Jar1 R. Ahlbeck and Stefan Ronnblad

Transcription

1 Removal of Sulfur Dioxide in a Fabric Filter Jar1 R. Ahlbeck and Stefan Ronnblad Abo Akademi University Abo, Finland In dry and spraydry flue gas desulfurization, a fabric filter significantly improves the removal of SO,. This report describes the reactivity of a deposited absorbent particle layer with SO2 by three reactivity parameters in a semiempirical model. These parameters for four different lime hydrates were obtained from fixedbed reactor experiments by regression analysis. The SO, removal in a fabric filter was then theoretically predicted. Results obtained with this method are in good agreement with pilot scale findings reported in the recent literature. The reactivity measurement followed by simulation procedure described here can be used to predict the influence of design and operating variables on the SO, removal in a fabric filter. The particulate absorbent at the fabric filter inlet in a dry or spraydry (SDA) flue gas desulfurization process (FGD) plant might consist of fly ash, calcium hydroxide, calcium sulphite hemihydrate, calcium sulphatedhydrate, calcium carbonate, and free liquid water. The amount of liquid water is dependent on the type of FGD process. In the case of an SDA process, the amount of liquid water is generally about 2.5 percent at an approachtosaturation at the reactor outlet of 15 C. If hygroscopic additives, or a fuel with high chloride content, are used, the liquid water content is even higher. If the approachtosaturation is increased, the liquid water content decreases. If a waterless Ca(OH), sample is placed into highly humid flue gas, the liquid water content increases continuously as a result of water absorption. The reactivity of the powder with SO, increases when the water content increases. An absorbent prepared by filtering and drying from a slurry has shown to be very reactive.' The previous history of the contact between the solid absorbent and the water thus seems to play a key role in the SO, removal. The significant influence of the liquid water is probably due to the fact that there is a predominant ionic reaction mechanism in the absorbent cake according to: This is, however, not the only possible reaction mechanism. The chemical system is extremely complex. Therefore the best way to obtain the efficiency of different types of absorbents is to measure the reactivity with SOz directly. Measurement of the hygroscopic properties of the absorbent might, however, provide some information about the reactivity of the absorbent., The use of hygroscopic additives such as CaCl, significantly increases the reactivity of the absorbent.3 The presence of additives might also increase the reactivity by improving the solubility of calcium hydroxide according to Equation C.4 SO,(g) + 2H,O(I) { HSO;+ H,Ot (4 (b) HSO, + Ca2+ + OH { CaS03(s) + H,O (1) Ca(OH),(s) { Ca H (4 H,O+ + OH { 2H20 (1) (4 Overall reaction: S02(g) + Ca(OH),(s) CaSO,(s) + H20(1) (e) Crystal water reaction: 2CaSO,(s) + H2C!(1) { 2CaS0,. 1/2H,O(s) (9 Copyright 1994 Air & Waste Management Association AIR & WASTE Vol. 44 April

2 The SO2 Removal The SO, removal in the filter, q, is defined according to: ms02"n msoyou1 r= (1) mso y In where mso2in in the mass flow of SO, into the filter and msoi out is the mass flow of SO, out from the filter. In the case where the water evaporation from the filter cake is small and the air in leakage is negligible, we can write: where C,, is the SO, concentration at the filter inlet and eo,, is the final concentration in the stack (moles of SO,/moles of flue gas). The removal might be dependent on the: Stoichiometric ratio (moles unreacted absorbent/moles SO,) at the filter inlet. Reactivity of the absorbent. Moisture content and the temperature of the flue gas. Weight concentration and properties of fly ash in and/or additives in the flue gas. Concentration of SO, in the flue gas at the filter inlet Design and operating parameters of the filter system. It is well known that the humidity of the flue gas is an important parameter.5 At temperatures close to the water dewpoint, SO, removal in a fabric filter of 100 percent might be possible.6 It is, however, not technically possible to operate at close proximity to the water dewpoint. At an approachtosaturation of 15"C, SO, removal in the filter of about 35 percent in an SDA process is reported by many authors. If the SDA reactor SO, removal then is 75 percent, the fabric