms abs particle size mass fraction C [wt% d.b.] 1.1 [%] < 0.4 [mm] 0.06 [-] H 6.60 [wt% d.b.] 0.5 [%] 0.4 [mm] [mm] 0.13 [-] N 0.06 [wt% d.b.] [%] 0.63 [mm] [mm] 0.26 [-] O (calc.) [wt% d.b.] 0.6 [%] 1.0 [mm] [mm] 0.30 [-] ash 0.20 [wt% d.b.] 0.0 [%] 1.6 [mm] [mm] 0.15 [-] water < 0.10 [wt% w.b.] - [%] > 2.5 [mm] 0.09 [-]

43 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Comparison with FID measurements Release of hydrocarbons measured with extractive FT-IR (for different fuels and combustion conditions) was cross-checked with measurements performed with FID equipment: C release in CxHyOz [mol/s] 3.50E E E E E E-03 FID: Carbon release in CxHyOz FTIR: Carbon release in CxHyOz 5.00E E time [s] Institute for Resource Efficient and Sustainable Systems Graz University of Technology Elemental recovery rates for reference experiments Elemental recovery rate r j relates the measured (flue gas concentration) yield of each element to the total amount of the element in the experiment (fuel analysis): sawdust bark MDF recovery rates [wt%] r-tot r-c r-h r-o

44 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Results reference experiment bark temperatur [ C] time [s] 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) Institute for Resource Efficient and Sustainable Systems Graz University of Technology Results reference experiment MDF sample mass [g] time [s] Temperatur [ C] time [s] 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 )

45 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Results reference experiment sawdust [g] time [s] temperatur [ C] time [s] 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) Institute for Resource Efficient and Sustainable Systems Graz University of Technology Conversion rates for different fuels Conversion rate u i relates the yield of each nitrogen species (with exception of N 2 ) to the total amount of nitrogen in the fuel: 90 nitrogen conversion rate [%] reference experiment sawdust reference experiment bark 70 reference experiment MDF board u NO u NH3 u HCN u NO2 u N2O

46 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Influence of the fuel N content on the total conversion rate for different fuels Conversion rate u TFN relates the yield of all nitrogen species (with exception of N 2 ) to the total amount of nitrogen in the fuel: 120 conversion rate TFN [%] waste wood bark sawdust MDF board fibreboard nitrogen content [wt% d.b.] Institute for Resource Efficient and Sustainable Systems Graz University of Technology Modelled release of N species based on experimental data sawdust experiment u = k + d i i i concentration [ppmv d.b.] NH3 modelled NH3 experiments concentration [ppmv d.b.] model HCN modelled HCN experiments normalised length on grate [-] concentration [ppmv d.b.] normalised length on grate [-] NO modelled NO experiments normalised length on grate [-]

47 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Summary and conclusions I Fuel analysis was performed for bark, MDF and sawdust. Good quality of the experiments performed at the pot furnace was achieved: 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. NH 3 was found to be the main NO x precursor for MDF board and bark. NH 3 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 and Sustainable Systems Graz University of Technology 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 CFD simulations of gas phase combustion and NO x formation.

48 Annex 4. CFD modelling of NOx formation in biomass grate furnaces with detailed chemistry Selma Zahirovic, Graz University of Technology, Austriai CFD modelling of NO x formation in biomass grate furnaces with detailed chemistry Robert Scharler, Selma Zahirovic, Emil Widmann, Ingwald Obernberger Institute for Resource Efficient and Sustainable Systems Graz University of Technology International workshop Modelling and process control of grate furnaces 30 September 2005 Innsbruck, Austria Institute for Resource Efficient and Sustainable Systems Graz University of Technology Overview Scope of work Modelling Empirical fixed bed modelling Modelling of turbulent reactive flow basic combustion modelling CFD NO x postprocessing Test of the CFD NO x postprocessor methodology and discussion of results Simulation of a 440 kw th pilot-scale plant (fibre board as fuel) Simulation of a 7.2 MW th industrial-scale plant (waste wood as fuel) Summary and conclusions

49 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Scope of work Development of a 3D CFD NO x 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 NO x 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 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Empirical fixed bed model basic 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 CH 4, H 2, CO, CO 2, H 2 O, and O 2 in the flue gas released based on literature data and lab-scale experiments Stepwise balancing of mass, species and energy Temperature [K] 1600 Temperature wt% H2O (w. b.) wt% H2O (w. b.) Length on grate [m] Example: Calculated profiles of temperature and H 2 O concentration in the flue gas along the grate 0

