validation of a cfd based modelling approach to ...

VALIDATION OF A CFD BASED MODELLING APPROACH TO PREDICT COAL COMBUSTION USING DETAILED MEASUREMENTS WITHIN A PULVERIZED COAL BOILER

S. Piffaretti1, A. Abdon1, E. G. Engelbrecht1, M. Orth2, M. Toqan3. 1

CPS Creative Power Solutions AG - 5400 Baden, Switzerland CPS Creative Power Solutions USA - AZ 85259-3774 3 CPS Creative Power Solutions ME - Abu Dhabi, P.O. Box 107950, United Arab Emirates 2

Abstract

In order to optimize the Four Corners pulverized coal fired power plant Arizona Public Service Company carried out a detailed mapping of temperature, CO concentration and excess O2 concentration for a number of configurations of their boiler at several planes. This data set forms an excellent basis to validate a CFD based model of the aerodynamics and chemical processes within the boiler. The modelling approach, sensitivity analysis and correlation with the Four Corners’ measurements are presented. In this way, this validation demonstrates the value of CFD as a tool to be used together with other diagnostic methods for evaluation and improvement of the combustion performance of pulverized coal furnaces.

Keywords

Furnace modelling, CFD simulation, coal combustion, turbulent flames, particle transport modelling.

INTRODUCTION Pulverized coal firing systems have existed for several decades and have served as one of the major work-horses in the energy supply of industrialized countries. The design of coal fired furnaces is however in many cases predominately based on empirical data and experience collected during the years of technology development. Only recently numerical engineering tools have emerged that allow for detailed predictions of the combustion process inside the furnaces. This is due to the increase in computational power and the integration of advanced simulation codes. Today the numerical simulation of reacting flows is an established and powerful method which may be utilized to understand and isolate the mechanisms that govern turbulent combustion. The cost benefit of using simulations to predict the performance and characteristics of engineering designs rather than relying solely on experiments and testing programs is substantial. Simulation results together with semi-empirical relationship and experimental studies can be used to assist qualified combustion and boiler heat transfer specialists in guiding combustion and NOx emission control system design, and assessing vendor claims. Numerical approaches also provide engineers a platform for creativity in pursuit of alternative design solutions.

Combustion models need to be continuously reviewed and refined in order to include more and more physical details in their codes. Comparison against experimental data is usually performed to validate numerical approaches. However, in most cases only mean experimental measurements are obtainable at the outlet of the combustion chamber of an industrial coal furnace. In this study a comprehensive set of internal data of an industrial coal boiler is available for validation purposes. The aim of this work is to use these measurements to test a documented coal combustion model. More precisely, experimental data taken inside the boiler in the upper part of the furnace for two different operating points will be compared against the simulation results. This comparison will demonstrate the suitability of the particular CFD tool in predicting the qualitative aerodynamic behaviour of the coal fired furnace in question.

NUMERICAL SIMULATIONS In this section the characteristics of the numerical method used in this study will be described. First, the geometry and the boundary conditions of the cases simulated in this work will be shown. Then, the details of the combustion model will be discussed.

Computational domain and furnace conditions The furnace studied here corresponds to the one in use at the Four Corners power station (Unit 2). The Four Corners (4C) power plant is one of the largest coal-fired generating stations in the United States. The plant is located on Navajo land in Fruitland, New Mexico. It was the first mine-mouth generation station to take advantage of the large deposits of bituminous coal in the Four Corners region. The plant’s five units generate 2,060 Megawatts (website: www.pnm.com, 2008). Units 1 and 2 are sister units that fire pulverized coal in single reheat, parallel back-pass 170 MWe boilers. The two furnaces are front fired with 3 rows of 6 burners each. In late-1998, Unit 2 received new low NOx burners in order to reduce the emissions of nitrogen oxides. In Figure 1 the CFD geometry and the numerical mesh used in this study are shown. The CFD geometry was designed to match the key dimensions of the Four Corners Unit 2 boiler. A model was initially built of the entire boiler and run using a simplified homogeneous gas model (no particulate tracking). The results of this model, flame shapes and furnace exit parameters, were then compared to those of a half furnace model. It was determined that the deviation between models was slight enough to justify use of a half boiler model. As already mentioned above the coal burners are located in 3 rows (elevation A, B and C). In the horizontal direction the burners are placed very close to each other. In the upper part of the furnace the planes D and E mark the locations where measurements have been collected inside the furnace. The numerical and experimental distributions of the different combustion variables on these planes will be intensively analysed in the next section. The numerical grid employed to compute the flow contains more than 2 million tetrahedral elements and it is refined in the vicinity of each burner in order to capture better the steep gradients that characterize the turbulent coal flames inside the furnace.