filter will thus increase the total removal to 84 percent. The filter SO, removal is dependent on the reaction time which is equivalent to the gas residence time in the filter cake. If t* equals to buildup time of a cake, the gas residence time, tms, can be written)' according to: where E is the porosity of the cake, c is the mass concentration of powder in the flue gas (kg/m3), andp is the density of the powder particles (kg/m3). From this simplified equation we obtain that the reaction time and thus also the SO, removal is dependent upon the time since the last cleaning, the mass concentration of absorbent in the flue gas and the porosity of the cake and the particle density, but independent of the flue gas velocity and the surface area of the filter. A SemiEmpirical Model for an Absorbent Sample Measurements of the reactivity of different types of absorbents under conditions of highly humid gas were reported by Yoon, et a1,8 Ruiz Alsop,S Seeker, et al,3 and Peterson and Rochelle.' The reactivity of an absorbent with respect to SO, is usually tested in a fixed bed batch reactor where a sample is placed in contact with a gas containing SO,. The sample conversion, F, is defined according to: F= nr (4) nk where nk = initial moles of absorbent in the sample and nr = moles of reacted absorbent. Models for the sample conversion based on reaction kinetics have been presented by RuizAlsops and Westerlund, et al.9as the models contain some unknown constants and barely measurable (3) mass transfer surfaces which must be calculated from experiments anyway, it seems reasonable to construct a simplified model that fulfills the following experimentally and theoretically observed boundary conditions: I. F = 0 at the reaction t = 0 The sample conversion is zero at the beginning 11. &/at + o when t + co This condition is due to the fact that the reaction rate decreases rapidly towards the end of the measured period due to increased mass transfer resistance af/& * 0 if the SO, concentration, C = 0. When to sulfur dioxide is present, no reaction takes place. IV. df/& * owhen t c+ 00 This condition which has been observed by many authors, for example Kramlich et al.,lo is due to the fact that at low values of the SO, concentration (4000 ppm) one of the limiting mechanisms is the external SO, mass transfer. At high concentration, the limiting mechanism gradually changes to an internal mass transfer controlled absorption where the influence of the SO, concentration in the surrounding environment disappears. A semiempirical reactivity model can now be written according to: where the three absorbent reactivity parameters are: (1) F,, = asymptotical value of the sample conversion (2) t, (s) = time dependence parameter, high value means slow reaction (3) C, (moles SO,/moles flue gas) = concentration dependence parameter Now we can measure the reactivity parameters for different types of absorbents and at different values of the temperature and moisture content of the flue gas. The parameters can easily be calculated from fixed bed absorption experiments by nonlinear regression analysis. A Process Model for a Fabric Filter Unit The stoichiometric ration,n&, at the filter inlet can be expressed as a function of the overall stoichiometric ration, No,,, and the SO, removal, nr, is the previous part of the FGD process (for example the SDA reactor), according to: In the case of different absorbent recirculation strategies, corresponding equations can easily be derived from mass balances.11 The amount of unreacted absorbent, dn,, that enters the filter during the time interval, dt, can be written according to: dn, = N, Cin Odn, (7) 414 April 1994 Vol. 44 AIR & WASTE

3 where dn, equals the corresponding amount of flue gas (moles), and C,, is the SO, concentration (moles/moles) at the filter inlet. It might now be noticed that the following calculations can be performed with any type of mathematical model for the sample conversion. It is also possible to use the numerical values from a fixed bed experiment to predict the SO, removal in a fullscale fabric filter unit directly. In this report, the semiempirical model (5) is used. From Equations 4 and 5 we obtain, Now combining Equations 7 and 8 with the stoichiometric balance: we obtain from Equations 12 and 13, n(t*) = F,,,. (let'/te) (14) which means that the SO, removal for a stoichiometric ratio equal to one and internal mass transfer controlled absorption, is dependent on the buildup