a model for dry sodium bicarbonate duct injection flue gas desulfurization

7/16/2019 A Model for Dry Sodium Bicarbonate Duct Injection Flue Gas Desulfurization

http://slidepdf.com/reader/full/a-model-for-dry-sodium-bicarbonate-duct-injection-flue-gas-desulfurization 1/12

Advances in Environmental Research 8 (2004) 655–666

1093-0191/04/$ - see front matter ᮊ 2003 Elsevier Science Ltd. All rights reserved.doi:10.1016/S1093-0191(03)00038-8

A model for dry sodium bicarbonate duct injection flue gasdesulfurization

Changfa Wu , Soon-Jai Khang *, Tim C. Keener , Sang-Kwun Leea a , b c

Department of Chemical Engineering, Cincinnati, OH, USAa

Department of Environmental Engineering, University of Cincinnati, Cincinnati, OH, 45221, USAb

Department of Environmental Science and Engineering, Hankuk University of Foreign Studies, Yoingin, Kyonggido 449-791,c

South Korea

Accepted 4 April 2003

Abstract

A mathematical model is developed for simulation of dry sodium bicarbonate (NaHCO ) duct injection for the3

removal of sulfur dioxide (SO ) in flue gases across a fabric filter (baghouse). The model employs parallel reaction2

kinetics and assumes that the sodium bicarbonate injection process can be separated into two stages. The first stageis a transport duct section where NaHCO particles are injected into the sulfur dioxide laden gas stream. The second3

stage is the fabric filter section where sodium sorbents are collected and behave as a variable depth fixed bed reactor.The process simulation for the efficiency of desulfurization in flue gas is performed and evaluated for a variety of operating conditions such as system temperature, particle size, residence time, normalized stoichiometric ratio,

concentration of sulfur dioxide and decomposition time. It is found that the removal of SO within the duct section2

is small and negligible for most practical conditions, with a contribution normally less than 5% of total SO removal.2

The major removal of SO occurs across the filter cake, which accumulates the sorbent particles on the fabric filter.2

These particles are periodically disposed as the filter is cleaned. The major factors for the process are temperature,particle size and SO gas concentration for all operating conditions. At low temperatures, the removal of SO2 2

increases as temperature increases, but the removal decreases at higher temperatures due to the impact of the thermaldecomposition reaction of NaHCO on SO removal. It was found that the temperature for the highest removal of 3 2

SO is within the range of 127–150 8C and the removal efficiency also depends on particle size.2

ᮊ 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Sodium bicarbonate; Sulfur dioxide; Desulfurization

1. Introduction

Sulfur dioxide (SO ) emissions related to industrial2

operations primarily occur from combustion sources andthermal processes, such as power plants (coal or oilfired), incinerators, steam generation equipment, processheaters, chemical reactors and other similar equipmentand processes. All of these emissions must follow EPAregulations set by the 1990 Clean Air Act Amendment

*Corresponding author. Tel.: q1-513-556-2789; fax: q1-513-556-2789.

E-mail address: [email protected] (S.-J. Khang).

(CAAA). Recently, as the construction of new powergeneration facilities is emphasized and most of thefacilities have plans to use coal, a renewed and moreinterest in economical methods of controlling SO emis-2

sions will be needed. It is reported that more than 250techniques for flue gas desulfurization (FGD) have beenproposed or developed on a worldwide basis (Oxley,1991). However, relatively few of those processes arecurrently in use (Makansi, 1993).

The major FGD systems can be broken down intosemi-dry processes, such as lime or limestone spraydryers, wet limestone or lime processes, ammonia injec-tion processes, caustic scrubbing processes and a dry



656 C. Wu et al. / Advances in Environmental Research 8 (2004) 655–666

process using sodium bicarbonate andyor trona, a natu-rally occurring mixture of sodium bicarbonate andsodium carbonates.

Full scale demonstrations of using sodium bicarbonatesorbent as a process for flue gas desulfurization have

shown that this process can achieve 90% percent remov-al of SO and some reduction of NO with or without2 x

injection of ammonia or urea (Fuchs et al., 1990; Bland,1990; Darmastaedter, 1990; Hooper, 1988). The resultshave shown a great potential for large-scale use of thistechnology with low investment to achieve the goal of compliance with regulations on SO removal and NO2 x

reduction. In dry sodium sorbent injection process,sodium bicarbonate (NaHCO ), or trona (sodium ses-3

quicarbonate; Na CO ØNaHCO 2H O) is injected after2 3 3 2

the pre-heater. This process is attractive for small plantsand other utilities in western states, because of itssimplicity and ease of retrofit with those power plantsequipped with bag house filters. Although sodium bicar-bonate is more expensive than lime or limestone, it canachieve very high sodium utilization and potentiallyproduce high value by-products such as Na SO which2 4,

