Acta Polytechnica doi:10.14311/AP.2014.54.0068 Acta Polytechnica 54(1):68–73, 2014 © Czech Technical University in Prague, 2014 available online at http://ojs.cvut.cz/ojs/index.php/ap NOX PREDICTION FOR FBC BOILERS USING EMPIRICAL MODELS Jiří Štefanica∗, Jan Hrdlička CTU in Prague, Faculty of Mechanical Engineering, Department of Energy Engineering, Technická 4, 160 00 Prague 6 ∗ corresponding author: jiri.stefanica@fs.cvut.cz Abstract. Reliable prediction of NOX emissions can provide useful information for boiler design and fuel selection. Recently used kinetic prediction models for FBC boilers are overly complex and require large computing capacity. Even so, there are many uncertainties in the case of FBC boilers. An empirical modeling approach for NOX prediction has been used exclusively for PCC boilers. No reference is available for modifying this method for FBC conditions. This paper presents possible advantages of empirical modeling based prediction of NOX emissions for FBC boilers, together with a discussion of its limitations. Empirical models are reviewed, and are applied to operation data from FBC boilers used for combusting Czech lignite coal or coal-biomass mixtures. Modifications to the model are proposed in accordance with theoretical knowledge and prediction accuracy. Keywords: NOX prediction, FBC, empirical models. 1. Introduction Fluidized bed combustion (FBC) technology provides an efficient and ecological way for low quality fuel combustion. The fuel is combusted in a bed of inert material that is brought into a fluidized state by pass- ing through air, which leads to very intensive mixing of gases and solids inside the bed. A high degree of mixing enhances the heat and mass transfer by orders of magnitude compared to other combustion technologies. Intensive heat transfer is beneficial for keeping the combustion temperature low and uniform throughout the bed. The mass transfer helps to keep high combustion efficiency even for low-quality fuels, and facilitates emission control. Nitrogen oxides NO and NO2 (referred to as NOX) are pollutant gases that cause photochemical smog, respiratory problems and damage to organisms. Emis- sions of these gases are therefore monitored and must be kept at a minimal level. Although transportation (internal combustion engines) is the major source of NOX, control of NOX emissions is efficient only in stationary combustion sources. Under typical combus- tion conditions of a solid fuel, about 95 % of the total NOX is in the form of NO, and just 5 % is in the form of NO2, which is much more noxious. Emissions of nitrogen oxides are influenced by fuel properties, com- bustion conditions and combustor design. Authors of NOX prediction models for FBC usually combine a kinetic modelling approach with FBC hydrodynamic models. However, all existing models suffer from in- accuracy, overcomplexity, or both. A much simpler approach can be found for pulverized coal combus- tors (PCC), where the application of empirical models leads to very simple correlations that can achieve good agreement with experiments. However, these correlations are used exclusively for PCC, and no ref- erence is available for modifying this method for FBC conditions [1, 2]. 2. Theory 2.1. Formation of nitrogen oxides Many authors have already written in detail about the formation of nitrogen oxides, see e.g. [2]. In general, there are three mechanisms of NO formation that are generally accepted: thermal, prompt and fuel. NO2 is formed through oxidation of NO by HO2 radicals that are present in low temperature regions of the flame. N2O is formed from NO by reaction with NCO or ammonia radicals. In FBC conditions, the vast majority of NO has its origin in fuel. Thermal and prompt NO formation mechanisms are insignificant, due to the low tempera- ture in a fluidized bed. A further reduction compared to PCC can often be achieved, because the most of the NO is reduced to N2 or N2O. Homogeneous re- duction occurs both in the freeboard and in the bed by reaction with CO and volatiles. Heterogeneous reduction takes place on the surface of devolatilized char particles inside the bed. Ash and bed material can have a catalytic effect on NO reduction. The low combustion temperature enhances the reduction of NO to N2O (with the exception of biomass combus- tion, which is for example a case of waste combustion) [2, 3]. 