50 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Extension of the fixed bed model release of N species The empirical fuel bed combustion model was extended in order to describe the release of N species (NO and NH 3 as well as HCN) which are relevant for the formation of fuel NO x in biomass grate furnaces (fibre board, waste wood, bark) Conversion functions (as a function of local ) were defined for the investigated fuels based on lab-scale pot furnace (batch) reactor experiments; NH 3 showed to be the predominant NO x precursor, HCN was found only in very low concentrations [wt% NH3, HCN - wet flue gas] NH3 HCN NO length on grate [m] [wt%no - - wet flue gas] Example: calculated profiles of NH 3, 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) [wt% NH3, HCN - wet flue gas] NH3 HCN NO length on grate [m] [wt% NO - wet flue gas] Institute for Resource Efficient and Sustainable Systems Graz University of Technology CFD models Modelling of turbulent reactive flow basic combustion simulation Turbulence Realizable k- model Gas phase combustion Eddy Dissipation model (A mag = 0.6, B mag = 0.5) / global methane 3-step mechanism (CH 4, CO, CO 2, H 2, H 2 O und O 2 ) Radiation Discrete Ordinates model Modelling of NO x formation postprocessing mode Eddy Dissipation Concept (EDC) Kilpinen 92 mechanism (50 species, 253 reactions) ISAT (In-Situ Adaptive Tabulation) algorithm for reaction kinetics

51 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Eddy Dissipation Concept Eddy Dissipation Concept (EDC) implementation in Fluent 6.1 based on Gran and Magnussen (1996) Net production rate R i [kg/m 3 s] R i 2 = Y * ~ Y * 3 i i 1 time averaged ( - ) density [kg/m 3 ] * residence time fine structures [s] = f(t k ) = f(, ) modelled length scale of fine structure regions [-] = f(k,, ) modelled Y i Favre-averaged ( ~ ) and fine structure values ( * ) of species mass fraction Y i [-] 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 ( 3 ) Fine structures are treated as ideal reactors (in FLUENT plug flow reactor) => integration of reaction kinetics / closure of equation system CFD model boundary/furnace outlet Institute for Resource Efficient and Sustainable Systems Graz University of Technology Test runs 440 kw th pilot-scale plant kw th Conventional flue gas analysis at boiler outlet (NO x, CO, O 2, CO 2 ) In-situ FT-IR measurement ports I - III (CH 4, CO, CO 2, H 2 O, NH 3 ) case A: port III; case B: port II Temperature measurements (thermocouples T 1 T 3 ) Additionally: data from literature, experience and labscale pot furnace experiments concerning relevant species concentrations (NO, NO 2, HCN, NH 3 ) PCZ primary combustion zone SCZ secondary combustion zone operation data case A case B fuel fibre board fibre board water content wt% d.b. nitrogen content wt% d.b. fuel power related to NCV kw th lambda fuel bed eff lambda primary eff total air ratio flue gas recirculation ratio adiabatic flame temperature C measured NO x emissions ppmv

52 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Simulated NH 3 profiles 440 kw th pilot-scale plant kw th NH 3 is consumed not immediately above the fuel bed but somewhere in the furnace depending on the stoichiometry (earlier for higher ) => confirmation by in-situ FT-IR measurements and pot furnace experiments NH 3 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 prim < 1 Case B prim > 1 NH 3 mole fraction [-] profiles in the symmetry plane of the pilot-scale biomass grate furnace Institute for Resource Efficient and Sustainable Systems Graz University of Technology Simulated NO profiles 440 kw th pilot-scale plant kw th Simulated NO x emissions consisted mainly of NO; NO 2 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 prim < 1 Case B prim > 1 NO mole fraction [-] profiles in the symmetry plane of the pilot-scale biomass grate furnace