can be sold to photographic, detergent, chemical, glassand paper industries (Darmastaedter, 1990). Some stud-ies have been conducted on the reaction mechanismsresponsible for the sodium bicarbonate or carbonatereaction with SO to provide insight into this desulfur-2

ization process. However, it seems that there are noavailable models or methods to predict the removalextent of SO across a fabric filter with which to provide2

a guide for practical process operation and design forsuch operational parameters as flue gas temperature,proper sorbent particle size, or overall sorbent residencetime for large scale operation. In lieu of this, a mathe-matical model for sodium bicarbonate injection has beendeveloped to simulate the process in order to evaluatethe performance of the desulfurization efficiency forvarious operational conditions. The model takes intoaccount the effects of operating parameters such astemperature, flue gas SO concentration, particle size of 2

sodium sorbent, stoichiometric ratio and sorbent resi-dence time on the removal of SO .2

2. Chemistry of sodium bicarbonate reaction with

sulfur dioxide

In the dry sodium FGD process, the removal of SO 2

is achieved by adsorption, andyor reaction with drysorbents. The reactions include sodium bicarbonate reac-tion with SO andyor thermal decomposition, with the2

product of decomposition, sodium carbonate, directlyreacting with SO to produce NaSO .2 3

Keener and Davis (1984) studied the reaction of sulfur dioxide with sodium carbonate and sodium bicar-

bonate over various temperatures and particle sizes. The

results for both sodium carbonate and freshly formedsodium carbonate obtained from the decomposition of sodium bicarbonate are quite different. The reason forthis is ascribed to the high level of reactivity of theimmediately formed surface of sodium carbonate fol-

lowing the thermal decomposition of sodium bicarbon-ate. It has been shown (Keener and Khang, 1993) thatsodium bicarbonate can react with SO directly at2

temperatures lower than 340 K before a large amountof decomposition has occurred. Kimura and Smith(1987) studied the kinetics of sulfur dioxide withsodium carbonate over different temperature ranges.They proposed that the Na SO is formed by two paths2 3

according to the temperature range. The direct reactionwith SO occurs at temperatures above 413 K and2

irreversible adsorption of SO occurs at temperatures2

from 353 to 413 K by the formation of an intermediate,followed by desorption of CO . It has been shown that2

there is a large increase in pore volume after thermaldecomposition and a first order reaction with SO was2

observed for this process up to high levels of conversion(Hu et al., 1986). However, at high temperatures ()

316 8C), sintering was reported to occur. Finally, sodiumsulfite reacts with O (from the excess air used in2

combustion), resulting in the formation of sodium sul-fate (Helfrich and Bortz, 1992).

Sodium bicarbonate decomposes at temperaturesgreater than 60 8C (although it decomposes slowly atthese low temperatures), resulting in the fresh formationof sodium carbonate with high surface area. The freshly

formed material is very active with SO . Keener and2

Khang (1993) modeled the kinetics of the NaHCO –3

Na CO –SO reaction by using a parallel reaction path.2 3 2

They proposed the following overall reaction paths forthe sodium bicarbonate desulfurization system.

k1

path 1(a): 2NaHCO qSO™Na SO q2CO qH O

(1a)3 2 2 3 2 2

kd

path 1(b): 2NaHCO™Na CO qCO qH O (1b)3 2 3 2 2

k2

path 1(c): Na CO qSO ™Na SO qCO (1c)2 3 2 2 3 2

where, k , k and k in Eq. (1a)–Eq. (1c) are reaction1 d 2

rate constants. The gas–solid reactions in Eqs. (1a) and(1c) occur on a non-porous surface, forming behindporous micro-grains due to the evolution of gaseousCO . The thermal decomposition reaction in Eq. (1b)2

is taking place in a micro-grained pellet.Further, sodium sulfite is oxidized into sodium sulfate

in excess oxygen conditions as shown in Eq. (2).

Na SO q1y2 O ™Na SO (2)2 3 2 2 4



657C. Wu et al. / Advances in Environmental Research 8 (2004) 655–666

As given in Eq. (1a), NaHCO reacts directly with3

SO to form Na SO , but may also thermally decompose2 2 3

resulting in the formation of Na CO with a larger and2 3

newly formed surface area. The subsequent reactionwith SO is shown in Eq. (1c). A kinetic expression2

has been developed (Keener and Khang, 1993) todescribe reactions Eq. (1a)–Eq. (1c). The equationswere based on the assumptions of first order reactionsfor all gas–solid reactions, in which an unreacted coremodel can be applied (since the original NaHCO3

particle is almost non-porous). A zero order reactionwas also assumed to describe the thermal decompositionreaction given in Eq. (1b). In addition, a pore pluggingexpression was used to account for the pore closure andsubsequent loss of sorbent reactivity with increasedaccumulation of product.

