Enhanced flue-gas denitrification using ferrous.cntdot.EDTA and a

Mar 1, 1991 - Enhanced flue-gas denitrification using ferrous.cntdot.EDTA and a polyphenolic compound in an aqueous scrubber system. M. H. Mendelsohn ...
0 downloads 0 Views 576KB Size
Energy & Fuels 1991,5, 244-248

244

Enhanced Flue-Gas Denitrification Using FerrousDEDTA and a Polyphenolic Compound in an Aqueous Scrubber System M. H. Mendelsohn* and J. B. L. Harkness Argonne National Laboratory, Energy Systems Division, 9700 S. Cass Ave., Argonne, Illinois 60439 Received October 4,1990. Revised Manuscript Received December 21, 1990

Previous work in this laboratory has involved studying the possibility of combined N0,/S02 scrubbing using various aqueous chemistries with a metal chelate additive. Recently, we have focused our work on the metal chelate ferrous.EDTA [Fe(II).EDTA]. A problem encountered in the practical application of Fe(I1)-EDTA is that the ferrous ion has been found to oxidize to the corresponding ferric species, leading to a decrease in NO, removal for the scrubbing solution containing the additive. We have found that addition of a polyphenolic compound leads to a sustained high NO, removal under various oxidizing conditions. We believe that the improved performance of Fe(II).EDTA is due to the known capabilities of these organic compounds both to inhibit oxidation of ferrous chelates by dissolved oxygen and to rapidly reduce any ferric ions back to the original ferrous species. These effects are illustrated by the following chemical reactions: 02(1)+ organic oxidized organic, and Fe(II1) organic Fe(I1) + oxidized organic.

+

-

-

Introduction The use of metal chelate additives in an aqueous scrubbing environment for combined N0,/S02 removal from oxygen-containing flue gases has been investigated in this laboratory for several Recent work with the metal chelate Fe(II).EDTA has shown initially high NO, removals. However, these removals decline with time as a function of the amount of oxygen gas in the feed gas stream. Because of this dependence on oxygen concentration in the feed gas, we have attributed the decline in NO, removal to the oxidation of the Fe(II).EDTA additive to the ferric form. One possible solution to this problem would be to add a secondary additive to the system, one that is capable of either preferentially reacting with any dissolved oxygen or reducing any oxidized ferric species back to the ferrous form. These chemical reactions may be summarized simply as follows: 02(1)+ additive oxidized additive Fe(II1) + additive

-

Fe(I1)

+ oxidized additive

From an examination of previous work in the literature, we have found one class of compounds that is capable of performing both of the above-stated reactions. Theis and Singe+ found that certain polyphenolic compounds, which are products of natural vegetative decay, are capable of significantly affecting the rate of oxidation of ferrous iron. For example, an equimolar amount of tannic acid was able M to maintain a ferrous iron concentration of 5 X unchanged for 7 days in the presence of 0.5 atm of OF This study also showed that such compounds as tannic acid, pyrogallol, and gallic acid can rapidly reduce aqueous ferric (1) Harkness, J. B. L.; Doctor, R. D. "Development of Combined Nitrogen Oxide/Sulfur Oxide Environmental-Control Technology*. Argonne National Laboratory Report ANL/ECT-14, Argonne, IL, Aug. 1985 (also available through NTIS). (2) Harkneea, J. B. L.; Doctor, R. D. Simultaneous NO SO, Removal In Aqueous Scrubber Chemistries. Presented at the A.I.6h.E. National Meeting, New Orleans, LA, April 1986. (3) Harkness, J. B. L.; Doctor, R. D.; Wingender, R. J. US.Patent No. 4,612,175 (1986). (4) Theis, T. L.; Singer, P. C. Enuiron. Sci. Technol. 1974, 8, 569.

ions to ferrous ions. Also, phenols (such as gallic acid) are well-known antioxidant^.^ Because of these properties of polyphenolic compounds, we have investigated the effects of tannic acid, pyrogallol, and gallic acid as secondary additives in aqueous scrubbing systems containing the primary additive Fe(II).EDTA. Using these secondary additives, we have been able to maintain NO, removals as high as 6045% for up to 2 h.