Figure 1: Schematic of the symmetric CFD boiler model (right picture) and its numerical grid (left picture).

The combustion air passing through the burners is divided into primary, secondary and tertiary air. The primary air, which carries the pulverized coal, flows into the system perpendicular to the furnace’s walls. The tertiary air flows down the center of the burner, also perpendicular to the furnace wall. The secondary air which flows down the outer annuli of the burner, instead, has a strong swirling motion in order to stabilize the coal flames and control flame length inside the combustion chamber. In this study two different furnace operating conditions are investigated: Test Case 1 (TC1) and Test Case 2 (TC2). Both cases correspond to a boiler under steady-state, full load conditions where the total fuel mass flow rate is kept constant. The fuel composition of the two tests is listed in Table 1. The excess oxygen at the exit of the furnace is different in TC1 and TC2. During TC1 the boiler operated at a higher excess oxygen level (approximately 1% higher) than during TC2. However, the main difference between TC1 and TC2 exists in the different burner operation settings. The secondary air flow through each burner is dependent on the specific register and damper settings. TC2 has identical SA air register and sleevedamper settings for all burners, whereas the TC1 features a “biased” mode with more SA to the most outer burner columns (column 1 in Figure 1). The individual normalized secondary air mass flows that go through each burner of the half furnace are listed in Table 2.

TC1 TC2 Proximate Analysis [% - mass] Fixed carbon 35.0 % 35.0 % Volatiles 29.9 % 29.5 % Ash 20.7 % 21.1 % Moisture 14.4 % 14.4 % Ultimate Analysis [% - mass] C 50.2 % 49.3 % H2 3.9 % 3.8 % N2 1.0 % 1.2 % S 0.9 % 0.7 % O2 8.9 % 9.5 % Table 1: Coal properties of the two test cases.

The swirl generated by the secondary air registers has a crucial impact on the combustion processes as it governs the mixing between fuel and air and thereby influences emissions and flame lengths. In order to investigate the flow conditions under which the low NOx burner operates, single burner CFD simulations have been performed. The single burner results have then been used to extrapolate the furnace’s boundary conditions for the swirling air. Besides the data listed in Table 2, also the tangential and axial velocities at the burner throat have been calculated with the help of these single burner runs. TC1 TC2 Burner Elevation Column 1 Column 2 Column 3 Column 1 Column 2 Column 3 C 1.194 0.976 0.805 0.981 1.014 1.014 B 1.260 0.976 0.786 1.014 1.014 0.981 A 1.260 0.956 0.787 1.014 0.994 0.972 Table 2: Relative secondary air massflow distribution for the two test cases. Values are normalized with the overall averaged massflow.

Coal combustion model The combustion process in a pulverised coal boiler is very complicated. Fuel enters the combustor as a combination of solid particles and gaseous vapour after it undergoes the mechanical process of pulverisation. The solid particles of coal contain combustible species some of which are released as a gas when the particles are heated by the flame and some which burn directly via surface reactions. A proportion of the coal particles consist of ash which will remain after combustion is complete. To model the main processes a two stage approach was used. An Eulerian method was adopted to model the continuous phase while a Lagrangian approach modelled the coal particulate and ash. Full coupling of the two approaches was achieved via an iterative solver.

PPDF

NO

NOx model

Volatiles O2

O2

HCN

HCO NO

PPDF

PPDF N2

PPDF

Char & O2

Volatiles

O2

PPDF

O2

EDM CO

O2

EDM + PPDF

CO2

Figure 2: Reaction Paths Modelled (EDM – Eddy Dissipation Model, PPDF – Presumed Probability Density Function Model, EDM + PPDF – rate determined as the slower of the two models)

