Dry Regenerable Metal Oxide Sorbents for SO2 Removal from Flue

Nov 21, 2011 - Elevance Renewable Sciences, Woodridge, Illinois 60517, United States. ‡. Illinois Institute of Technology, Chicago, Illinois 60616, ...
0 downloads 0 Views 2MB Size
ARTICLE pubs.acs.org/IECR

Dry Regenerable Metal Oxide Sorbents for SO2 Removal from Flue Gases. 3. Long-Term Durability Vasudeo S. Gasvaskar† and Javad Abbasian*,‡ † ‡

Elevance Renewable Sciences, Woodridge, Illinois 60517, United States Illinois Institute of Technology, Chicago, Illinois 60616, United States ABSTRACT: The modified expanding grain model, presented in the part-2 of this series of papers, is extended to describe the sorbent behavior in a long-term cyclic process using the experimental data obtained in a 25-cycle sulfation/regeneration test. The model extension uses one adjustable parameter to describe the changes in the diffusional resistances occurring within the sorbent particle to predict the behavior of the sorbent undergoing an extended cyclic process. The model suggests that during the cyclic process, because of the cracks and fissures developing within the product layer of the grain, the product layer porosity and the product layer diffusivity of the grain as well as the tortuosity parameter gradually increase, resulting in lower intergrain diffusivity and lower overall sorbent reactivity. The model was used to predict the long-term performance of the sorbent to estimate the fresh sorbent makeup rate in the regenerative copper-oxide process. The results indicate that the sorbent makeup rate generally correspond to a useful sorbent life of higher than two years, but is highly sensitive to the slope of the extrapolation line for the sorbent deactivation.

’ INTRODUCTION Part 1 of this series of papers1 described the development and experimental evaluation of an alumina-supported-copper oxidebased sorbent (i.e., Cu-2). The important and relevant chemical and physical properties of the sorbent are presented in Table 1. The sorbent was evaluated in a thermogravimetric analyzer (TGA) and in a batch fluidized-bed reactor over 25 sulfation/regeneration cycles. The operating conditions for these tests are presented in Table 2. In both tests, the sorbent was loaded into the reactor and the reactor was heated to 450 °C in nitrogen. The reactor gas was changed to the SO2-containing mixture when the reactor reached 450 °C. In the TGA tests, the reaction continued until the sample weight gain during sulfation and weight loss during regeneration approached zero. In the fluidized bed experiments, the tests were terminated when the SO2 concentration in the reactor exit stream reached 100 ppmv during sulfation and dropped down to 5 ppmv during the regeneration. The detailed description of the reactor setup and the experimental technique were provided in part 1.1 Part 2 of this series of papers2 focused on the modeling of the experimental data obtained in part 1 using a modified expanding grain model, which describes the porous solid as an assemblage of a large number of small grains, surrounded by macro-pores, through which the gaseous reactant (i.e., SO2) diffuses to reach the surface of the grains.3 The expending grain model was used to relate the sulfation characteristics of the sorbent with its physical and chemical properties such as BET surface area, copper content, and particle porosity. In part 2, the diffusivity of SO2 through the product layer of the grains was also estimated and related to the chemical composition of the product layer. The effective diffusivity of the gaseous SO2 through the porous particle was described using a tortuosity parameter (α), which accounts for the structural changes occurring in the sorbent particle affecting the diffusion path of SO2 through the particle. The model was used to predict the sorbent behaviors in a r 2011 American Chemical Society

Table 1. Physical and Chemical Properties of the Sorbent

a

1st cycle

25th cycle

fresh

(oxidized form)

(oxidized form)

copper content, % aluminum content, %

14.1 43.2

14.4 44.1

14.1 44.2

oxygen (by difference)a, %

42.7

41.5

41.7

average particle diameter, μm

240

240

240

porosity, %

48.14

46.40

45.35

BET surface area, m2/g

150

125

111

bulk density, g/cm3

0.92

0.94

0.95

Oxygen corresponding to CuO and Al2O3.