When the unreacted core model is applied withassumptions of first order reaction for path Eq. (1a)

and zero order reaction for path Eq. (1b), the overallconversion of NaHCO to Na SO are dependent upon3 2 3

the decomposition of NaHCO , and are divided into3

two time regions. Before and after the time (t ) forb

complete thermal decomposition of NaHCO , the extent3

of conversion ( X ) of NaHCO to Na SO are written R 3 2 3

as follows:For reaction time, t Ft :b

w z6k C b B C2 SO2 2 3y tytŽ . px |X s A tyt qt e y t q tŽ .R p pr t 2 3y ~b b

3w zB EtC Fx |q 1y 1y (1yb) (3)D Gty ~b

For reaction time, t Gt :b

w z6k C b B C2 SO2 2 3y t ytŽ .b px |X s A t yt qt e y t q tŽ .R b p p b br t 2 3y ~b b

2y t yt y tytŽ . Ž .b p pUq(1yb)q D e ye (4)Ž .

rb

B E B E2 22 3C F C F Ast q t q t

p p p2D G D Gt tb b

22t 2t p p Bs q

t tb b

t pCs

2tb

3 2 3w zB E3k C t b 2t 2t 2t2 SO p p p p2 t ytb p C Fx | Ds e y t q q p2 2D Gt t t ty ~b b b b

where, X is the extent of conversion of NaHCO to R 3

Na SO and t is the time for complete thermal decom-2 3 b

position of NaHCO in Eq. (1b). is the concentra-C3 SO2

tion of SO , r , molar density of NaHCO , t , a pore2 b 3 p

plugging time constant and b, the fraction of thermaldecomposition. These expressions are used for thescrubbing model described in the following section.

3. Model development

In the sodium bicarbonate FGD process, the sorbentsare injected into the duct section either upstream or atthe entrance of the fabric filter. The sorbent particlesare transported with the flue gas to the fabric filterwhere they are collected along with the entrained flyash. The particles react with SO while they are in the2

entrained flow mode during their transport phase aswell as on the filter surface, where the major removalof SO takes place. When the pressure drop of fabric2

filter reaches a preset value, the particles on the surfaceare removed by using one of a variety of fabric cleaningtechniques. The total residence time of the sorbentparticles ranges from 30 to 60 min on average. The drysodium injection process is described schematically inFig. 1.

The variables affecting on the removal efficiency of SO include the intrinsic reaction kinetics, sorbent sur-2

face area, particle size, mass ratio of sorbent to SO ,2

system temperature, solid particle residence time andbehavior in the duct and fabric filter. Keener and Biswas(1989) studied the sodium carbonate dry injection pro-cess using a semi empirical kinetic model. Garding andSvedberg (1988) studied the overall removal of SO for2

sodium bicarbonate injection with a lumped reaction

model with Na CO , SO , O to form Na SO , H O2 3 3 2 2 4 2

and CO by employing an unreacted-core model and2

ignoring the thermal decomposition reaction. With thismodel, they achieved some agreement with experimentalresults over a limited range of Damkohler number. Butthe model did not follow the particle flow pattern anddid not consider the reaction within the filter cake,where the major removal takes place. In this study, atwo-stage model is developed by assuming that thereaction occurs in two different sections: A duct sectionwhere the sorbent particles are in an entrained flowmode, and a filter section where the particles are in a

packed bed mode. The model is described in brief asbelow. The reader is referred elsewhere (Wu and Chang-fa, 2000) for a more detailed explanation.

3.1. Duct section

In the duct section, the following assumptions areused to describe the process.

1. Solid sorbent particles are monodisperse, and uni-formly distributed across the duct section and remainso within the whole section.

2. Neither particle aggregation nor breakageyattrition

occur in this stage.




Fig. 1. Schematic of dry sodium bicarbonate duct injection for FGD.

3. Axial diffusion is negligible compared to convectiveflow for the high velocity of the flue gas and solidparticles, which implies that a plug flow can be usedfor both gas and solid.

With those assumptions, a mass balance for the SO2

in a control volume of the duct section is described bythe following equation:

d(C U)SO2y sR (t)N (5) p

d x

where, U is superficial velocity of flue gas in duct (cmys), R (t ) overall reaction rate per particle (molysy p ,

particle), N , particles number per unit volume of fluegas in duct section (1ycm ), and x , position along duct3

section (cm). In order to obtain the relationship betweenthe removal rate of sulfur dioxide and the sorbentconversion, the mass balance over a particle is taken,producing the following differential equation (Lee andGeorgakis, 1981),

dX R (t)R ps (6)

dt N p

where N is the initial number of moles of NaHCO per p 3

particle. When the previous kinetic Eq. (3) and Eq. (4)

are combined into Eq. (6), we can obtain the conversionrate of NaHCO in the duct section as follows.3

For reaction time, t Ft (before complete decompo-b

sition of NaHCO )3

wdX 6k C bR 2 SO2 2y tytŽ . pxs A 1ye y BtqCtŽ .dt r t yb b

2zB E B E3 tC F C F |q (1yb) 1y (7)D G D Gt t ~

b b

For reaction time, t Gt ; (after complete decomposi-b

tion of NaHCO )3

dX 2 DR ytyt( ) ps e (8)dt r tb b

In order to convert the above equations to dimension-less form, we need to define the dimensionless quantitiesof concentration, distance and particle residence time as

follows:

C x LSO2* *C s ; X s ; T s (9)R0C L USO2

where is inlet concentration of SO (molycm ), L,0 3CSO 22

duct length (cm), and T , particle resident time in duct R

section (s). Since t s x yU s( X )(T ) and ,* *t sP yC R p SO2

where P is pore plugging constant (molØsycm ), t yt * 3 p

can also be written by the dimensionless number groupas below.