time exactly in the same way as the sample conversion is dependent on the reaction time in a corresponding fixed bed experiment. When much time has passed, the SO, removal approaches the value of the maximum possible sample conversion, Fmm. For a stoichiometric ratio greater than one, the SO, removal can reach a value that is significantly higher than Fma. The integrated mean value of the SO, removal is the filter unit, nf, is defined according to: P t, 1 dc dn, = dn, (9) we obtain a simple differential equation describing the profile of SO2 through the absorbent cake where the time, t, is proportional to the location in the cake so that t = 0 at the inlet contact surface between the cake and the flue gas, and increases in proportion to the downstream distance from this surface. We now want to calculate the outlet SO, concentration, from which we can calculate the time dependent SO, removal in the filter. Equation 10 has a solution according to: et/te. [CE. etne. ln(ec/ce 1) N,,. C,,. F,,,] = Q (11) where Q is the integration constant. Inserting the boundary condition C (t = 0) = C,, gives: C,. ln(ec1~ce 1) N,,. C,,. F,,, et/te. [CE. et/te. ln(ec/ce 1) N,,. C,,. F,,,] = 0 (12) Now if we know the values of the reactivity parameters, Fma, and C,, we can easily calculate any vale of C(t) by using a numerical rootsolver. From the value C(t = t*) where t* is the buildup time of the cake, and using definition (l), we can calculate the time dependent removal, n(t*), according to, (13) If we now compare this model with the simple expression for the gaseous SO, reaction time (2), we notice that we have a proportional relationship according to Equation 2 between the gaseous SO, reaction time and the buildup time. According to Equations 12 and 13, the SO, removal is also a function of the buildup time. We also have a proportional relationship according to Equation 2 between the mass concentration of absorbent in the flue gas and the reaction time. Equations 12 and 13 give a relationship between the SO, removal and the mass concentration defined as the inlet stoichiometric ratio. The porosity and density of the cake are now included in the reactivity parameters. There is also a relationship between the inlet SO, concentration and the removal. For the special case when, C, c c Ci, C, c c C(t*), and Ni, = 1 where tf is the time interval between two cleanings. Equations 2,12 and 13 claim that the SO, removal in the filter is independent of the surface area of the filter and the gas velocity through the filter, providing that the cleaning interval is constant. An experimentally observed correlation between the area of the filter and SO, removal observed by Stromberg and Karlsson (1988) was due to the fact that their experiments were performed with a constant pressure drop, kpa. When the filter area was doubled, the SO, removal rate increased four times because the cleaning interval could also be increased four times. A significant dependence of the filter removal on the gas velocity at constant value of the cleaning interval is due to the fact that the reactivity parameters are slightly dependent on the gas velocity. It is thusimportant that the fixed bed reactor experiments are performed with a gas velocity close to that in a fullscale filter. If calculations with this model are combined with a model for the pressure drop, economical optimization of the fabric filter can be performed. A Fixed Bed Reactor The objective of the experimental design was to provide an environment in which gas containing SO, could be passed through a fixed bed absorbent under carefully regulated conditions. The key regulated variables were bed thickness, approach velocity, temperature, gas composition and exposure time. Figures 1 and 2 illustrate the apparatus. The reacting gas, normally consisting of air, water vapor, and SO,, is contained in a 1 m3 batch which makes it possible to regulate the gas composition. The water vapor is introduced by a separate heater from which a small amount of water can be vaporized into the batch. The resulting partial pressure of vapor is measured by a hygrometer. The SO, is introduced through a rotameter and the resulting concentration is measured by an IRtype analyzer. The batch is heated by an electrical heater. The absorbent, about 2 g, is placed across a filter material on a bed. Initially SO,free gas from the batch is passed through the absorbent to bring the absorbent cake into equilibrium with the moisture in the reacting gas. The process signals are connected to a computer. The reacting gas flow through the bed is controlled by a small fan and a valve. The pressure drop is also measured. At the start of the test, SO, is introduced from the gas bottle until the desired starting concentration ( ppm) is achieved. From the decreasing SC, concentiatiofi, the ccmesponding sample conversion, F, as a function of the reaction time, is calculated by the computer from the stoichiometric balance. AIR & WASTE Vol. 44 April