fluidized-bed reactor with very good accuracy. The results showed that the model can quantify the relationship between the physical/chemical properties and the sulfation behaviors of the copper-based sorbent.2 The model was used to predict the optimum combination of the physical and chemical properties of the sorbent formulation that corresponds to the maximum sulfur loading. It should be noted that the regenerable copper oxide sorbents have been shown to have excellent catalytic activity for reduction of NOX to nitrogen (through reaction with ammonia) and are capable of reducing over 99% of NOX in the flue gas.4 This paper focuses on the modeling of the experimental results obtained in the 25-cycle sulfation/regeneration test performed with the Cu-2 sorbent to estimate the useful sorbent life in the regenerative copper oxide process. Received: September 1, 2011 Accepted: November 21, 2011 Revised: November 18, 2011 Published: November 21, 2011 213

dx.doi.org/10.1021/ie201974s | Ind. Eng. Chem. Res. 2012, 51, 213–217

Industrial & Engineering Chemistry Research

ARTICLE

Table 2. Operating Conditions in TGA and Fluidized-Bed TGA operating variable

fluidized-bed reactor

sulfation regeneration sulfation regeneration

initial temperature, °C

25

heat-up rate, °C/min

5

5

reaction temperature, °C

450

450

450

gas flow rate, SLM (STP) 0.85

0.85

0.667

0.667

weight of sample, g

0.012

30

30

0.012

25 450

gas composition, vol.% SO2

0.25

0.25

O2 N2

3.7 balance

3.7 balance

CH4

100

100

Figure 1. Structural changes occurring within a grain subjected to cyclic sulfation-regeneration operation.

Modeling of the Cyclic Test. The modified expanding grain

model2 was used to model the experimental results obtained in the 25-cycle sulfation/regeneration test performed with the Cu-2 sorbent in a thermogravimetric analyzer (TGA) and in a batch fluidized-bed reactor. It has been shown1 that the overall BET surface area and the particle porosity of Cu-2 sorbent decreases as the sorbent is subjected to the cyclic process, indicating changes in the grain radius and the pore structure of the sorbent particles. The chemical analyses performed on the spent sorbent confirmed that the chemical composition of the sorbent particle does not change during the cyclic operation (see Table1). It has been reported in the literature that the regenerable sorbents, when subjected to cyclic operation, can exhibit changes in the sorbent morphology caused by the migration of the active species toward the surface of particle. Mojtahedi and Abbasian5 reported that the decrease in the reactivity of a zinc-based regenerable sorbent was due to the migration of zinc under constant exposure to hightemperatures and pressures (650 °C and 20 bar). Although in this study, because of the relatively moderate temperatures encountered, such migration of the active species is not envisioned, other changes occurring in the properties of the product layer, such as its porosity(εs) can affect the product layer diffusivity of SO2 and the expansion factor Zv.2 A schematic depiction of the possible structural changes that may occur within a single grain and its product layer undergoing successive sulfation/-regeneration cycles is presented in Figure 1. During the cyclic operation, the grain radius is expected to gradually increase with the increasing number of cycles, affecting the pore size distribution and resulting in decreasing surface area and porosity.2 These changes can be described in the expanding grain model by the changes occurring in the product layer of the reacted grain. In part 2,2 it was shown that the expansion factor (Zv) (the ratio of the molar volume of the product to the reactant), defined as Zv ¼

Freactant  MW product Fproduct  MW reactant  ð1  εs Þ

Figure 2. Model fits of the 25-cycle sulfation-regeneration test performed in the TGA with the Cu-2 sorbent.

conversion, and resulting in a gradual decrease in the overall sulfur capacity of the sorbents (see Figure 4). To model the reactivity of the sorbent in this cyclic process, the product layer porosity of the grain (εs) was selected as the adjustable parameter, and the diffusion of SO2 through the product layer was assumed to vary with εs using the relationship ðDg Þj ðDg Þ1