Experimental Setup and Procedures An overall block diagram of the experimental setup is shown in Figure 1. A description of each section of the laboratory facility follows. Flue-Gas Feed System. The flue-gas feed system used pure gases to produce a controlled gas composition (see Figure 2), which allowed for significantly better control than would have been possible with a coal-fired boiler. At the same time, the system gave a realistic gas composition for a cod-fired utility boiler using high-sulfur coal. The operator could also make controlled variations in the gas composition in order to investigate the effects of individual components. The feed system was designed to deliver 200 L/min of flue gas a t standard conditions. The normal flow rate was 100 L/min. At the maximum flow rate, the feed system could synthesize a flue-gas composition with SO2 concentrations up to 3000 ppm, NO concentrations up to 1000 ppm, and NOz concentrations up to 450 ppm. The feed system consisted of two parallel streams. The first stream metered nitrogen and air through the flowmeter and the humidifier. Gaseous N2 was supplied by vaporizing liquid N2in an external heat exchanger, and air came from the laboratory supply system. In order to maintain the pressure in the liquidnitrogen dewar, a small fraction of the nitrogen vapor was reinjected into the dewar by a small refrigeration compressor. This recycled nitrogen passed through an activated-charcoal filter to remove any possible contaminants introduced by the compressor. Both the nitrogen and the air flow rates were controlled by metering valves operating in conjunction with downstream flow controllers and were piped to a single Flowtech flowmeter, the primary flow-measurement instrument in the feed system. The second stream consisted of carbon dioxide and the pollutant gases. ( 5 ) Loginova, L. F.; Medyntaev, V. V.; Khomutov, B. I. Zh.Obshch. Khim. 1972, 42, 739.

0887-0624/91/2505-0244$02.50/00 1991 American Chemical Society

Energy & Fuels, Vol. 5, No. 2, 1991 245

Enhanced Flue-Gas Denitrification Flus-Gas

I Gas-Feed System

Figure 1. Block diagram of aqueous-phase scrubber system.

lEIIIBUI( System

Low-Temp Gas-Phase Control A-Scrubber Analyzer

U Distilled Water

U

-

I I

uu

Figure 2. Feed system for simulated flue gas.

COPwas supplied from a single cylinder, connected in parallel with a standby cylinder to an automatic switchover system. This switchover system monitored the supply presure from the active C02cylinder and, when the pressure dropped below 50 psig, shut off the active cylinder and started the flow from the standby cylinder. The switchover system also sounded an alarm so the operator could replace the spent cylinder with a fresh one. The COPflow rate was also controlled by a metering-valve/flow-controller combination. The COPflowed to a manifold, where it picked up and diluted the pollutant-gas flows (NOand NO,). The SO2 was piped separately to the main feed stream because of concern over homogeneous gas-phase reactions between the NO, and the SO2. For safety, each of the pollutant gases was supplied from a cylinder either in the bench hood or directly in the air sweep from the laboratory airspace into the hood. For the second stream, containing the pollutant gases, SO, was supplied a t 39-40 psig from a single-stage regulator on a 5-lb cylinder of 99.9%-pure liquid. NO and NO, were supplied from size 1A and 3P cylinders (Matheson Gas,Inc.), respectively. The 99.0%-pure NO was delivered to the feed system a t 25-26 psig from a two-stage regulator. The 99.9%-pure NO2 cylinder was heated and adjusted to about 20 psig by using a two-stage regulator. The three gases were piped separately to three-way valves, which either directed the pollutant gases to their flow-control valves or shut off the gas flow and directed nitrogen to the feed system instead. This use of three-way valves permitted a quick purge from the feed system of all gases representing a health or safety hazard. Each flow-control valve was stabilized by a downstream flow controller. SO2 gas flow was monitored with a gas rotameter. Flue-gas moisture was added to the nitrogen/air stream by controlling the liquid operating temperature of a humidifier, which was heated by a pair of mantle heaters. The humidifier was a 6-Lglass vessel partially filled with 6-mm glass beads. The nitrogen/air stream was introduced below the liquid level of distilled water and bubbled through the packed bed of glass beads to the exit in the top of the vessel. The glass beads ensured uniform contacting and mixing between the gases and distilled