4 The test is ended by stopping the air flow and removing the filter. The sample is removed and analyzed for sulfur (ASTM D2795) in order to check the mass balance. The mean value of the SO, concentration between the start and the finish of the experiment is calculated. From one run a curve of F versus t is obtained for this mean value of the SO, concentration. The first estimates of tt and F,, can then be calculated. Any computer program for nonlinear multiple regression analysis can be used. By repeating the experiment with the same amount of the same absorbent and with the same moisture content in the batch, but with new values of the initial SO, concentration, a set of curves is created from which C, can be obtained. Results Experimental Results from the Literature and Model Prediction A number of fixed bed experimental results have been reported in recent literature. There are, however, only a few authors that have tested the absorbent both as a sample in a fixed bed reactor and as a continuously injected absorbent in a fabric filter. Seeker et al.3 have reported measurements of the sample conversion, F, of a calcitic atmospheric hydrate called Longview lime in a bench scale reactor. They have also measured the baghouse SO, capture n(t*o, of the same hydrate in a pilot scale facility. From the fixed bed sample conversions measured by Seeker, we obtain a value of F,, 40 percent, and a value of the time constant, te 15 min. The approachtosaturation was 9 C. The value of CE cannot be determined from these experiments, but because the SO, concentration was as high as 1900 ppm, we can assume that the limiting mechanism is internal mass transfer ((CE << C(t*)). The authors have then injected the absorbent with a stoichiometric ration of 2.5 into a gas stream and obtained experimental values of n(t*) in a fabric filter as a function of the buildup time of the cake, t*. These measurements are shown in Figure 3 as dots. The approachtosaturation was the same as in the fixed bed experiment, 9 C. Now inserting the reactivity parameters from the fixed bed experiment in Equations 12 and 13 we obtain the theoretically predicted values of n(t*) illustrated in Figure 3. The theoretically predicted removal is slightly higher than th experimentally obtained removal at high values of the buildu time. This might be due to partial bypass around the filter nor uniformity of the filter cake. We have, however, reason to belie\ that the theory can be very useful when estimating the influenc of different design and operating parameters on the absorption c SO, in fabric filters. Reactivity Test of Calcium Hydrates In our experiments, four different calcium hydrates wer tested. All hydrates were manufactured and prepared in differei ways. When the experiments were performed at 58 C, and 79 percer R.H., we obtain the absorbent sample conversion as a function c time according to Figure 4. We notice that absorbent IV is le: reactive than the other absorbents. The parameter C, was obtaine changing the initial SO, concentration. A value of C, 190 ppn was obtained. From the curves in Figure 4 we obtain values shown in Tab1 I of the reactivity parameters F,,,, and te. Prediction of 30, Removal We can now use the reactivity parameters F,,, te, and C, i Table I and calculate the integrated mean filter SO, removal, n, according to Equation 15 as a function of the cleaning interval, ti of a filter unit. We can also calculate this removal as a function c the stoichiometric ratio at the filter inlet. F IS0 mm Figure 2. The fixed bed. A piece of fabric cloth is mounted between tw flanges. The absorbent sample (2 g) is placed on the surface of the cloth. 60 Humldlfler 1 m Fllbr 50 baghouse t t* min Figure 1. The experimental setup. Figure 3. Experimental results by Seekers (dots) compared with predictec results (curve) for a pilot scale fabric filter. N, = April 1994 Vol. 44 AIR 81 WASTE