0t LCSO2* *sX C (10)*t UP p

Assuming plug flow for gas and particle within theduct, the reaction time of a sorbent particle is equal toits residence time. The relationship between b and t ,b

can be obtained as follows:

b k (1yb) 2k Cd 1 SO2s , s (11)t r R t r Rb b o b b o

By combining the above Eqs. (9)–(11) into Eqs. (7)




Fig. 2. Solid particle accumulation and reaction on fabric filter surface.

and (8), we can derive the following dimensionlessequation for the conversion rate of NaHCO and con-3

centration of SO in the duct section.2

For reaction time, t Ft :b

0 *dX 6k k T C C0 *R 2 d R SO2 yX*C* LC yUPŽ .SO2ws A 1yeŽ .* 2dX r Rb o

0 *6k T C C1 R SO2* 2 *2xy BT X qCT X qR Rr Rb o

2B ETR *

C F= 1y X (12)D Gtb

or

* *SdC 6k k CT0 *2 d

yX*C* LC yUPŽ .SOU 2wsNN T A 1yeŽ .T p R* 2VdX r Rb o

*6k C1* 2 *2xy BT X qCT X qR Rr Rb o

2B ETR *C F= 1y X (13)∂D Gtb

For reaction time, t Gt ;b

dX 2T D* * *R R

yX C L C yUP0Ž .SO2s e (14)*dX r tb b

or

*dC 2T D* * *R

yX C L C yUP0Ž .SO2sN N e (15) p* *dX r Pb

where the boundary conditions for the above equations

are C (1 at X s0.* *

The above equations used for the duct section arehighly nonlinear. Therefore, a numerical solution is usedto integrate the above equation over the duct length. Anormalized stoichiometric ratio, NSR, which is themolar ratio of sodium bicarbonate to SO , is used to2

describe the amount of sorbents injected for desulfuri-zation. By using the NSR and by assuming the flue gasas an ideal gas, the particle density, N , can be writtenin terms of system pressure, P and temperature, T asfollowing.

(2NSRym )y(r n )(V C )(PyRT)Na P P G SO2 2NsVG

2NSR C (PyRT)SO2s (16)m (rP n )Na P2

3.2. Fabric filter section

A simplified process for particle capture and reactionon the fabric filter is shown in Fig. 2. Particle reactionand accumulation on the filter surface are dynamicprocesses. During the build-up of particles on the filtersurface, the concentration is changing across the cakewith time. In order to simplify the model development,a pseudo steady-state is assumed. Since the growth rateof the filter cake is relatively small during a short timeperiod, it is reasonable to assume the pseudo steady-state within a short time period (5–10 s). Under theassumption, we can estimate the overall SO uptake by2

considering the process as a fixed height of the particlelayer and by integrating over the layer’s thickness. Foreach time period, the height of particle layer is calcu-lated, and a fixed bed model is applied to perform thecomputation until the time set for the periodic cleaning.

It is also assumed that particles deposit immediately




Table 1Reaction constants and their activation energy

Reaction constant Constants Activationcoefficient (k )0 energy

(CarlymoleØK)

k (molycm s)2d 5.84=102 20 000

k (cmys) 2.62=106 13 512k (s )y1

2 5.87=104 7324

Data from Keener and Khang, 1993.

and uniformly on the surface of the fabric filter withoutany loss of solid particles (perfect filter efficiency), andthat the change of particle size is negligible as thereaction on the filter surface continues. Therefore, froma mass balance on the sorbent over the filter surface

area at any time t (the total reaction time through afilter plus the time spent in a duct section, t ), thed

thickness of the sorbent layer on the fabric surface maybe described by the following equation:

NU V f phs t (17)

(1y´ )

where h is the height of the particle layer on the filter,V s4y3p , the volume of a particle, ´ , the particle3R p 0

layer porosity, and U , superficial filtration velocity. f

The particle layer is assumed to be a fixed bed under

isothermal conditions with axial diffusion only. Bydefining , and Z s zyh(t ), and by con-* 0 *C sC yCSO SO2 2

sidering a mass balance for sulfur dioxide over a smallparticle layer, we obtain following dimensionlessequations;

For (t qt )-t ;d b

2 * * 2 *Sd C dC 6(1y´ )h (t)C k kT 2 dU wyP y A(1Te*2 *VdZ dZ D R re 0 b

tqt yt( )d pye )y B(tqt )qC(td

B Etqtd2 C Fxqt ) qk 1y s0 (18)d 1 ∂D Gtb

For (t qt )))t ,d b

2 * * 2d C dC (1y´ )h (t) 2 Dy tqt yt( )d pyP y e s0 (19)e*2 * *dZ dZ D Pe

where P is the Peclet number defined as U Z y D . Thee f e

boundary conditions for the above equations may be setat the fabric filter entrance and at the outlet of thesorbent cake:

1

CSO2* *C s at Z s0 (20)0CSO2

*dC*s0 at Z sh(t) (21)

*dZ

where is the outlet SO concentration from the duct1CSO 22

section.

