Development and application of an acid rain precursor model for

May 21, 1993 - L. Douglas Smoot,* Richard D. Boardman,* 1. B. Scott Brewster,1 Scott C. Hill,1 and. A. Kwasi Foli§. Advanced Combustion Engineering ...
1 downloads 0 Views 1MB Size
Energy & Fuels 1993, 7, 786-795

786

Development and Application of an Acid Rain Precursor Model for Practical Furnaces L. Douglas Smoot,* Richard D. Boardman,+ B. Scott Brewster,$ Scott C. Hill,$ and A. Kwasi Folio Advanced Combustion Engineering Research Center, 75 CTB, Brigham Young University, Provo, Utah 84602 Received May 21, 1993. Revised Manuscript Received September 8, 1993"

Control of emissions of sulfur (SOz, so3, HzS) and nitrogen (NO, NOz, N20, HCN, NH3) pollutants from fossil-fuel-fired furnaces and gasifiers remains a vital worldwide requirement as the utilization of fossil fuels continues to increase. Development and refinement of a predictive model for these acid rain precursors (MARP) has reached the point where this technology can contribute to acid rain control. In this paper, model foundations and recent developments are summarized, including formation of thermal and fuel NO, and sorbent capture of sulfur oxides. The method includes global formation, capture, and destruction processes in turbulent, reacting, particle-laden flows. This submodel has been combined with comprehensive, generalized combustion models (PCGC-2, PCGC3) which provide the required local properties for the combustion or gasification processes. The submodel has been applied to NO, formation in a full-scale (85 MW,), corner-fired utility boiler, where recent in situ NO, measurements were made, with variations in coal feedstock quality (including fuel N percentage) load-level and percentage excess air. Predictions are also made for in situ sorbent capture of sulfur pollutants in both combustion (fuel-lean, Sod, and gasification (fuel-rich, HzS) laboratory-scale reactors. Limitations of MARP are identified and work to improve the submodel is outlined.

Introduction Coal contributes about 57% of the electric power generation in the US. through combustion and gasification processes.' An estimated one billion short tons of coal per year are produced in the and projections suggest that coal use will increase by an estimated 43% in the next 15years.' Projections of increasing coal use are based on substantial reserves (coal is 94% of the known US. fossil fuel reserves)l, stable prices, and improving technology for clean and efficient use. Coals typically contain about 1% of nitrogen and 1-6 % of sulfur.3 During combustion or gasification, these fuel species are converted to nitrogen-containing (NO,) pollutants (NO, NOz, NzO, HCN, NH3) and sulfur-containing (SO,) pollutants (SOz, SOa, HzS, COS, CSz). The species formed depend principally on the fuel/oxygenratio. When emitted into the atmosphere, some of these species lead to acidic compounds such as sulfuric and nitric acids, which can be absorbed into atmospheric moisture and deposited to the earth's surface as acid rain.4 Detrimental effects of acid rain on plant and animal life and human health are well do~umented.~Control of these species through postcleanup processes is commercially viable but more

* Author for correspondence. Phone: (801)378-4024. Fax: (801)3783831. Dean, Engineering and Technology, and Director, ACERC. t Dodoral Candidate, currentlywithWestinghouseIdaho Nuclear Corp. t Research Associate, ACERC. 1 Post-Doctoral Associate, ACERC. Abstract published in Advance ACS Abstracts, October 15, 1993. (1)Siegel, J. Energy World 1992, 196, i. (2) National Data Book, 112thEditionStatisticalAbstract ofthe United States, U S . Department of Commerce, Washington, DC, 1992. (3) Boardman, R. D.; Smoot, L. D. In Fundamentals of Coal Combustion; Smoot, L. D., Ed.; Elsevier: Amsterdam, 1993; Chapter 6. (4) Longhurst, J. W. S.; Raper, D. W.; Lee, D. S.; Heath, B. A.; Conlan, B.; King, H. J. Fuel 1993, 72, 1261. Q