The devolatilization of the coal was modelled following the approach of Badzioch and Hawksley (1997), which assumes that the rate of devolatilization can be related to a single expression in Arrhenius form. During the devolatilization process a gas consisting of CO2, CO, H2, H2O, HCN, hundreds of different hydrocarbons, etc as well as tars and hydrocarbon liquids (Smoot et al, 1985) will be released. To simplify this it is assumed that the vapours released during devolitisation consist of H2O, HCN and a Fuel Gas which subsequently reacts to form CO. The tars and liquids formed were neglected. The oxidation of the remaining char was modelled following the approach of Field (1969, 1970, Wall et al, 1976). The reaction rate was given as a function of the diffusion rate of oxygen to the surface of the particle and the chemical reaction rate at the particle surface. The slower of these two rates dominates the overall rate. For the Eulerian models time averaged Navier-Stokes equations were solved. Reynolds stresses were closed using the standard k-ε turbulence model. The way in which turbulence chemistry interaction was modelled was dependent on whether the reaction rates were considered to be fast compared to the aerodynamic time scales such as for the oxidation of carbon or slow such as for the formation of thermal NOx. For fast reactions, chemical rates were closed using the standard Eddy Dissipation model of Magnussen et al (1976) which limits the mean reaction rate to the slowest mixing rate of fuel, oxidiser and products. Thus the chemistry rates, which are considered to be infinitely fast, are ignored. For slow reactions, chemical rates were closed using a rate expression in Arrhenius form. This was then time averaged using a presumed probability density function (PDF) closure similar to that first proposed by Mao et al (1970). A β function which was parameterised by the first two moments of the temperature was used to define the PDF similar to the model described by Borghi (1988). For reactions which were intermediate in speed a combined model using both approaches was adopted which would take the slowest reaction rate of either the Eddy Dissipation model or the presumed PDF model.

A plot of the reaction paths considered and the combustion model applied to these paths can be seen in Figure 2. In addition to the main heat release a model for the oxidation of nitrogen was also implemented. Fuel bound nitrogen is one of the greatest sources of NOx in coal burning furnaces. The model established by De Soete (1974), Fenimore (1978) and Smoot et al (1985) for the consumption of HCN forms the basis of the NOx model which was implemented (Figure 2). Additional source terms for the oxidation of N2 following either the prompt (Fenimore, 1970) or thermal (Zeldovich, 1946) route were included. For Thermal NOx, because of the low temperatures in the boiler, oxygen equilibrium was assumed and chemical rates were taken from Westenberg (1975). For prompt NOx the single step model proposed by De Soete (1974) was adopted. Both of these sources were averaged using the presumed PDF approach. The discrete transfer model of Shah (1979) was considered to model radiation within the boiler. A gray spectral model, which assumes that the radiation quantities are uniform throughout the spectrum, was adopted. Gas emissivity was set equal to a function of temperature and species concentration based on the work of Hemsath (1969) and Zinser (1985).

RESULTS AND DISCUSSION This section presents the comparison between the results of the CFD calculations and the experimental data measured in the upper part of the furnace. In Figure 3 and Figure 4 the normalized temperature, oxygen and CO mass fraction distributions in the upper part of the furnace are shown. In Figure 3 the distributions on the horizontal plane D (see Figure 1) are depicted, while in Figure 4 the experimental and numerical data on the vertical plane E (see Figure 1) can be seen. The numerical results are shown in the two figures with contour plots. The experimental values measured in the furnace are displayed with the help of small squares. The same colour scales have been used to compare the experimental and numerical results. The symmetry line (CL) of the furnace lies on the left side of the contour plots. The numerical temperature distributions are in good agreement with the experimental data and demonstrate that the CFD model presented in this study is definitely able to capture the trends observed in the measurements. On the horizontal plane (plane D, Figure 3) low temperatures are observed along the front furnace walls where the radiant, superheaters are located. Then, as we move towards the back walls the temperatures rapidly increase and in the centre of the furnace the highest temperature spots are observed. On the vertical plane, instead the low temperature zones for both test case 1 and 2 are located at the top and sides of the furnace and as we move down the temperatures steadily increase. The measurements of the oxygen mass fraction inside the furnace show an inverse trend with respect to the temperature fields. Along the front wall, over the burners, high concentrations of O2 are observed (Figure 3) and these peak concentrations rapidly decrease as we approach the centre of the furnace. On the other hand, close to the vertical furnace exit (Figure 4) high O2 mass fractions are located in the upper part