53 Institute for Resource Efficient and Sustainable Systems Graz University of Technology NO x emissions 440 kw th pilot-scale plant kw th Very good agreement of measured NO x emissions at boiler outlet and simulations Simulated NO x emissions at furnace outlet are lower for case A => confirmed by conventional flue gas analysis at boiler outlet Case A: simulated NO x emissions based on the release of NH 3 from the fuel bed as predominant NO x precursor (lab-scale experiments) are closer to the NO x measurements at boiler outlet than based on a release of NH 3 and HCN in similar concentrations (literature data) source data empirical fixed bed model NO x emissions [ppmv] case A case B measured calculated TU Graz literature Explanations: Case A prim < 1; case B prim > 1; literature data NH 3 and HCN in concentrations with same order of magnitude; experimental data TU Graz NH 3 predominant species, HCN is negligible Institute for Resource Efficient and Sustainable Systems Graz University of Technology Test runs 7.2 MW th industrial-scale scale plant MW th SCZ SCZ Conventional flue gas analysis at boiler outlet (NO x, CO, O 2, CO 2 ) Additionally: data from literature, experience and lab-scale pot furnace experiments concerning relevant species concentrations (NO, NO 2, HCN, NH 3 ) secondary air flue gas recirculation above the grate PCZ - cooled walls PCZ PCZ biomass fuel bed flue gas recirculation below the grate PCZ primary combustion zone SCZ secondary combustion zone CFD model boundary/ furnace outlet primary air operation data fuel waste wood water content wt% d.b. nitrogen content 1.20 wt% d.b. fuel power related to NCV 7,570 kw th lambda fuel bed eff lambda primary eff total air ratio flue gas recirculation ratio adiabatic flame temperature 1,120 C measured NO x emissions 140 ppmv

54 Institute for Resource Efficient and Sustainable Systems Graz University of Technology NO x emissions 7.2 MW th industrial-scale scale plant MW th Larger deviations between measured and simulated NO x emissions than for the pilot-scale plant Simulated NO x emissions based on a release of NH 3 from the fuel bed as predominant NO x precursor (lab-scale experiments) are closer to the NO x measurements at boiler outlet than based on a release of NH 3 and HCN in similar concentrations (literature data) Simulated NO x emissions decline with reduced temperatures in the primary combustion zone source data empirical fixed bed model note NO x emissions [ppmv] measured 140 calculated TU Graz 233 literature 293 TU Graz lowered temperature peaks 213 Explanations: Literature data NH 3 and HCN in concentrations with same order of magnitude; experimental data TU Graz NH 3 predominant species, HCN in low concentrations; lowered temperature peaks...peak values of mean flue gas temperature in the primary combustion zone were lowered with a damping function Institute for Resource Efficient and Sustainable Systems Graz University of Technology Simulated temperature and NO profiles 7.2 MW th industrial-scale scale plant MW th Very high fine scale temperatures => possible errors of fixed bed modelling and basic combustion simulation with the EDM Simulated NO x emissions decrease with reduced temperatures in the primary combustion zone => predicted hot spots may cause thermal and prompt NO x Profiles of fine scale temperature [ C] (left) and NO mole fraction [-] (right) in the symmetry plane of the industrial-scale biomass grate furnace

55 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Summary I Lab-scale pot-furnace experiments revealed that NH 3 is the dominating species released from the fuel bed for fibre board, waste wood and bark => the results are an important basis for CFD NO x postprocessing 3D simulations of biomass grate furnaces with the new CFD NO x 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, NO 2, HCN, NH 3 ) with measurements under different operating conditions as well as with data from lab-scale experiments, experience and literature Institute for Resource Efficient and Sustainable Systems Graz University of Technology Summary II Pilot-scale plant: very good agreement of NO x 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 NO x 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 NO x formation rates above the fuel bed were too high due to over-predicted flue gas temperatures ( hot spots )

56 Institute for Resource Efficient and Sustainable Systems Graz University of Technology Conclusions The newly developed NO x postprocessor has been successfully tested The NO x 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 NO x 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 NO x formation

57 Annex 5. Biomass combustion on grates and NOx formation mechanisms Claes Tullin, SP, Sweden Biomass Combustion on Grates and NOx-formation mechanisms Claes Tullin Marie Rönnbäck Jessica Samuelsson SP Swedish National Testing and Research Institute Outline Introduction What goes on in a fixed biomass bed? N-conversion in a fixed biomass bed SP Swedish National Testing and Research Institute

58 Do we know enough? Available data and models Boundary conditions Reasonable knowledge Limited knowledge Combustion processes: Drying Nitrogen chemistry Fair knowledge Very limited knowledge Devolatilisation and gas phase combustion SP Swedish National Testing and Research Institute Char combustion 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.. SP Swedish National Testing and Research Institute

59 Part 1 What goes on in a biomass fuel bed? SP Swedish National Testing and Research Institute Videoanalysis of a fuel bed in a 12 MW boiler SP Swedish National Testing and Research Institute

60 Video SP Swedish National Testing and Research Institute SP Swedish National Testing and Research Institute Gas composition in a fixed bed of biofuel - measurements in and above a downward propagating ignition front