2.2. Prediction of NOX for FBC boilers The complexity of NOX chemistry and the large num- ber of influencing parameters make an accurate pre- diction very difficult. The most common approach for predicting the emissions of nitrogen oxides of FBC boilers is kinetic modelling combined with a detailed 68 http://dx.doi.org/10.14311/AP.2014.54.0068 http://ojs.cvut.cz/ojs/index.php/ap vol. 54 no. 1/2014 NOX Prediction for FBC Boilers Using Empirical Models Premixed Diffusion Staged flame flame combustion k1 285 340 150 k2 1280 835 80 k3 180 20 −30 k4 −840 −395 100 Table 1. Pohl’s correlation coefficients for PCC boilers. FB hydrodynamic model of the bed and freeboard. Recent models taking into account all occurring phe- nomena contain hundreds of reversible chemical reac- tions, and divide the bed into control volumes that can respect different flow patterns and hydrodynam- ics in different parts of the bed. These arrangements increase the complexity beyond acceptable limits. An undisputed advantage of kinetic models is the prediction of nitrogen oxide concentration profiles through the bed and freeboard, which can be used to identify and validate the detailed chemistry. However, this information is not necessary for predicting stack emissions. [2] [4] [5] 2.3. Empirical modelling approach The main advantage of empirical models is their sim- plicity. The data required is usually easy to obtain through proximate and ultimate analysis of the fuel and combustion parameters. By contrast with kinetic models, there is no need to solve an extensive equa- tion system, or to have high for computation capacity available. The prediction is based on experimentally derived correlations accounting for the dependency of the emis- sions on the influencing parameters. The parameters that have been identified to have the largest influ- ence, and that are used in empirical models, can be classified within three groups: • Fuel related (nitrogen content, volatile matter con- tent, etc.) • Boiler design related (staged/ unstaged combustion, extent of fuel - air mixing, etc.) • Boiler operation related (excess air, combustion temperature, etc.) The influence of individual parameters can be observed experimentally by keeping the other parameters con- stant. However, this approach presumes independent effects of the parameters, and this is not necessarily valid for all fuels and combustion conditions. The main disadvantage of empirical models is un- certainty originating from lack of input data, e.g. ash composition can promote NO reduction under certain conditions, petrographic composition can significantly influence the devolatilization and char formation pro- cess. Another consideration is the extent of mixing of fuel and combustion air. To minimize the uncertainty, Figure 1. Pohl’s correlation coefficients for PCC boilers [9]. correct parameters must be used in the model in order to cover all important factors, and at the same time not to increase the complexity. Influencing parameters not included in the input data are taken into account via constants, and their applicability determines the limitations of the model. Input parameters and selection of constants should be carefully considered. Nevertheless, deviations in NOX concentration can be measured in the flue gas stream due to inhomogeneity, so prediction reliability of ±50 ppm can be considered acceptable. From the models found in the literature, only Pohl’s and Ibler’s were chosen and applied to boiler data, because they assume general applicability [1, 6]. 2.4. Pohl’s model A simple correlation was developed by Pohl et al. [1] to estimate NO emissions for controlled mixing conditions (various types of PCC flames, cf. Table 1): NO[ppm] = k1 + k2 N daf 1.5 + k3 VM 40 NOeq3200 + k4 FC 60 NOeq 3200 , (1) where NOeq [ppm] is the maximum emission of NO provided that all fuel nitrogen converts to NO, N daf [%] is the nitrogen content in combustible, VM [%] is combustible volatile matter, and FC [%] is the fixed carbon content. NOeq can be calculated from the nitrogen content in the fuel and dry flue gas volume 69 Jiří Štefanica, Jan Hrdlička Acta Polytechnica Steam output 125 t/hour Steam temperature 490 °C Steam pressure 7.3 MPa Table 2. Steam nominal parameters for Komořany I CHPP. LHV 13 MJ/kg W r 28 % Ar 25 % N daf 1 % Table 3. Fuel parameters for Komořany I CHPP. Vfd [N m3/kg], ash content Ar [–] and water content W r [–], by NOeq[ppm] = 2.1422N daf (1 − Ar − W r) Vfd · 106. (2) A different set of constants will presumably be needed for FBC conditions. As can be seen from Figure 1, only three combustion regimes are accounted for, and other fuel-air mixing regimes are not defined. Pohl’s model was constructed on the basis of a wide range of experimental data from PCC boilers (diffusion flame) [1, 6–8]. 