¼

ðεs Þj ðεs Þ1

ð2Þ

where (Dg)j is product layer diffusivity of SO2 corresponding to the jth cycle (cm2/min), (Dg)1 is product layer diffusivity of SO2 corresponding to the first cycle (cm2/min), (εs)j is porosity of the product layer during the jth sulfation (adjustable parameter), and (εs)1 is porosity of the product layer during the first sulfation (assumed as 0.1). The tortuosity parameter α for the corresponding cycles was determined using its relationship with Zv, which was developed in part 2.2 Figure 2 shows a very good agreement between the experimental results and those predicted by the model. The increase in the sulfur loading during the initial cycles is a common characteristic of regenerable sorbents.5,6 The calculated values of the initial grain radius, the product layer porosity, the expansion factors, the product layer diffusivities, and the tortuosity

ð1Þ

determines the changes in the grain radius. As shown in Figure 1, the product layer of the grain may grow increasingly hollow because of the cracks and fissures developing with successive cycling. This ultimately leads to a greater expansion of the grains (see eq 1), in turn affecting the diffusivity of the gaseous SO2 through the porous sorbent particle. Additionally, as was explained in part 2,2 the initial porosity of the sorbent particle decreases during the sulfation/regeneration cycles, decreasing the maximum local 214

dx.doi.org/10.1021/ie201974s |Ind. Eng. Chem. Res. 2012, 51, 213–217

Industrial & Engineering Chemistry Research

ARTICLE

Figure 3. Effect of cycling on the product layer porosities (εs).

Figure 5. Extrapolation of the BET surface area changes during extended cycling.

Figure 4. Comparison of the model predictions of the experimental data on sulfur loadings in a fluidized-bed reactor.

Figure 6. Extrapolation of the particle porosity changes during extended cycling.

parameter for selected cycles are presented in Table 2. As shown in Figure 3, the change in the product layer porosity appears to stabilize after about twenty cycles. Furthermore, the estimated values of εs appear to decrease during the first few cycles, which may be attributed to the initial activation/conditioning during the initial cycles.5,6 The reduction in the product layer porosities results in a more compact product layer, which leads to lower values of Zv and α and results in higher overall reaction rates and higher sulfur loadings during the first few cycles. The calculated values of Dg and α determined in part 22 were used to predict SO2 breakthrough curves and the sulfur loading of the sorbent (at an exit SO2 concentration of 100 ppmv). A comparison of the experimental results of sulfur loadings for Cu2 sorbent obtained in the 25-cycle test with those predicted by the model is shown in Figure 4, indicating that model is capable of predicting the behavior of the sorbent in cyclic tests conducted in both the thermogravimetric analyzer (TGA) and the fluidizedbed with acceptable accuracy. Process Implication. A preliminary process analysis was performed to estimate the fresh makeup rate and the useful sorbent life to assess the technical viability of the regenerable copper oxide process. Given that the reactivity and hence the absorption capacity of regenerable sorbents are expected to gradually decrease in the cyclic process, regenerative processes require replacement of the spent sorbent with fresh sorbent (i.e., makeup) to maintain the required bed inventory at the required sulfur absorption capacity. If the rate of sorbent loss due to attrition/elutriation is

higher than the fresh makeup rate dictated by the sorbent deactivation, the makeup rate is simply the same as the rate of sorbent loss. Otherwise, the rate of makeup is dictated by the sorbent deactivation. In this case, additional amount of sorbent must be removed from the reactor. The “forced” sorbent removal rate is calculated from the following equation: sorbent removal rate ¼ make-up rate dictated by deactivation  sorbent loss rate due to attrition=elutriation ð3Þ

To estimate the rate of sorbent loss by elutriation, it was assumed that the expected elutriation rate of solid sorbents in fluid bed processes (i.e., Floss) is proportional to the attrition index (AI) determined by the ASTM-D5757 method.7 Furthermore, the useful sorbent life dictated by sorbent loss due to elutriation is simply calculated from the following equations: Floss µ ðAIÞ