61 Propagation of ignition front SP Swedish National Testing and Research Institute SP Swedish National Testing and Research Institute Experimental rig and fuel Fuel: pellets of compressed sawdust diameter 8 mm moisture 11 %

62 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 SP Swedish National Testing and Research Institute Experimental rig ignition front counter-current to the air flow SP Swedish National Testing and Research Institute Grate: 0.35 m x 0.35 m, height 0.7 m

63 Measurement set-up SP Swedish National Testing and Research Institute Results: Measured concentrations for a batch Superficial velocity 0.14 m/s, pellet with moisture 11 % Conc. (Vol-%, wet wet gas) gas) O N2 N2 N2 H2O CO CO2 H2 H2 H2 5 THC CH Time (min) SP Swedish National Testing and Research Institute Nitrogen (Vol-%, wet gas) Nitrogen (Vol-%, wet gas)

64 Results: Measured concentrations in the front Superficial gas velocity 0.14 m/s, pellet with moisture 11 % Conc. (Vol-%, wet gas) Position in conversion front (mm) O H2O CO2 CO H2 THC CH Time (min) SP Swedish National Testing and Research Institute Nitrogen and hydrogen Comparison exp data and mass balance calculations H2, H2, N2 N2 (Vol-%, wet wet gas) N2 H Time (min) Thin lines: results from mass balance Thick lines: measured with bag sampling SP Swedish National Testing and Research Institute

65 Results: Tar Tar (kg/kgs) (kg/kg) Time (min) SP Swedish National Testing and Research Institute Tar (kg tar/kg devolatilized fuel) All hydrocarbons that condense > 190 C) Part 2 NOx-formation mechanisms NOx mechanisms Primary NOx-reduction methods Secondary NOx-reduction methods SP Swedish National Testing and Research Institute

66 NOx-mechanisms Fuel-N oxidation Thermal NOx Prompt NOx SP Swedish National Testing and Research Institute Conversion of Fuel-N to NOx Fuel-N ~20 % 80 % Vol-N Char-N NH i O2 NO NO Fuel-N is the major source for NOx during biomass combustion Important parameters: - Fuel-N content - Temperature N 2 NH i SP Swedish National Testing and Research Institute -O2

67 Oxidation of N2 in air Equilibrium ppm ppm NO O N2 2 NO Temperature [ C] 21 % O Equilibrium concentrations of NO in a gasmixture of O2 and N2. SP Swedish National Testing and Research Institute Thermal NOx Oxidation of air N2 Extended Zeldovich mechanism N + O NO + 2 N N + O 2 NO + O N + OH NO + H ppm/s % O Temperature [ C] Formation of thermal NOx negligeable at T < 1400 C. SP Swedish National Testing and Research Institute

68 Prompt NOx Oxidation of N2 in air What is prompt NOx? 1. Oxidation of nitrogen in air involving hydrocarbon radicals CH + N 2 HCN + N 2. Reaction via N2O N O + M N O + M N O + O NO + NO 2 (3. Thermal NOx at supercritical equilibrium concentrations of radicals) T < 1400 C Negligeable formation of prompt NOx SP Swedish National Testing and Research Institute 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 reactions Protein depolymerisation Char formation Emitted from fuel particles as NH3, HCN, HNCO, NO or N2 Heterogeneous (char catalysed) reactions Influence of inorganic material.. Complex homogenous gas phase chemistry SP Swedish National Testing and Research Institute NOx emissions = NOx formed NOx destroyed

69 Methods for NOx-reduction Primary methods Combustion control Air staging Fuel staging Flue gas recirculation Secondary methods O2 NO NH i NO SCR Selective Catalytic Reduction SNCR Selective Non-Catalytic Reduction N 2 SP Swedish National Testing and Research Institute Measurements in burning fuel bed Suction probe Cooler Major species: O2, H2O, CO2, CO, H2, N2, THC Nitrogen compounds: NH3, HCN, NO, NO2, N2O Heated filter Tar trap Absorption Filter Dry air Mass flow controller Cooler FTIR SP Swedish National Testing and Research Institute CO 2 CO/ CH4 O2 bag Heated sampling line NO THC

70 Low T s indicate that thermal NOx is not important C min Temp in reaction front at different air flows (m/s) SP Swedish National Testing and Research Institute Gas composition in a fuel bed % HCN ppm NH3NO H 2 O THC CO O 2 CO min SP Swedish National Testing and Research Institute

71 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 SP Swedish National Testing and Research Institute 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 SP Swedish National Testing and Research Institute