0887-0624/93/2507-0786$04.00/0

expensive than control during combustion.6 Thus, substantial research has been conducted to reduce these acid rain pre~ursors.~Control of NO, during combustion processes has been achieved through such methods as air and fuel staging, low-NO, burners, or product gas recirculation, all of which aim to reduce NO, formation through control of local fuel/air ratio. Reductions of up to 5-fold have been demonstrated with these technologies6 which are sometimes used in combination. Reduction of SO, is not amenable to control through combustion, since essentially all of the sulfur forms SO, species upon complete combustion. There is no alternative nonpollutant species for S as there is for N, namely Nz. Thus, SO, control during combustion has been achieved through use of naturally occurring sorbents, such as limestone, dolomite, or hydrated lime which contain CaC03 or Ca(0H)Z. After COz or HzO is driven off the solid sorbents by the hot gases, the remaining CaO reacts with sulfur species (e.g., SOz, HzS), retaining the sulfur in solid form7 for subsequent low-cost removal. Commercial practice in pulverized coal furnaces has shown 50-90 76 sulfur removal by sorbents;a sulfur removal efficiency by sorbents in fluidized beds can be higherS3 In this work, a generalized model for acid-rain precursors (MARP) has been developed, evaluated and integrated into a comprehensive combustion model. The model, (5) Smoot, L. D., Ed. Fundamentals of Coal Combustion: For Clean and Efficient Use; Elsevier: Amsterdam, 1993. (6)Lisauskas, R. A.; Snodgrass, R. J.; Johnson, S.A.; Eskinaze, D. Joint Symposium on Stationary Combustion NO. Control, Boston, MA,

1985. (7) Silcox, G. D. Ph.D. Dissertation, University of Utah, Salt Lake City, UT, 1985. (8)Angleys, M.; Lucat, P. Proc. Second Znt. Conf. Coal Combust.

Beijing, China 1991, 584.

0 1993 American Chemical Society

Energy & Fuels, Vol. 7, No. 6,1993 787

Acid Rain Precursor Model

which considers SO, and NO, formation, reduction and control processes, is presented and applied to practical systems.

Model Theory and Framework Assumptions. The development of an effective mathematical model to predict NO, and SO, formation and capture of SO, by injected sorbents during pulverized coal combustion requires an adequate description of (1) nitrogen and sulfur release from reacting coal and char particles, homogeneous and heterogeneous reactions of nitrogen and sulfur-containingspecies, (3)sorbent particle dispersion, and (4) SO, capture by the sorbents. Several key assumptions have been made to simplify the descrip tion of each of these steps while accounting for rate-limiting formation of NO,, chemistry-turbulence interactions for homogeneous reactions, and diffusive and kinetic resistances exhibited during sorbent capture of SO,. The model assumptions are generally based on experimental observations from a range of test conditions (i.e., low- and highrank coals, fuel-lean and fuel-rich stoichiometry, swirling and nonswirling flows, and near- and post-flame regions), and have been discussed previously.@-13 Both the NO, submodel and SO,/sorbent reactions submodel are decoupled from the generalized combustion model and are executed after the flame structure has been predicted. The basis for this assumption is that the formation of trace pollutant species does not affect the flame structure which is governed by fast fuel-oxidizer reactions. This requires that the concentrations of some radical species, such as OH and 0 which are involved in the formation of NO,, be estimated. Decoupling of the submodelfrom the main combustion code also limits SO,/ sorbent reaction predictions to low mass injection rates of sorbent particles. The advantage of this approach is computational efficiency. Further, to solve the pollutant model equation jointly with the combustion model equations is far more complex and raises some fundamental questions. The time required to solve the system of equations for the combustingfuel can require many hours of computer time while the pollutant submodels typically converge in a fraction (-10%) of the time required to converge the combustion case. Thus, submodel parameters and pollutant formation mechanisms (e.g., the BET surface area of sorbents and sorbent injection location) can more easily be investigated by solving submodel using restart files for a precalculated flame structure. Conservation Equations. A Favre-averaged species continuity equation is used to calculate the time-mean mass fractions of the NO, and SO, species (NO, HCN, NH3, SOZ, and HzS are considered) throughout the turbulent flow field:13J4

a Yi

wi = &-)

a Yi

+ &-)

-

(;)(ig) -

For the NO, species (NO, HCN, and NH3), Wi is the overall mean chemical reaction source or sink term calculated by summing all individual (global and elemen(9)Boardman, R. D. Ph.D. Dissertation, Department of Chemical Engineering,Brigham Young University, Provo, UT, 1990. (IO) Boardman,R. D.; Eatough, C. N.; Germane, G. J.; Smoot, L. D. Combust. Sei. Technol. 1993,20,1.