ð4Þ

ðLU Þloss ¼ W0 =Floss

ð5Þ

ðLU Þloss µ 1=ðAIÞ

ð6Þ

where W0 is the weight of the sorbent in the reactor and (LU)loss is the estimated useful sorbent life dictated by sorbent loss due to attrition/elutriation. In these calculations, the fluid catalytic cracking (FCC) catalyst was selected as the baseline material 215

dx.doi.org/10.1021/ie201974s |Ind. Eng. Chem. Res. 2012, 51, 213–217

Industrial & Engineering Chemistry Research

ARTICLE

the sorbent is calculated from FC ¼ W0 =θ

ð7Þ

where θ is residence time of the sorbent, which is an operational parameter calculated based on the sorbent reactivity and the reactor configuration. The fresh sorbent makeup rate and the useful sorbent life dictated by the chemical deactivation, Fr‑CD, can be calculated from

Table 3. Model Generated Parameters for the Cyclic Test grain radius cycle Microns

product layer product layer diffusivity, porosity, εs

tortuosity

Dg, cm2/min

parameter, α

1

9.02

6.7  1010

3.5

3

11.18

0.083

5.53  1010

3.1

7 10

11.66 11.97

0.13 0.15

8.6  1010 1.01  109

4.2 4.9

15

12.07

0.18

1.21  109

6.1

0.1

9

20

12.07

0.185

1.24  10

6.3

25

12.18

0.2

1.34  109

7.1

for estimation of the rate of attrition of the Cu-2 sorbent. Since the estimated life of a typical FCC catalyst with an attrition index of 4% is about six months in the FCC fluid-bed process,8 the estimated (LU)loss of the Cu-2 sorbent with an attrition index 1%, was estimated to be 2 years (i.e., 4 times that of the FCC catalyst). It should be noted that the assumed linear dependence of the useful life of the sorbent on the reciprocal of the attrition index is a very conservative estimate because the physical attrition encountered in a fluidized-bed reactor system is expected to be far less stringent than that in the ASTM-D5757 test. In other words, the estimated useful life of the Cu-2 sorbent formulation (due to attrition) is expected to be longer than the assumed 2 years. To estimate the useful sorbent life dictated by sorbent deactivation, the model-estimated parameters for the 25-cycle sulfation-regeneration test were extrapolated (Figures 5 and 6) to predict the sorbent reactivity after a significantly large number of cycles. Since both the surface area and the particle porosity appeared to stabilize after about 25 cycles, both parameters were linearly extrapolated. The porosity of the product layer, which is related to the cracks and the fissures occurring in the copper sulfate product, is expected to remain significantly lower than the particle porosity, and as shown in Figure 3, appears to stabilize at about 0.2. Therefore, since the product layer porosity is not expected to increase significantly with increasing cycles, it was assumed to level off at a value of about 0.22. These extrapolated values were used in the expanding grain model to calculate the maximum number of cycles and the fresh sorbent makeup rate due to sorbent deactivation. The solids circulation rate (FC) of

ð8Þ

ðLU ÞCD ¼ θ  Cmax

ð9Þ

where Cmax represents the maximum number of cycles at which the sorbent can meet the SO2 emission limit. Since Fr‑CD is expected to be sensitive to the slope of extrapolation line (baseline slope) of the experimental data, a sensitivity analysis was performed by assuming two additional extrapolation lines of the deactivation slopes (i.e., 1/2  baseline, and 2  baseline) shown in Figures 5 and 6. The dependence of the estimated useful life of the sorbent on the slope of the extrapolation line is presented in Figure 7. The results indicate that, if the sorbent characteristics deteriorate at a rate equal to or lower than that shown in Figures 5 and 6, the makeup rate will be generally dictated by the physical attrition (which was assumed to be two years). However, at a higher rate of deterioration (e.g., 2  baseline), the useful life of the sorbent will be dictated by the deterioration of the sorbent reactivity and will be significantly lower.