4. Results and discussion

The FGD model for dry sodium injection is divided

into two stages and so the program for computation is

completed based on two stages. Since the differentialequations are highly nonlinear due to the complexity of reaction mechanisms, an analytical solution is not avail-able. For the duct section, the equations have beensolved numerically by simultaneously solving both con-version and concentration differential equations with a

fourth-order Runge–Kutta integration method. For thebag house filter section, the particle cake is divided intomany thin layers. Within a short time period such asapproximately 5 or 10 s, it is reasonable to make apsudo-steady state computation. This procedure is essen-tially a finite difference method with each layer as anindividual calculation unit. The concentration of SO is2

assumed to be uniform within this layer. Depending onthe flow rate of particles, the surface area of filter, theparticle size and the concentration of SO , the height of 2

a layer is approximated to be at the same order of magnitude of the particle diameter. Then, we calculatethe concentration of SO in each layer until it reaches2

the surface of filter.

4.1. Data used in the computation

Data required for this model are listed in Table 1.The kinetic constants for the model come from thekinetic study by Keener and Khang (1993). The cakeporosity, ´ , is taken as 0.5 from similar cases in theliterature (Mycock et al., 1995; Calvert and Englund,1984). The pore-plugging constant is estimated fromthe empirical equation developed by Keener and Biswas(1989). The initial particle molar density is taken as

0.0263 molycm (Keener and Khang, 1993).3

Other operational parameters are based on typicalpilot and full demonstartion test conditions (Bland,1990) are listed in Table 2. Most values of the opera-tional parameters used for the computation of a modelare presented in various graphs illustrating the effectsof parameters.

4.2. Model sensitivity

A sensitivity analysis provides a means for checkingthe behavior of the model under a variety of conditions.

The sensitivity of the model to each input parameter is




Table 2Operational parameters used for the computation of a model

Operational parameters Input values

Temperature (8C) 107–204 (380–477 K)

Particle size (mm) 5–90Particle density, b (molycm )3 0.0263Particle layer porosity, ´ 0.5Superficial velocity within 15

duct, U (mys)

Inlet SO concentration (ppm)2 500–2500Sodium to sulfur ratio 0.5– 2.5Volumetric flowrate (m ys)3 9.17–6.4=10y5

Air to cloth ratio (mys) 6.45=10 –2.03=10y3 y2

Table 3Standard input values for the model sensitivity analysis

Parameters Input values

Temperature (8C) 127 (400 K)

Particle size (mm) 20Inlet SO concentration (ppm)2 2000Sodium to sulfur ratio 1

necessary for identifying the degree of accuracy towhich parameters need to be estimated. The main runs

of sensitivity analysis consist of temperature, particlesize, inlet SO concentration and sodium to sulfur ratio.2

Each of these runs involves a discrete input parameter

from a standard condition. The standard conditions forthe input parameters are given in Table 3.

4.2.1. Temperature sensitivity

The model results are sensitive to temperature in

several important ways. The temperature affects on theextent of sorbents decomposition and also plays animportant role in the sulfation parameter computations.

The net effect of temperature on model results wasexamined by plotting a series of sensitivity curves. Asshown in Fig. 3a, the rate of increase in removal

efficiency is very quick and fairly constant after 1200 sat temperatures of 127 8C (400 K) and greater. However,maximum removal efficiencies occur under nearly 1278C (400 K) of temperature for all operating parameters.The lower temperature leads to smooth increase inremoval efficiency, with lower removal efficiencies at

1200 s.

4.2.2. Sensitivity of particle sizes

The sensitivity runs reveal the influence of particlesizes on decomposition and sulfation parameter com-putations. Both decomposition and sulfation at smallerparticle sizes lead to higher estimates of the reaction

parameter and, hence, higher removal estimates. Themodel sensitivity to particle size is illustrated for fivedifferent diameters in Fig. 3b. As the particle sizes

increase, the rate of increase in removal efficiency isdramatically fast, with higher removal efficiencies at alloperating conditions. The sensitivity of the model is

dependent of the particle size, but the maximum removalefficiency of smaller particle less than 20 mm is rela-

tively insensitive to the particle size.