1. N z + O W N O + N 2. Oz+NNO+O 3. N + O F N O + H

7

"C :1 NO

Coal Nitrogen and C h 0, a r r NH3

Surface

Figure 1. Global kinetic mechanism incorporated into a generalized nitric oxide predictive model for joint or separate prediction of fuel NO and thermal NO.

tary-step) mean reaction rates, Gj. These rates involve formation or destruction of each species in each computational cell in the reactor, plus the contribution (or source) of the species as it is released from the coal and/or char particles. It is assumed that the individual gas-phase reaction rates in the NO, mechanism can be time-averaged by convoluting the instantaneous rates over the random fluctuations of the two mixture fraction progress variables f (primary gas mixture fraction) and 7 (coal offgas mixture fraction). 3,= PSS,(wi(ftrl,h)/~(f,rl,h)) h9

h)drl df

(2)

This approach is discussed in detail by Smith et al.13 The main assumptions for this equation are (1) that the instantaneous reaction rates are uniquely correlated to the mixture fractions f and 7 and (2) that fluctuations of f and rl are not strongly cross-correlated and may be statistically separated. The validity of the first assumption has recently been addressed by Boardman and Smoot.ls In the SO, submodel, the reaction term wi for SO2 or HzS includes the evolution of sulfur from the reacting coal and char particles plus the depletion of HzS or SO2 through capture by calcined sorbents. The SO, submodel can currently treat capture of either SO2 or HzS,but not both simultaneously.

NO, Submodel Mechanism Reactions and Species. Fuel NO formationis assumed to proceed through HCN and NH3 which are oxidized to NO while being competitively reduced to Nz as illustrated in Figure 1. Global reaction rates based on those reported by de Sobtele are employed for the fuel NO mechanism shown in Figure 1while the kinetic parameters reported by Bose et al." are used for an alternative global fuel-NO mechanism which has also been testedeg Fuel nitrogen is assumed to evolve from coal as HCN or NH3. Experimental evidence indicates that HCN is normally the first measurable volatile nitrogen species.18J9 The appearance of fuel nitrogen as NH3 in the gas phase may be important (11)Smith, P. J.; Smoot, L. D. AIChE J. 1986,32,1970. (12)Hill, S. C. Ph.D. Diesertation, Department of Chemical Engineering, Brigham Young University, Provo, UT,1983. (13)Smith, P. J.; Hill, S. C.; Smoot, L. D. Nineteenth Sympoeium (Znternutionul) on Combustion; The Combustion Institute: Pittaburgh, PA, 1982;p 1263. (14)Boardman, R. D.;Brewstar, B. S.; Smoot, L. D.; Silcox, G. D. Trans. ASME, Znt. Joint Power Conf., Atlanta, GA 1992. (15)Boardman, R.D.; Smoot, L. D. Combust. Sci. Technol., in press. (16)de Soiite,G.G. Fifteenth Symposium (International) on Combustion; The Combustion Institute: Pittsburgh, PA, 1975;p 1093. (17)Bose, A. C.;Dannecker,K. M.;Wendt, J. 0. L. EnergyFuels 1988, 2, 301. (18)Ghani, M. U.; Wendt, J. 0. L. Twenty-third Symposium (Znternutionul) on Combustion;The Combustion Institute: Pittaburgh,PA, 1990; p 1281. (19)Freihaut, 3. D.;Proscia, W.M.Presented at the First International Conference on Combustion Technologies for a Clean Environment, Vilamoura (Algawe), Portugal, 1991.

788 Energy & Fuels, Vol. 7, No. 6, 1993

during char oxidation20or during high-temperature gasification of low-rank coals as indicated by experimental data of Brown et a1.21 Prompt NO formation and NO recycle reactions have been neglected in this NO, model. This is due to the increased complexity of the nitrogen chemistry and also intimate coupling of these reactions with the fuel oxidation steps. Despite the increasing availability of highly efficient, memory-extensivecomputers, it remains impractical to model comprehensive gas reaction mechanisms during coal combustion while including the important effects of turbulence interactions. The current model cannot be used to predict the effects of fuel-staging (i.e., reburning) with natural gas or other hydrocarbon. Thermal NO formation is governed by the extended Zel’dovich mechanism which is shown by the elementarystep reactions above the fuel NO mechanisms in Figure 1. The overall rate for the three reversible reactions is22

A simplified expression is obtained by assuming initial concentrations of NO and OH are low so that only the forward rates of the Zel’dovichmechanism are significant.23

-d[Nol - 2kl[01[N21

(4) dt Radical Estimates. Rate equations (3) and (4) are coupled to the fuel oxidation process through competition for the oxygen atom, whose local concentrations must be estimated since a comprehensive kinetic scheme is not used to compute the fuel oxidation chemistry. In fuellean, secondary combustion zones, where CO is oxidized to C02, oxygen atoms are often assumed to be in equilibrium with 02.