5 Because the curves are calculated using reactivity parameters bbtained at 79 percent R.H. and 58 C, they are only valid for these :onditions. The whole procedure can, of course, be repeated for Ither experimental conditions and for other absorbents. From Figure 5 we assess that the SO, removal increases when he time interval, tf, increases. At the same time, the pressure drop ncreases. This can, however, be avoided by increasing the surface rea of the filter. The difference in reactivity between absorbent I and absorbent V which was measured in the fixed bed experiment causes a ignificant difference in filter SO, removal. At tf = 20 min we 60 I c f react ion time lgure 4. Measured sample conversion as a function of the time for four ifferent lime hydrates. z a.i i. E p: 5? 20 % I 80 I I I I I I 1 80 % k 3 60 z. L p N 20 tp = 10 min I IV I I I I I stoichiometric ratio I I t I I m stoichiometric ratio gure 5. Predicted SO2 removal versus stoichiometric ratio for two lime drates (I and IV) and two values of the cleaning interval. Table I. Reactivity parameters of four lime hydrates at 750 ppm SO*, 79 percent R.H. and 58 C. The parameter C, 190 ppm for all samples. I II 111 1v Fm I ie o mln achieve a removal of 50 percent with absorbent I at a stoichiometric ratio of 2.2. If we use absorbent IV, we need a stoichiometric ratio of 3.5 which means that we must use 60 percent more absorbent if we want the same level of SO, removal. Discussion A process model describing the contribution of a fabric filter to the SO, removal in a spraydry or coolside FGD process has been presented. An experimental method has also been presented for determination of the absorbent reactivity parameters used in the model. Because the objective of this report is limited to experimental and modeling methodology, the dependence of these parameters on different experimental conditions and different chemical and physical properties of the absorbents is not treated. An extensive analysis of different properties of the absorbents, such as particle shape, diameter distribution and hygroscopic properties, might be of interest. These questions will be treated in forthcoming studies at the Process Design Laboratory. References 1. J.R. Peterson, G.T. Rochelle, Aqueous Reaction of Fly Ash and Ca(OH)2 to Produce Calcium Silicate Absorbent for Flue Gas Desulfurization, Environ. Sci. Technol22: 1299 (1988). 2. J. Ablbeck, A Study on Spray Dry and Coolside Desulfurization, Doctoral Dissertation, Process Design Laboratory, Ab0 Akademi University, W.R. Seeker, S.L. Chen, J.C. Kramlich, S.B. Greene, B.J. Overmoe, Fundamental Studies of Lowtemperature Sulfur Capture by Dry Calcitic Sorbent Injection, Joint Symposium on Dry SO, and Simultaneous SO,/NO, Control Technologies, Volume 2, EPA600/ b, U.S. Environmental Protection Agency, NTIS PB (1986). 4. J. Klingspor, An Experimental Study on Spray Dry Scrubbing, Doctoral Dissertation, Department of Chemical Engineering, Lund Institute of Technology, Sweden, R.N. RuizAlsop, Effect of Relative Humidity and Additives on the Reaction of Sulfur Dioxide with Calcium Hydroxide, Doctoral Dissertation, The University of Texas at Austin, R.G. Rhudy, G.M. Blythe, Recent Results from the EPRI 2 1/2 MW Spray Dryer Pilot Plant, The 9th EPAEPRI Symposium on FGD, Washington, DC, EPA 600/986033b, T. Petersen, H.T. Karlsson, The Significance of Fly Ash in Wetdry Scmbbing of SO,, Chem. Eng. Technol. 11:298 (1988). 8. H. Yoon, J.A. Withum, W.A. Rosenhoover, F.P. Burke, Sorbent Improvement and Computer Modeling Studies for Coolside Desulfurization Joint Symposium on Dry SO, and Simultaneous S02/NOf Control Technologies, Volume 2, EPA600/986029b, U.S. Environmental Protection Agency, NTIS PB , T. Westerlund, S. Ronnblad, K. Fagervik, J. Ahlbeck, On the Modeling of Flue gas Desulfurization, Report 8783A, Process Design Laboratory, Department of Chemical Engineering, Abo Akademi University, J.C. Kramlich, D.K. Moyeda, G.H. Newton, R. Payne, Rate Controlling Processes in Humidification for Duct SO, Capture First Combined FGD and Dry SO, Removal, Symposium, St. Louis, T. Westerlund, Mass Balances in Flue gas Desulfurization, Report 66, Process Design Laboratory, Department of Chemical Engineering, Ab0 Akademi University, About the Authors The authors are research associates with the Department of Chemical Engineering, Ab0 Akademi University, Biskopsgatan 8, SF20500 Abo, Finland. This manuscript was peer reviewed. The revised manuscript was received on June 12, AIR & WASTE Vol. 44 April