4.2.3. Sensitivity of inlet SO concentrations2

The result obtained from sensitivity runs with differ-ent SO concentrations clearly shows higher removal2

efficiencies for higher inlet SO concentrations at the2

constant operating conditions maintained in the analysis.

As shown in Fig. 3c, model sensitivity to inlet SO2

concentration is more significant for lower concentra-tions. A two-fold increase in inlet SO concentrations2

less than 1000 ppm, results in about doubling increaseof predicted removal efficiency, with less increase ininlet SO concentrations greater than 1000 ppm.2

4.2.4. Sensitivity of sodium to sulfur ratio

In the model, sorbent injection rate was directlyproportional and most sensitive to sulfur dioxide remov-al efficiency. A series of sensitivity curves are plottedfor 0.5, 0.8, 1.0, 1.5, 2.0 and 2.5 of sodium to sulfur

ratio, and shown in Fig. 3d. A two-fold increase insodium to sulfur ratio results in a doubling increase of predicted removal efficiency for all conditions. In allcases, higher sodium to sulfur ratio yield higher effi-ciencies, but lower sorbent utilization.

4.3. Model validation

The data presented by Bland (1990), has been usedto validate the model performance. The sodium bicar-bonate injection rate may be compared to the NSR fora given SO concentration and the flue gas flow rate.2

Model results are shown in Fig. 4 and compared withtest data. The results show that as the NSR increases,the removal of SO increases, regardless of temperature,2

concentration or the particle size. At NSRs1, theremoval should be 100% if a complete reaction isachieved. However, the actual results are less becauseof the limitations of reaction kinetics and reaction timeas well as the pore plugging throughout the combinedprocess. The model results underestimate the test data,which may be due to the fact that the reported particlesize may be less than the actual particle size and thetemperature may have been varying throughout the

process. Smaller particles may be suspended in the fluegas for longer times than other larger sizes, and thushave different exposure time to SO .2

The relationship between the utilization of sodium




Fig. 3. Sensitivity analysis of model parameters; (a) Temperature, (b) Particle size, (c) Inlet SO concentration and (d) Sodium to2

sulfur ratio.

bicarbonate and the particle size is shown in Fig. 5. Anupper limit on utilization is predicted for particles lessthan 20 mm. As the particle diameter increases, theconversion of sorbent quickly reduces. These results aresimilar for other gas temperatures. Again, the differencesbetween the model predictions and test data might becaused by the use of an average particle size (wide sizedistribution) and the non-uniform particle velocity inthe duct.

The removal of SO occurs in the fabric filter section,2

where particles are accumulating on the surface of bagfilter. The concentration of SO and the conversion of 2

particles vary from time to time. However, as timeincreases, the outlet SO concentration levels off. The2

time needed to reach this level-off is reduced as thetemperature increases. This time span can be thought of

as the transient portion of the operating cycle. The time

to reach the maximum removal amount is comparedwith demonstration results of system recovery time aftercleaning (Bland, 1990) as shown in Fig. 6. The com-puted results compare favorably with the test data.

4.4. SO removal rate within the duct section2

After injection, it is assumed that SO removal starts2

immediately. The degree of SO removal was computed2

for 1 s of duct residence time(fixed) for different NSRsand temperatures, and for particle sizes of 5 and 10mm. The results are shown in Figs. 7 and 8. The gasvelocity used was 15.0 mys, which is a typical flue gasduct velocity.

As seen from Figs. 7 and 8, the concentration changeswithin the duct are quite small, and are less than 5%

for most temperatures. For higher temperature and very




Fig. 4. SO model predictions compared to test data.2

Fig. 6. Time to reach steady state vs. gas temperature at NSRs

1.0.

Fig. 5. Comparison of modeled sorbent conversions with test

data. Fig. 7. SO duct concentration profile vs. NSR at T s177 8C.2

fine sodium bicarbonate particles, the duct contributionto the total removal is approximately 10%. This smallremoval is due to the short contact time, which resultsin the particles not decomposing to a significant level.Smaller particle sizes result in higher removal efficien-cies within the duct, as shown in Fig. 8.

4.5. SO Removal in fabric filter section2

The primary removal of SO occurs in the fabric2

filter section where the particles are accumulating onthe surface of the fabric filter and continuing to react

with SO until they are removed from the system. The2

removal of SO is a dynamic process at different2

positions across the particle layer. However, the wholeprocess may reach a near steady state level after acertain period of operation. For the following discus-sions, the computations for all situations are based onthe operational period after the beginning of sorbentinjection. For most cases of interest and from actualdemonstration test results, the time to reach a steadyremoval is from a few minutes to 20 min or so. Thecomputations carried out here have been conducted tosimulate 40 min of operation, which is enough for mostcases of interest.

4.5.1. Effect of particle size on SO removal rate2

Fig. 9 shows the computed results of the effect of particle size for removal across the fabric filter as afunction of filtration time for a gas temperature of 127




Fig. 8. SO duct concentration profile vs. gas temperature at2

NSRs1.0 and Dps5 mm.