Smoot et al. general estimation of the atomic oxygen concentration is far from resolved. The importance of accounting for chemistry-turbulence interactions when predicting NO, concentrations was illustrated by Smith and Smoot.” There is a significant difference between the NO profiles predicted from mean temperature and composition profiles and those obtained when the kinetic rates were integrated with respect to fluctuating properties. Evaluation. Evaluation of the NO, model has been conducted for a variety of experimental, two-dimensional, coal reactor configurations and conditions, including swirlingdiffusion flames, external air-staging, and oxygenblown gasification.11@~25The capability of predicting trends associated with variation in operating parameters such as equivalence ratio, swirl number, air-staging location, coal type, particle size, and reactor pressure was demonstrated for bituminous and subbituminous coals. The thermal NO mechanism has also been tested for predicting trends with variation in equivalence ratio and swirl number for a natural gas diffusion flame in a large, laboratory-scale combustor.1° Several model predictions have alsobeen compared to detailed local flame data. These include coal combustion and coal gasification cases as well as a natural gas flame. Table I summarizes the results of these various evaluations. The table illustrates the effects of flame conditions and combustor operating parameters such as inlet stream swirl number, stoichiometry, and fuel type. The sensitivity of an early version of the NO, submodel to kinetic parameters was evaluated by Hi11.I2 The sensitivity of the predictions to the form of the fuel NO mechanism was discussed by Boardmans but has not been fully determined. The simplified thermal NO rate expression (eq 4) yielded approximately the same results as the complete thermal NO rate expression (eq 3) for coal combustion and natural gas cases. However, eq 3 better predicts thermal NO concentrations for fuel-rich cases, such as gasification. Equation 3 provides a path for fuel NO to be destroyed through the reverse Zel’dovich reactions.

SOJSorbent Reactions and Submodel Mechanism In the flame region where hydrocarbons are consumed, oxygen atom concentrations often significantly exceed the level which is predicted by expression 5. A “partial equilibrium” expression has also been developed for fuelrich zones where 0 and OH radical concentrations may be approximated by fast-shuffle oxy-hydrogen reactions and where CO is further oxidized to C02.

Equations 5 and 6 yield equivalent 0 concentrations when CO, C02, and 02 are in complete chemical equilibrium. Equation 6 will yield different results when a rate expression is also used to calculate CO oxidation to C02. Use of eq 5 gave far better results for combustion of natural gas and air with equivalence ratios in the range from about 0.9 to 1.0 in a laboratory reactor.1° However, reliable (20) Peck, R.E.; Glarborg, P.; Johnsson, J. E. Combust. Sci. Technol. 1991, 76, 81. (21) Brown, B. W.; Smoot, L. D.; Hedman, P . 0. Fuel 1986,65, 673. (22) Westenberg, A. A. Combust. Sci. Technol. 1971,9,59. (23) Caretto, L. S. Fourteenth Symposium (International) on Combustion; The Combustion Institute: Pittsburgh, PA, 1977; p 803.

Reactions and Species. The SO,/sorbent reaction submodel requires simultaneous prediction of gas-phase sulfur concentrations and sulfur species capture by sulfation of entrained, calcium-basedsorbent particles. Thus, the trajectories and conversion of the sorbent particles must be tracked throughout the reactor or furnace. Sulfurcontaining pollutants formed by burning fossil fuels ~ normal include S02, SO3, H2S, COS,and C S Z . ~Under boiler operating conditions with excess air, virtually all of the sulfur is converted to SO2 with small quantities of so3. Fuel-rich species (primarily H2S) exist only in regions of the reactor where the coal particles are rapidly devolatilizing and oxygen is depleted. Normally, the sorbent particles are injected downstream of the burner to avoid sintering of the sorbents in the high-temperature region near the burner. At this point in the reactor, the gaseous sulfur will be at or near equilibrium concentrations and (24) Hill, S. C.; Smoot, L. D.; Smith, P. J. Twentieth Symposium (International)on Combustion; The Combustion Institute: Pittsburgh, PA, 1984; p 1391. (25) Boardman, R. D.; Smoot, L. D. AZChE J. 1988,34, 1573. (26) Kramlich, J. C.; Malte,P. C.; Grosshandler, W. L. Eighteenth Symposium (International)on Combustion;The Combustion Institute: Pittsburgh, PA, 1981; p 151.