of the cross section. In other words, in the upper part of the furnace peak temperatures are associated with low O2 concentrations and high O2 mass fractions correspond to relatively low temperatures. These trends are clearly shown also in the numerical distributions, though the peak O2 mass fractions are slightly overestimated in the CFD. The formation of such an oxygen blanket that runs along the furnace front wall is a typical feature of front wall fired furnaces. This phenomenon is well depicted in Figure 5 where the normalized O2 concentration for TC2 on a vertical plane is shown. This plane is taken at the centreline of the second column (see Figure 1). The contours clearly show how the oxygen is consumed in front of the burners in the coal flames. However, they also show how some amount of air escapes the reaction zones and does not mix immediately with the fuel gases in the flames. This fact is particularly true right above the last burner row. In this case the oxygen that does not participate directly in the combustion travels along the front walls until it reaches the top of the furnace. This creates a high oxygen region observed in both the numerical and experimental data of Figure 3 and Figure 4. This characteristic of the flow is undoubtedly an aspect that will be evaluated for improvement. The total amount of secondary air should be directly involved in the main reaction zones in front of the different burners and is a challenge to achieve. Future parametric studies of furnace and single burner simulations should be performed in order to further evaluate and propose potential burner settings that prevent these inefficiencies in the coal flames. Both the numerical and the experimental O2 distributions in Figure 3 and Figure 4 visibly show the qualitative differences between the Test Case 1 (TC1) and the Test Case 2 (TC2). As already stated in the previous section, TC2 has an even distribution of the burners’ secondary air. This operating condition leads to an inhomogeneous distribution of the air along the furnace front wall (Figure 3) and along the roof of the boiler (Figure 4). The oxygen mass fraction in the furnace is clearly lower in the regions near the side wall (reference Figure 3 and Figure 4) and in the middle of the furnace, close to the centreline of the boiler, a higher amount of O2 is observed. The CFD simulations reflect this behaviour well for both the horizontal and vertical planes. TC1, instead, presents a slightly different aerodynamic behaviour. In this case, in order to better balance the temperature distribution at the furnace exit and prevent local overheating of the Superheater, the external burners are run leaner. In this way more air is injected close to the side walls and the O2 distribution becomes more balanced. Also in this case the numerical model predicts the behaviour of the furnace very well. Changing the air distribution inside the furnace means changing the temperature distribution. In fact the same comments made above for the O2 mass fraction are also valid for the inverse of the temperature distribution. The CFD model captures the differences between TC1 and TC2 especially well on the horizontal plane (plane D). In TC2, the horn of the high temperature spots in plane D is longer and extends itself towards the corner between the front and side walls leading to an uneven temperature distribution at the surfaces of the superheater. The combustion model seems to overestimate the absolute values of the CO mass fraction inside the furnace (reference Figure 3 and Figure 4). Nevertheless, the high CO spots are located in the right position and coincide with the high CO regions measured experimentally.

TC1

TC2

Figure 3: Normalized Temperature, Oxygen and CO distributions at plane D. Squares represent experimental data. (Left: data of TC1, Right: data of TC2)

TC1

TC2

Figure 4: Normalized Temperature, Oxygen and CO distributions at plane E. Squares represent experimental data. (Left: data of TC1, Right: data of TC2)

Figure 5: Normalized oxygen mass fraction in TC1.

Figure 6: CO contours on plane D and coal particles streamlines for TC1. Streamlines are coloured to show the mass fraction of unburned carbon contained in the particles.

It is interesting to observe the strong correlation between the CO and the amount of unburned carbon in the solid particles. Figure 6 shows the three dimensional furnace of TC2 (viewed from above). In this picture the CO distribution at plane D (the same distribution as depicted in the lower right plot of Figure 3) are shown together with the coal particle streamlines. These streamlines are coloured according to the percentage of unburned carbon attached to the ash particles. It can be observed that at plane D the particles that have high carbon content are located in the corner between the back and the side wall of the furnace. This is exactly the position where high CO concentration can be found in the numerical and the experimental results. High CO spots could be seen as an indicator for rich conditions inside the furnace as little oxygen is available to oxidize completely the fuel. Because of the lack of available O2, high CO concentrations may also be associated with higher than normal levels of unburned carbon in the ash. In addition, CO concentrations are also useful to evaluate the potential risk of slagging inside the furnace. Rich conditions generally make the slagging problem worse inside a furnace. The iron compounds of the ash are usually responsible for the

sticky behaviour of the particles. Under rich environments, these compounds can not be fully oxidized and they make the ash stickier.