2.5. Ibler’s model Ibler et al. [10] proposed the following correlation for predicting fuel nitrogen conversion to NO: NO NOmax [–] = 7 · 10−5K CO2 3 √ T − 1025, (3) where K [–] is a fuel related constant (Ibler recom- mended using values of constant K between 4 and 6 for Czech coals), CO2 [%] is the flue gas oxygen con- centration and T [K] is the combustion temperature. The predicted concentration in ppm can be calculated by multiplying the fuel nitrogen conversion by NOeq from Equation (2). The constant 7 · 10−5 in Equation (3) represents the PCC conditions, and a different constant will presumably be needed for FBC conditions. As can be seen from Equation (3), Ibler’s model is targeted more on combustion conditions than on fuel properties, which are characterized by constant K only. 3. Experimental The main aim of this paper is to make an evaluation of real measurement NOx emissions data from two large- scale fluidized bed boilers, and to make a comparison with the NOx levels predicted by Pohl’s model and by Ibler’s model. Steam output 140 t/hour Steam temperature 535 °C Steam pressure 12.5 MPa Table 4. Steam nominal parameters for Mladá Bole- slav CHPP. Hard Lignite Biomass coal coal pellets LHV 24.31 18.77 15.23 W r 13.2 28.18 13.66 Ar 11.67 6.37 4.51 N daf 0.89 1.38 1.9 Table 5. Fuel parameters for Mladá Boleslav CHPP. 3.1. Komořany I CHPP The K3 FBC boiler with a bubbling bed at the Ko- mořany I combined power plant was used as the first reference. The lower part of the combustion cham- ber containing the bed is lined and contains an in- bed evaporator. The upper part contains wall and grid parts of the evaporator. The convection part, which follows the combustion chamber, contains the superheaters (primary, secondary and output) and the economizer. A tube-type air heater with a sep- arate part for fluidization and secondary air is the last heat transfer surface of the boiler. The boiler is equipped with a bed material recirculation system as well as bed height control. The combustion process is controlled by the fluidization air flow rate and the fuel input. The steam parameters are adjusted by feed water injection before the last superheater. The steam nominal parameters are shown in Table 2. The lignite coal used was analysed before each combustion test. The results of the analysis were coupled with the NOX emissions for the model predictions. The average parameters of coal are shown in Table 3. 3.2. Mladá Boleslav CHPP The K90 FBC boiler with a circulating bed at the Mladá Boleslav combined heat and power plant was used as the second reference. This boiler is designed for hard coal combustion, but the recent fuel is a mixture of hard and lignite coal with the addition of biomass. The combustion chamber with a lined lower part contains the membrane-wall type evaporator. Af- ter the combustion chamber there is a cyclone for coarse particle separation. The second duct contains a membrane wall, tube and wall type superheaters and an economizer, followed by a hopper. The third duct contains a tube-type air preheater, which is the last heat transfer surface of the boiler. The steam pa- rameters are controlled by feed water injection before the second and last superheater. The steam nominal parameters are shown (for hard coal) in Table 4. The 70 vol. 54 no. 1/2014 NOX Prediction for FBC Boilers Using Empirical Models Komořany Mladá Boleslav Oxygen concentration after economizer [%] 3.8–4.7 4–5 Fluidized bed temperature [°C] 815–866 873 Boiler load [%] 100 75–100 Fuel mixture by mass [%] – lignite coal, biomass, hard coal 100, 0, 0 40–85, 0–25, 0–50 CO concentration [mg/m3] 79–222 Table 6. Combustion parameters. Figure 2. NOx prediction reliability using the original Pohls method. hard coal, lignite coal and biomass that were used were analysed before each combustion test. The results of the analysis were coupled with the NOX emissions for the model predictions. The average fuel parameters are shown in Table 5: 4. Results Experimental data from combustion tests on these boilers was taken from [11] and [12]. Combustion tests were carried out in these boilers covering the combustion conditions described in Table 6. 4.1. Pohl’s model results The coefficients for the staged combustion model were adopted as a basis for NOX prediction using Pohl’s model for FBC boilers. Three options were explored. The first option used Pohl’s original model, as it was proposed by the authors in Equation (1). As expected, the reliability was very low, see Figure 2. In the second option, coefficient k1 was optimized by the least squares method for a better fit with the x = y line. The best fit with determination index R2 = 79.43 % was found for k1 = 19.87 — see Figure 3. The third option incorporated the temperature and the excess oxygen dependency proposed by Ibler into coefficient k1: k1 = 0.17C 2O2 3 √ T − 1025. (4) The modified constant k1 was consecutively optimized by the least squares method for the best fit with the Biomass −2.2 Hard coal 3.1 Lignite coal 1 5 Lignite coal 2 3 Table 7. Values of fuel constant K. x = y line, see Figure 4. The modified version of Pohl’s method showed slightly better results with the determination index R2 = 79,54 %. 4.2. Ibler’s model results The prediction results using Ibler’s model were not in good agreement with the measured data, see Figure 5. To increase the reliability, the combustion constant was modified from 7 · 10−5 to 2.9 · 10−4 and fuel constants K were optimised to the values presented in Table 7, in both cases using the least squares method fitting the x = y line. See the results in Figure 6 with R2 = 81.3 %. 