water. As distilled water was vaporized into the gas stream, a level controller added makeup water. Following the humidifier, the gas stream was heat-traced to keep its temperature above the dew point, and the carbon dioxide and pollutant gases were then added to the nitrogen/air stream. The feed stream was heattraced following the humidifier up to the three-way valve used to direct the flue gas to the experimental system or to the exhaust system. A second three-way valve was used to direct a purge stream of Nz either to the exhaust or to an experiment. This valving configuration permitted the flue gas to bypass the experiment while flow rates and gas compositions were being adjusted and while the scrubber was brought to the target conditions of an experiment. Whenever the flue gas was bypassed, a nitrogen flow was directed to the experimental system at approximately one-quarter of the flue-gas flow rate. A sample of the flue gas was withdrawn just before the main three-way switching valve and piped directly to the flue-gas analyzer system. This sample line also was heat-traced, from the sample point to the ice traps immediately before the analyzer system. The precise order in which the simulated flue-gas components are mixed has been modified from that used To obtain the results described below, flue gas was prepared by first setting the NO level at 450 ppm in the presence of C02,02,and N2 gases only. In all runs, the fluegas mixture contained 14.5% COP,5.4% O,, and N, as the balance. After the metering valve for the NO gas was set to give 450 ppm, a shut-off valve was closed, and NO2 was then set in the same COP,O,,and Nz mixture. NO,, calculated as a difference between measured NO, and measured NO, was set (except where noted below) at around 75 ppm. The preset amount of NO was then added to the NOz. Finally, SO, was added to the feed-gas mixture and adjusted to 3000 ppm (except where noted below). This new feed-gas preparation procedure has improved the reproducibility and reliability of our removal measurements compared with the previously used method.lS2 Except where noted, approximately 8% water vapor was also added to the simulated flue-gas mixture. Flue-Gas Analyzer System. The flue-gas analyzer system was originally assembled by Beckman Instruments, Inc., following a design developed jointly with the Argonne project staff; later, this system was extensively modified to improve its performance. The analyzer flow sheet shown in Figure 3 represents the current configuration of this system. The system consisted of a sampleconditioning section; individual gas analyzers for CO,, O,,SO,, and NOJNO; and a six-point recorder. The following Beckman gas analyzers were used C02, Model 864 IR Analyzer; 02,Model 755; SO2,Model 865 IR analyzer; and NO/NO,, Model 951A. The system was originally equipped with a timer-operated switch to allow automatic, alternate monitoring of two gas streams. However, the stream-selection function was later controlled by the computer/data-logger system, so the user had three options for controlling the analyzer system: (1)manual control (all stream and NO,-mode selection was done by the user); (2) automatic control (the time switched between gas streams according to a user-set program, although the user still had to switch between NO, modes manually), and (3) computer-operated control (the computer controlled both switching,functions). Both flue-gas and effluent samples were delivered to the analyzer in Teflon tubing surrounded by heat-traced copper tubing. Each stream first passed through a trap submerged in a wet-ice bath to remove excess moisture in the samples. A t a flow rate of approximately 2 L/min, the gas residence time in these traps was less than 10 s. In several checks for possible gas scrubbing in the traps, no removal of SO2 or NO/NO, was observed. Following passage through the traps, each sample flowed through in-line filters to protect the downstream valves and the sample pump from entrained solids. The two streams flowed through separate rotameters equipped with flow-control valves to maintain the flow rates at 2 L/min. Lower flow rates would result in erroneous NO,-analyzer readings and slow the response time of all the analyzers. The sample-stream selection was achieved by a pair of solenoid-operated three-way valves. These two valves were normally closed, and the switching control system was designed so that only one of them could be opened a t a time. When a valve was in the closed configuration, the sample stream was diverted to the exhaust system (using an air-operated jet pump to maintain the sample flow rate, so that a fresh sample would

Mendelsohn and Harkness

246 Energy & Fuels, Vol. 5, No. 2, 1991 I

Zero Gas

Calibratbn Gases

Figure 3. Configuration of gas-analysis system. be available for analysis a t all times). The feed-sample purge line also was equipped with a flow-control valve in order to maintain the feed-sample flow rate at 2 L/min even when the feed sample was being diverted to the exhaust system. This provision ensured that the flow rate to the experiment would not be affected by stream switching. The stream to be analyzed flowed to the sample pump, a diaphragm pump (Air Dimensions, Inc., Model M19310) that could deliver 18 L/min a t open flow. The flue-gas sample passed through a pair of membrane driers in series. These driers (Perma Pure Products, Inc., Model PD-625-24E) contained multiple tubes that were permeable to water but not to any of the other flue-gas components being analyzed water passed through the tubes and was removed by a flow of dry air moving through a shell containing the tubes. The dry air was supplied to the sample driers from a heatless, regenerative drier (General Cable Co., Model HF200). For safety, the wet purge air from the driers was pumped to the exhaust system by a second air-operated jet pump. The regenerative drier also supplied air for the ozone generator in the NO, analyzer. The flue-gas sample then passed to a pair of manually operated stream-selector valves, used in conjunction with the analyzer zeroing and calibrating functions. These five-way selector valves directed the flue-gas sample to the individual analyzers. The gas samples leaving the analyzers were piped directly to the exhaust system. Aqueous-Phase Scrubber System. Figure 4 shows a flow diagram of the aqueous scrubber system that was used. A disk-and-donut scrubber having four stages was used instead of the previously described flooded column.'tZ The dimensions of the scrubber are approximately 7.6 cm (inner diameter) by 52.7 cm from the sieve plate to the bottom of the inlet nozzle. A sieve plate having holes 0.48 cm (3/16 in.) in diameter with a totalopen area of 10.3% was placed a t the bottom of the scrubber in order t o provide the capability of having some liquid holdup in the column. Also, an approximately 10-L holding tank was added to the system and connected to the bottom of the scrubber column. Circulation rates from the holding tank to the top of the scrubber could be varied from about 330 to 1420 mL/min. For the experiments described below, an average circulation rate of 890 mL/min was used. With this circulation rate, the liquid level was maintained at 33-36 cm above the sieve plate. This level was controlled by manually varying the circulation rate in the range of 790-985 mL/min. At this liquid level, the SOz removals in all the runs described below were 99% or better. All of the scrubbing solutions used in the experiments contained 0.31 M sodium