Figure 7. Dependence of estimated useful sorbent life on the slope of extrapolation.

(rg  103),

Fr-CD ¼ FC =Cmax

’ CONCLUSIONS The modified expanding grain model described in part 2 of this series of papers was extended to successfully model the sorbent behavior during the 25-cycle sulfation/regeneration test. The model suggests that during the cyclic process, the product layer porosity, the product layer diffusivity, and the tortuosity parameter gradually increase, resulting in lower intergrain diffusivity and lower overall sorbent reactivity during the cyclic process. The results generated by the model were shown to be in close agreement with the experimental data from both the TGA and the fluidized-bed reactors. The results also indicate that the estimated useful sorbent life is higher than two years. However, this estimate is highly sensitive to the slope of the extrapolation line for the sorbent deactivation. ’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ NOMENCLATURE Cmax = maximum number of cycles for which the sorbent can maintain the required SO2 removal efficiency (Dg)j = product layer diffusivity of SO2 corresponding to the jth cycle, cm2/min (Dg)1 = product layer diffusivity of SO2 corresponding to the first cycle, cm2/min FC = solids circulation rate, kg/min Floss = rate of sorbent loss due to attrition, kg/min Fr‑CD = fresh sorbent makeup rate due to chemical deactivation, ton/year (LU)loss = useful sorbent life dictated by sorbent loss due to attrition, year 216

dx.doi.org/10.1021/ie201974s |Ind. Eng. Chem. Res. 2012, 51, 213–217

Industrial & Engineering Chemistry Research

ARTICLE

(LU)CD = useful sorbent life dictated by chemical deactivation, year MWj = molecular weight of species j, g/mol W0 = weight of the solids in the reactor, kg Zv = expansion factor, dimensionless εs = porosity of the product layer owing to the cracks and fissures, dimensionless (εs)j = porosity of the product layer during the jth sulfation, adjustable parameter (εs)1 = porosity of the product layer during the first sulfation, assumed as 0.1 Freactant = density of the reactant, g/cm3 θ = sorbent residence time, min

’ REFERENCES (1) Gavaskar, V. S.; Abbasian, J. Dry regenerable metal oxide sorbents for SO2 removal from flue gases. 1. Development and evaluation of copper oxide sorbents. Ind. Eng. Chem. Res. 2006, 45 (1), 5859. (2) Gavaskar, V. S.; Abbasian, J. Dry regenerable metal oxide sorbents for SO2 removal from flue gases. 2. Modeling of the sulfation reaction involving copper oxide sorbents. Ind. Eng. Chem. Res. 2007, 46 (2), 1167. (3) Szekely, J.; Evans, J. W.; Sohn, H. Y. Gas-Solid Reactions; Academic Press: New York, 1976, Chapter 2. (4) Abbasian, J. A A regenerative fluid-bed process to control SOX and NOX.Final Technical Report to Illinois Clean Coal Institute, http:// www.icci.org/reports/01-1Abbasian2.2B-1.pdf, 2002. (5) Mojtahedi, W.; Abbasian, J. H2S Removal from Coal Gas at Elevated Temperature and Pressure in Fluidized Bed with Zinc Titanate Sorbents. 1. Cyclic Tests. Energy Fuels 1995, 9, 429. (6) Abbasian, J.; Slimane, R. B.; Regenerable Copper-Based, A Sorbent for H2S removal from coal gases. Ind. Eng. Chem. Res. 1998, 37, 2775. (7) ASTM Annual Book of Standards, Determination of Attrition and Abrasion of Powdered Catalysts by Air Jets, Petroleum Products and Lubricants (III): D-2981- Vol. 05.03, 1985. (8) Kunni, D; Levenspiel, O. Fluidization Engineering; ButterworthHeinemann: Boston, MA, 1977.

217

dx.doi.org/10.1021/ie201974s |Ind. Eng. Chem. Res. 2012, 51, 213–217