5. Conclusions This paper has discussed the advantages and limita- tions of empirical prediction of NOX emissions from the FBC boilers. With careful choice of input param- eters and constants, empirical modelling can be in very good agreement with experimental data while 71 Jiří Štefanica, Jan Hrdlička Acta Polytechnica Figure 3. NOx prediction reliability using Pohl’s method with modified constant k1 Figure 4. NOx prediction reliability using Pohl’s method with incorporated temperature and excess oxygen dependency. keeping the model simple and the input data easy to obtain. Given the inhomogeneity of the flue gas stream, prediction accuracy of ±50 ppm can be con- sidered reliable. For FBC conditions, neither Pohl’s model nor Ibler’s model for PCC boilers provided satisfactory results without modifications. The models were adapted by least squares methods to fit the experimental data from two FBC boilers, in Komořany (combusting lig- nite coal) and in Mladá Boleslav (combusting a coal- biomass mixture). The modified Pohl model for staged combustion with k1 = 19.87, which uses only fuel parameters as input data, shows relatively good agreement with the measured data with R2 = 79.43 %. The prediction accuracy increases to R2 = 79.54 % with the adoption of a modification to coefficient k1 for temperature and excess air dependency taken from Ibler’s model. However, most of the predicted value originates from the other constants, so the NOX prediction is limited to a quite narrow range (cca 80–120 ppm), irrespective of the combustion conditions, and does not follow the measured trend. Ibler’s model, which focuses more on combustion parameters (temperature and oxygen concentration), and accounts for fuel properties by constant K only, shows better agreement, with R2 = 81.3 % for the combustion constant 2.9 · 10−4 and fuel constants K taken from Table 7. The predicted NOX emissions are from a much wider range (50–180 ppm), and seem to follow the experimentally observed trend well. The prediction results from the modified models are in almost all cases within the 50 ppm limit, and can be considered reliable. However, the modified Ibler model has higher prediction accuracy and seems to be more suitable for FBC conditions. References [1] J. POHL, S. CHEN, M. HEAP and D. PERSHING: Correlation of NOx emissions with basic physical and chemical characteristics of coal, Proc. Joint Symposium on Stationary Combustion NOx Control, 1983 [2] J. A. MILLER and C. T. BOWMAN: Mechanism and modeling of nitrogen chemistry in combustion, Prog. Energy Combust. Sci., vol. 15, pp. 287–338, 1989 72 vol. 54 no. 1/2014 NOX Prediction for FBC Boilers Using Empirical Models Figure 5. NOX prediction reliability using Ibler’s method. Figure 6. NOX prediction reliability using Ibler’s method with optimized constants. [3] J. Kers, K. Priit, A. Aruniit, V. Laurmaa, P. Križan, L. Šooš, Ü. Kask: Determination of physical, mechanical and burning characteristics of polymeric waste material briquettes, Estonian Journal of Engineering. Vol. 16, No. 4, pp. 307–316, ISSN 1736-6038, 2010 [4] D. KUNII and O. LEVRNSPIEL: Fluidization engineering, Butterworth-Heinemann, 1991 [5] A. GUNGOR: Prediction of SO2 and NOx emissions for low-grade Turkish lignites in CFB combustors, Chemical engineering journal 146, pp. 388–400, 2009 [6] J. POHL, G. DUSATKO, P. ORBAN and R. MCGRAW: The influence of fuel properties and boiler design and operation on NOx emissions, Joint symposium on stationary combustion NOx control, pp. 24-1–24-28, 1987 [7] L. JUNIPER and D. HOLCOMBE: Formation and control of NOx emissions from coal fired boilers, AIE seminar on clean use of coal, 1992 [8] P. BENNET: NOx prediction research report 20, Final report of the Cooperative Research Centre for Black Coal Utilization, 2001, http://www.ccsd.biz/publications/files/RR/RR% 2020%20NOx%20prediction.pdf [9] J. Pohl et al.: Correlation of NOx, Emissions with Basic Physical and Chemical Characteristics of Coal. Proc. Joint Symposium on Stationary Combustion NOx Control. EPRI CS-3182, Vol. II., 1983. [10] Ibler, Z., Karták J., Mertlová J., Ibler Z.: Technický průvodce energetika, Nakladatelství BEN, Praha. [11] M. MÁK: Posouzení změny paliva u fluidního kotle, CTU in Prague, Faculty of Mechanical Engineering, Department of Energy Engineering, diploma thesis, 2011. [12] T. DLOUHÝ et al.: Spalovací zkoušky na fluidním kotli K3, unpublished report. 73 http://www.ccsd.biz/publications/files/RR/RR%2020%20NOx%20prediction.pdf http://www.ccsd.biz/publications/files/RR/RR%2020%20NOx%20prediction.pdf Acta Polytechnica 54(1):68–73, 2014 1 Introduction 2 Theory 2.1 Formation of nitrogen oxides 2.2 Prediction of NOX for FBC boilers 2.3 Empirical modelling approach 2.4 Pohl's model 2.5 Ibler's model 3 Experimental 3.1 Komorany I CHPP 3.2 Mladá Boleslav CHPP 4 Results 4.1 Pohl's model results 4.2 Ibler's model results 5 Conclusions References