Fig. 9. SO removal vs. sorbent particle size at T s177 8C, NSRs1.5 and inlet SO concentrations2000 ppm.2 2

8C (400 K), inlet SO concentration of 2000 ppm, and2

NSRs1.5. It can be seen that the removal of SO2

increases with the decrease of particle size. This is dueto the impact of thermal decomposition of the smallerparticles on the production of available surface area.

The diffusion of SO into a particle is very important2

for the whole reaction, especially for small pores in a

complex structure. Here, the initial sodium bicarbonateis considered as non-porous solid particle and is subjectto thermal decomposition at the system temperature.The decomposition process not only provides a pathwayfor further reactions, but also results in fresh unreacted

surface area with high reactivity.It can be seen in Fig. 9 that the removal of SO2

reaches the maximum level for particles less than 50mm during the 40 min simulation period. For theselarger particles, the time to reach this stage will increase.The larger particles would also theoretically reach thissame maximum given enough time on the filter. Thismay be due to the complexity of the decompositionprocess, which results in larger particles requiring longertimes to thermally decompose. As the operation timegets longer, the thickness of filter cake also increasesand thus may affect on the gradual increase in removalefficiency.

4.5.2. Effect of temperature

The system temperature has a significant effect onthe desulfurization process. Fig. 10 shows these effectsas a function of particle size at a NSRs1.0 for an inletSO concentration of 2000 ppm. From these results, the2

removal of SO increases with increasing temperature2

for all particle sizes until a maximum is obtained, atwhich point removal levels decease with further increasein temperature. Lower temperatures promote a moregradual production of highly active Na CO , which is a2 3

function of particle size. The SO uptake rate is more2

closely matched by the Na CO production rate along2 3

with adequate production of pore volume due to the




Fig. 10. SO removal vs. filtration temperature at NSRs1.02

and inlet SO concentrations2000 ppm.2

Fig. 11. SO removal vs. inlet SO concentration at Dps202 2

mm and NSRs1.5.

thermal decomposition reaction. This can result in highsorbent utilization, depending on the particle size. High-er temperatures increase the thermal decompositionproduction rate, which can result in low sorbent con-versions depending on the degree of SO uptake. At2

higher temperatures, the reaction comes to the finalremoval rate in a very short time period due to the fast

thermal decomposition reaction, which results in productbuild-up and pore plugging and closure at the outsidesurface of the particle. Pore plugging occurs due to themolar volume difference between the initial Na CO2 3

(42.25 cm ymol) and product Na SO (53.16 cm y3 32 3

mol).Increasing the sorbent particle size increases the

temperature where maximum removal occurs. Particlesof all sizes have almost the same conversion and thesame removal of SO at temperatures above 160 8C2

(433 K). This effect is different than that for sodiumcarbonate, which always shows a higher conversionwith increasing temperature (up to the point of sintering)

(Keener and Biswas, 1989). Hu et al. (1986) havereported that the total pore volume generated by thermaldecomposition of sodium bicarbonate at 100 C(373 K)8

is higher than that generated at 200 8C (473 K), butthe pore diameter distribution is quite different. At thelower temperature, the pore size distribution is narrowand has a peak at approximately 0.5=10 m, while aty6

the higher temperature, the pore size is spread and doesnot show the peak. This pore size distribution mayaffect the progress of reaction and cause the differencein the utilization of sorbent and the overall removal of SO . From these results, it can be concluded that it is2

essential to consider the temperature and particle size

together to achieve the optimal results for flue gasdesulfurization.

4.5.3. Effect of SO concentration2

The SO inlet gas concentration effects important2

operational parameters such as the total decompositiontime, the pore plugging time constant and the reactionrate as indicated in the previous discussion on kinetics.However, SO inlet concentration has its greatest impact2

on the reaction rate. Therefore, SO concentration has2

an influence on the total removal rate. Fig. 11 showsthe effect of inlet SO concentration on overall SO2 2

removal for the condition of NSRs1.5, and for 20 mmdiameter particles. The effect of inlet SO gas concen-2

tration on the desulfurization process is the result of acombination of factors, but higher concentrations of SO result in higher levels of sorbent utilization and2

higher levels of SO removal. Bland and Martin (1990)2

found that the inlet SO concentration effects the total2

removal of SO with sodium bicarbonate in a manner2

similar to that shown in Fig. 11.