CONCLUSIONS The above investigation has demonstrated the successful application of a CFD code to evaluate the main combustion and aerodynamic characteristics of a front wall fired pulverized coal furnace. It has been shown that the predictions are in good agreement with the field measurement data collected from the simulated unit. In particular, two different furnace operating conditions have been analyzed in this work. The CFD model is able to follow the trends observed experimentally between the two cases. The relative distributions at the furnace exit of the critical species of oxygen and CO match the sample data patterns. Also the predicted temperature field at this location is well reproduced by the model in both situations. The simulations have provided insight to the main aerodynamic mechanisms that govern the furnace exit conditions where the furnace measurement data has been collected. The improved understanding of the physics demonstrates the value of CFD as a tool to be used together with other diagnostic methods for evaluation of the combustion performance of pulverized coal furnaces. The described CFD model is currently used to evaluate modifications of the combustion system with the target to reduce emissions and to improve temperature distributions at the heat exchangers.

ACKNOWLEDGEMENTS The authors are grateful to the Arizona Public Service (APS) company for the excellent cooperation during this project, to J. Lee for providing the experimental data and to Prof. A. Benim, Dr. P. Stopford and Dr. S. Srinjvasachar for their helpful discussions.

REFERENCES Badzioch, S., Hawksley, P.G.W., “Kinetics of thermal decomposition of pulverised coal particles”, Industrial Engineering Chemistry Process Design and Development, V.9, p. 521, 1997. Borghi, R., “Turbulent Combustion Modelling”, Progress in Energy and Combustion Science, V. 14, pp. 245-292, 1988. De Soete, G.G., “Overall Reaction Rates of NO and N2 Formation from Fuel Nitrogen”, 15th Symposium (International) on Combustion, The Combustion Institute, pp. 1093-1102, 1974. Fenimore, C.P., “Formation of nitric oxide in premixed hydrocarbon flame”, 13th Symposium (International) on Combustion, The Combustion Institute, pp. 373-380, 1970. Fenimore, C.P., “Studies of Fuel-Nitrogen Species in Rich Flame Gases”, 17th Symposium (International) on Combustion, The Combustion Institute, pp. 661-670, 1978.

Field, M.A., “Rate of combustion of size-graded fractions of char from a low-rank coal between 1200 K and 2000 K”, Combustion and Flame, V. 13, pp. 237-252, 1969. Field, M.A., “Measurements of the effect of rank on combustion rates of pulverized coal”, Combustion and Flame, V. 14, pp. 237-248, 1970. Hemsath, K.H., “Zur Berechnung der Flammenstrahlung”, Dissertation, University of Stuttgart, 1969. Magnussen, B.F., Hjertager, B.H., “On Mathematical Modeling of Turbulent Combustion with Special Emphasis on Soot Formation and Combustion”, 16th Symposium (International) on Combustion, The Combustion Institute, pp. 719-729, 1976. Mao, K.W., Toor, H.L., “A diffusion model for reactions with turbulent mixing”, AICHE Journal, V. 16, pp. 49-52, 1970. Shah, N.G., “New Method of Computation of Radiant Heat Transfer in Combustion Chambers”, Ph.D. Thesis, Imperial College, London, United Kingdom, 1979. Singer, J.G., “Combustion Fossil Power: A Reference Book on Fuel Burning and Steam Generation”, Combustion Engineering, Windsor, 1991. Smoot, L.D., Smith, P.J., “Coal Combustion and Gasification”, Plenum Press, London, 1985. Wall, T.F., Phelan, W.J., Bartz, S., “The prediction of scaling of burnout in swirled pulverised coal flames”, International Flame Research Foundation Report F388/a/3 Ijmuiden, The Netherlands, 1976. Westenberg, A. A., Combustion Science and Technology, N. 4, p. 59, 1975. Zeldovich, Y.B., “The oxidation of Nitrogen in combustion and explosions”, Acta Physicochemica, USSR, V. 21, pp. 577-628, 1946. Zinser, W., “Zur Entwicklung mathematischer Flammenmodelle für die Verfeuerung technischer Brennstoffe”, VDI-Berichte, Reihe 6, N. 171, VDI Verlag, Düsseldorf, 1985.