C

h

d

Feed Tank

I I

Figure 4. Flow diagram of laboratory aqueous scrubber. carbonate and 0.067 M Fe(II)-EDTA in a total volume of approximately 3.6 L. This chemistry (without the Fe(I1)-EDTA) simulates the scrubbing step in an industrial double-alkali process, where the spent scrubber liquor would be regenerated by lime or limestone (the second alkali). In these experiments, the scrubber liquor was not regenerated. The pH was not specifically controlled for these experiments, but the following behavior was observed in all the tests reported here: At the beginning of each run, the pH was about 9.5, but it dropped to less than 8 after the first 10-15 min; thereafter, pH remained in a buffered regime and slowly dropped (during a 2-h experiment) from 7.5 to about 6.5. Finally, the scrubbing solution temperature was maintained a t 50 OC by using a thermostatically controlled heating tape around the 10-L holding tank containing the returned scrubbing liquor. Computer/Data-Logger System. The laboratory system included a data logger (Fluke Model 2242B) and an IBM PC Model 5150. The data logger could scan 60 channels of type T,

Energy & Fuels, Vol. 5, No. 2, 1991 247

Enhanced Flue-Gas Denitrification

0 with water vapor added 0 without water vapor added

0 Fe(il).EDTA abne 0 Fe(iI).EDTA with tannic aca

.- . 451 404

0

20

10

50

40

30 Time (min)

Figure 5. NO, removal for Fe(II).EDTA alone vs Fe(II).EDTA with tannic acid. 70

20

40

60

80

100

1 120

Time (min)

Figure 7. Comparison of NO, removal for Fe(II)-EDTA and pyrogallol with or without moisture.

t

254

0

0

60

10

20

30

40

50

60

70

80

90

IAO

Time (min)

Figure 6. Comparison of NO, removals with or without the secondary additive pyrogallol.

K, J, or D thermocouples or analog dc voltages of 0-40 mV, 0-400 mV, 0-4 V, or 0-40V. The data logger digitized these input signals with a resolution of 10 r V (corresponding to 0.2 "C) and converted the thermocwple voltage data to Celsius-scaletemperatures (using

75'

0

0

75pp"Q

OppmNq

70.

internal linearization and temperature-compensation routines). The data-logger control program waa simply a channel-by-channel description of each of the input variables, together with format requirements for the transfer of data vectors from the data-logger to the minicomputer. A permanent copy of the data-logger control program, along with the data from several experiments, was stored on 5.25-in. floppy disks. The control program logged three pairs of effluent samples and one pair of feed samples in each cycle. This control function was accomplished by using a two-bit chip in an interface board designed and built a t Argonne. Setting bit no. 0 to unity activates an optically coupled relay capable of switching 120-V ac loads. This relay activates a double-pole relay that controls two solenoid-operated valves for the analyzer stream selection (Le., either feed or effluent stream). Bit no. 1controls the NO,-mode selection directly through a device installed on the NO,/NO analyzer. This device operates an internal solenoid valve, which either directs the sample stream through the NO,-to-NO converter or bypasses the converter. If the sample stream is directed through the converter, the analyzer determines total NO, concentration; if the converter is bypassed, only the NO concentration is determined.