Acid Rain Precursor Model

Energy & Fuels, Vol. 7, No. 6,1993 789 Table I. Summary of NO, Submodel Validation Studies

variable coal moisture*#

flame conditionso wet and dry Wyoming subbit

nparticle sizebqc

165 and 35 pm Utah bituminous

swirl number(' (coal flames) swirl numbepf (natural gas flame)

0 Iswirl no 5 10 subbit and bitum

COn diluente

0 I COZ/Oz I 1.5 Colorado subbit

stoichiometry a. coal-air diffusion'

0.5 Ist ratio I1.2

b. coal-air premixed8 c. natural gas diffu6.e

~

-

0 5 swirl no. I5 different-primary tube diameters

-

0.8 I st ratio S 1.2 0.8 S equiv ratio I 1.1

air stage locations

primary-zone st. ratio and residence time Kentucky bitum

gasification pressurefa

1 and 5 bar Utah bituminous

char oxidatiod

fuel-rich, Utah coal char

0

remarks predictions match observed increase of NO with increasing moisture; 12% difference between exit values predicted observed trend of decreasing NO with increasing particle size; -16 % difference between exit values observed local minimum in NO concentrations predicted and explained; -8% difference between exit values observed trends predicted with only partial success by thermal NO mechanism alone; -differences attributed to prompt NO and possibly inaccurate prediction of mixing temperature insensitivity of fuel NO predicted; -15% difference between exit values trend for swirling and nonswirling conditions matched; 19% difference for exit values observed trend matched; -12% difference for exit values observed trend matched with thermal NO mechanism using partial equilibrium expression to estimate 0 concentrations observed trend of reduced NO concentrations correctly predicted for (1)different stoichiometric ratios in the primary zone and (2) increased residence time in the primary zone observed trend of decreasing NO concentrations with increasing pressure correctly predicted; -30% difference between exit values (2ppm for 2 bar case) peak NO concentrations correctly predicted but exit values differed significantly (>400%)

*

st. ratio = stoichiometric ratio. See ref 12. See ref 24. See ref 11. e See ref 10.f See ref 9. s See ref 25.

will exist mostly as SO2 or H2S. Gas-phase reactions are assumed to be fast, giving rise to local concentrations in equilibrium. The approach of this "first-generation" SOJsorbent reaction submodel has been to consider cases of sorbent reactions by only SO2 or H2S. Sulfur is assumed to be released from the coal at a rate proportional to the total coal mass loss and is partitioned between SO2 and H2S, according to chemical equilibrium in the effluent gas. The model does not currently allow for the conversion of SO2 to H2S or vice versa once it is formed. This limitation of the model can be alleviated by including the sorbent injection in the overall convergence scheme of the coal combustion model or by using a devolatilization submodel which predicts the evolution of individual gas species, including sulfur, and incorporating reaction kinetics for the sulfur species in the manner which NO, reactions are modeled. Alternatively, the need to partition the sulfur evolved from the coal or char between SO2 and H2S according to the overall flame stoichiometry might be eliminated by using eq 1to calculate a correction to the mass fraction due to capture, where the source term for sulfur evolving from the coal and char is neglected and only the capture term is included. This correction could then be added to the mass fraction without sulfur capture to get the mass fraction after sulfur capture. The shrinking-coregrain model developed and evaluated by Silcox et al.7027is used to predict sorbent capture. The overall reaction process involves mass transfer of gaseous sulfur species, and/or oxygen to the calcined sorbent surface, diffusion through the intergranular pores, solidstate diffusion through the Cas04 product layer, and reaction with solid CaO, according to the following overall reactions.

-- +

CaO(s) + '/202(g) + S02(g) CaO(s) + H2S(g)

CaS(s)

CaS04(s)

(7)

H20(g)

(8)

CaO(s)+ SO,(g) CaS04(s) (9) Because the molar volume of Cas04 is greater than the

molar volume of CaO, the formation of Cas04 product decreases the pore radius and eventually closes off the pores. Thus, there is a practical limit to conversion which favors smaller particles. The effects of the sorbent particles on the gas velocity and radiation fluxes are neglected. The particles are small (usually