5. Conclusions

A two-stage model has been developed to simulatethe sodium bicarbonate duct injection FGD process. Anentrained flow reactor model is applied to the ductinjection section, while a fixed bed model is employedto describe the development of accumulation of particlesand removal of SO in the fabric filter stage. It was2

found that within the duct section, the removal of SO2

is small and negligible for 1 s particle residence timeand typical particle sizes and gas temperatures. MostSO removal occurs within the sorbent layer in the bag2

house filter. The effects of operating parameters suchas particle size, temperature, concentration and normalstoichiometric ratio on the process efficiency have been




discussed and results validated with published test data.The simulation results indicate that for a given particlesize, there is a temperature where SO removal is2

optimized. For typical particle sizes less than 30 mm,the optimum temperature is between 127 and 150 8C

(400–423 K). The inlet concentration of SO has also2

been found to be an important factor on the desulfuri-zation efficiency. Higher SO concentrations improve2

the percentage of removal by increasing the reactionrate, which becomes a complex variable in the overallgas–solid removal mechanism.

6. Notation

:0CSO2Inlet concentration of SO , molycm3

2

:CSO2Concentration of sulfur dioxide, molycm3

D :e SO diffusivity in flue gas, cm ys22

Dp: Particle diameter, microns H : Thickness of particle layer on the filter, cmK :1 Reaction constant for the reaction of

NaHCO with SO , cmys3 2

K :2 Reaction constant for the reaction of Na CO2 3

with SO , sy12

k :d Reaction constant for thermal decompositionof NaHCO to Na CO , molycm s2

3 2 3

L: Duct length, cm M :G Mole flow rate of SO , molyh2

m :Na Mole fraction of sodium in the sorbentinjected

N : Particles number per unit volume of flue gasin duct section, 1ycm3

N : B Moles of NaHCO , mol3

N : p Initial number of moles of NaHCO per a3

particle, molyparticleP :* Pore plugging constant, molycm3

R :0 Particle radius, cmr :c Thermal decomposition reaction front radial

position, cm R : p Overall reaction rate per particle, molys

particlet :b Time for complete decomposition of

NaHCO , s3

t :d Particle resident time in the duct section, st : p Pore plugging time constant, sT : R Particle resident time in duct section, sU : Superficial velocity of flue gas in duct, cmysU : f Superficial filtration velocity, mysV :G Volume flow rate of flue gas, m yh3

V : p Volume of a particle, cm3

X : Position along duct section, cm X : R Conversion of NaHCO to Na SO3 2 3

r :b Molar density of NaHCO , molycm33

b: Fractional amount of NaHCO decomposing3

to Na CO2 3

´ : Filter cake porosity

References

Bland, V.V. 1990. Evaluation of sodium compound utilizationin combustion gas SO and NO reduction, EPRI, GS-6850.2 x

Bland, V.V., Martin, C.E. 1990. Full-scale demonstration of additives for NO reduction with dry sodium DFG, EPRI,2

GS-6852.

Calvert, S., Englund, H.M., 1984. Handbook of Air PollutionTechnology. John Wiley and Sons, NY.

Darmastaedter, E., 1990. Sodium bicarbonate injection: a smallplant SO yNO Option, Power Engineering.2 x

Fuchs, M.R. et al., 1990. Full-scale demonstration of desulfur-ization by dry sodium injection, EPRI GS-6860.

Garding, M., Svedberg, G. 1988. Modeling of dry injectionflue Gas desulfurization, JAPCA, 38(10).

Helfrich, D.J., Bortz, S.J. 1992. Combined SO and NO2 x

removal by means of dry sorbent injection, EnvironmentalProgress, 11(1).

Hooper, R. 1988. Full scale dry sodium injection demonstrationmeets SO BACT, Power Engineering, June.2

Hu, W., Smith, J.M., Dogu, T., Dogu, G. 1986. Kinetics of sodium bicarbonate decomposition, AIChE J. 32(9).

Keener, T.C., Khang, S.J. 1993. Kinetic of the sodium bicar-bonate-sulfur dioxides reaction, Chem. Eng. Sci. 48(16).

Keener, T.C., Biswas, P., 1989. A dry scrubbing model forSO removal, Chem. Eng. Comm. 81.2

Keener, T.C., Davis, W.T. 1984. Study of the reaction of SO2

with NaHCO and Na CO , JAPCA, 34(6).3 2 3

Kimura, S., Smith, J.M. 1987. Kinetics of the sodium carbonatesulfur dioxide reaction, AIChE J. 33(9).

Lee, D.C., Georgakis, C., 1981. A single particle size modelfor sulfur reaction in fluidized-bed coal combustion. AIChEJ. 27, 3.

Makansi, J. 1993. Controlling SO Emission, Power, March.2

Mycock, J.C., McKenna, J.C., Theodore, L., 1995. Handbook of Air Pollution Control Engineering and Technology. CRCPress, Inc, Boca Raton.

Oxley, J.H., Rosenberg, H.S., Barrett, R.E., 1991. Energy Eng.88, 6.

Wu, Changfa, A Scrubbing Model for Dry Sodium BicarbonateDuct Injection Flue Gas Desulfurization, Masters Thesis,University of Cincinnati, 2000.

a model for dry sodium bicarbonate duct injection flue gas desulfurization

Documents

removal efficiency

major removal

removal of so2

the3removal of sulfur

fabric filter section

highest removal of3

sodium sorbents

particle size