Results and Conclusions All experimental comparisons in this paper are made using total NO, removal data, because we have observed that the presence (as in the feed stream) or absence (as in the effluent stream) of sulfur dioxide can alter the NO or (NO,-NO)value but has little effect on the total NO, value. This "SOz effect" depends on the amount of unmixed nitrogen dioxide in the feed gas mixture and most likely arises from a gas-phase reaction between SOz and NOz. Because of the relatively small amount of NO2that

Time (min)

501 45

404 0

20

40

60

80

100

1

120

Time (min)

Figure 9. Comparison of NO, removals for three different NOz levels.

we are adding in our new feed-gas preparation procedure, as described above, this effect is small. In fact, although we still consider NO, removals more reliable, NO removals

Mendelsohn and Harkness

248 Energy & Fuels, Vol. 5, No. 2, 1991

t

801

751

0 Fe(ll).EDTA with pyrogallol 0 Fe(ll).EDTA with gallic acid

70

501

pyrogallol for feed-gas mixtures with and without added moisture. NO, removal with added moisture was consistently about 6% greater than without added moisture. This effect is probably indicative of gas-phase interactions of NO and/or NO2with water vapor, as discussed earlier.' The next three figures illustrate the effects on NO, removal of various changes in the feed-gas stream composition. Figure 8 compares NO, removals with pyrogallol for feed-gas mixtures containing 1500 and 3000 ppm SOp. NO, removal was 9% higher, on average, with 3000 ppm SOp However, we find it interesting that after 2 h of scrubbing with 1500 ppm SOz, the NO, removal had increased to about 56%, only slightly lower than the 60% removal with 3000 ppm SO2. Figure 9, a and b, shows NO, removals for feed-gas mixtures with 0 ppm versus 75 ppm NO2and 0 ppm versus 150 ppm NO2,respectively. Figure 9a shows that the removals were virtually identical for the first 90 min of each test; but for the last 30 min, the run without NOz yielded a removal about 4 % higher than the one with 75 ppm NO2. Figure 9b shows that the run with 150 ppm NO2gave slightly improved NO, removal for the 10-9O-min interval (about 3%), but, again, the run without NOz gave a removal about 3% higher for the last 30 min. Because we estimate the experimental error in our NO, measurements to be f 2 percentage points, the point to be stressed here is that NO2levels of 0-150 ppm make relatively little difference in terms of total NO, removal. Finally, Figure 10 shows NO, removals for the secondary additives gallic acid and pyrogallol under identical conditions. The NO, removal for these two additives was very similar. Although pyrogallol appeared to be slightly better in the 20-80-min interval, the average difference was only 3%, which is near the limit of our estimated experimental error.

457 400

20

40

60

80

100

120

Time (min)

Figure 10. NO, removal comparison for secondary additives pyrogallol and gallic acid.

were never more than a few percent different from the reported NO, removals in the cases discussed below. Our initial experiment was performed with tannic acid as the secondary additive, using the previously described flooded column Figure 5 shows NO, removal for a baseline run with a concentration of 0.067 M Fe(II).EDTA (Fe(I1):EDTA = 1:l) alone versus that for an identical run with the addition of 0.01 M tannic acid. This first-try experiment showed a significant improvement in NO, removal, from about 14% to about 40% in the stable portions of both curves. After this experiment, the scrubber column was changed from the flooded type to the disk-and-donut type described above. Because of several problems with tannic acid, including the viscosity changes it caused, its high molecular weight, and its relatively high cost, we performed the remaining experiments with the polyphenolics pyrogallol and gallic acid. After trying several ferrous:polyphenolicmolar ratios, we found the most effective molar ratio to be approximately 1:l. This ratio of primary additive to secondary additive was used in all the experiments described below. Figure 6 shows NO, removal for Fe(I1)-EDTA alone versus that with pyrogallol as a secondary additive. This figure clearly demonstrates the declining NO, removal with Fe(II)*EDTA alone versus the slightly increasing removal with pyrogallol. After 90 min, NO, removal with pyrogallol was twice that of Fe(II)*EDTA alone (64% vs 32%). The tests represented by this figure are the only ones reported in this paper that did not have moisture added to the feed-gas stream. Figure 7 compares NO, removals with

Acknowledgment. This work is supported by the U.S. Department of Energy, Assistant Secretary for Fossil Energy, under contract W-31-109-ENG-38,through the Pittsburgh Energy Technology Center (PETC). We acknowledge the support provided by Perry Bergman and Charles Drummond of the PETC, and the direction and encouragement given by C. David Livengood of ANL. In addition, we express our deep appreciation and gratitude to Sherman Smith for his invaluable contributions on the modifications and maintenance of the experimental apparatus, as well as the performance of the tests described herein.