Advanced CFD Post-Processing for P

Advanced CFD Post-Processing for P PREDICTING THE IMPACT OF WET FGD SCRUBBING ON HG EMISSIONS FROM COAL- FIRED POWER STATIONS Paper #590 Stephen Niksa Niksa Energy Associates 1745 Terrace Drive Belmont, CA 94002 neasteve@pacbell.net Naoki Fujiwara Coal Research Laboratory Idemitsu Kosan Co., Ltd. Chiba 299 0267, Japan nffish@po.iijnet.or.jp ABSTRACT This paper introduces a predictive capability for Hg retention in any Ca-based wet FGD, given Hg speciation at the FGD inlet, the flue gas composition, and the SO capture 2 efficiency. An analysis of data from 17 full-scale wet FGDs based on thermochemical equilibrium yields highly accurate predictions for total Hg retention with no parameter adjustments. For the most reliable data, the predictions were within measurement uncertainties for both limestone and Mg/lime systems operating in both forced and natural oxidation mode. With the ICR database, the quantitative performance was almost as good for the most modern FGDs, which probably conform to the very high SO 2 absorption efficiencies assumed in the calculations. The large discrepancies for older FGDs are tentatively attributed to the unspecified SO capture efficiencies and operating 2 temperatures and to the possible elimination of HCl in prescrubbers. The equilibrium calculations suggest that Hg retention is most sensitive to inlet HCl and O levels and the 2 FGD temperature; weakly dependent on SO capture efficiency; and insensitive to HgCl, 22 NO, Ca/S ratio, and slurry dilution level in limestone FGDs, and to MgSO levels in 3 Mg/lime systems. Consequently, FGDs will retain less Hg from flue gases derived from subbituminous coals than from typical hv bituminous coals, even when the levels of oxidized Hg at the FGD inlet are the same. Also, systems with prescrubbers to eliminate HCl probably retain less Hg than fully integrated FGDs. The analysis also predicts re-0emission of Hg, but only for inlet O levels that are much lower than those in full-scale 2 FGDs. 1 INTRODUCTION The coal burning utility industry is expected to mount a massive response to comply with as-yet unspecified regulations on mercury emissions from power plants that take effect in 2008, and a variety of control strategies are now under intensive development. The most economical approaches utilize existing particulate control devices (PCDs) and flue gas desulfurization (FGD) scrubbers, often in conjunction with additives, sorbents, catalytic pretreatment reactors, and other supplemental measures to enhance Hg recoveries. A multitude of factors affects the performance of PCDs and scrubbers via their impact on Hg speciation at the inlet to the control device. Oxidized vapor species (Hg(II)), particularly HgCl, are water-soluble and therefore easily dissolved in scrubbing 20solutions, whereas elemental (Hg) vapor is insoluble and therefore able to pass through 0scrubbers into the smokestack. Some of the dissolved Hg(II) may be re-emitted as Hg. Under the best circumstances with high volatile (hv) bituminous coals, over 90 % of the 1inlet Hg vapors are retained in wet FGDs on scrubber solids and in wastewater, but Hg retention can also be negligible with subbituminous coals and lignites. Even with very similar hv bituminous coals, Hg retention in full-scale wet FGDs ranges from 40 to 85 % (as seen below). Hg retention is affected by the liquid-to-gas ratio, but is insensitive to slurry pH over the normal operating range, and to whether a limestone slurry is processed 2in forced or natural oxidation mode. The impact of flue gas composition, especially acid gas (HCl, NO, SO) concentrations, has not yet been characterized, but will probably be XX strong because the Hg species concentrations are much too low to affect the chemistry of any of the other major species. ompanies contemplating Hg control with wet FGDs need to know in advance how C much of the Hg at the FGD inlet will be retained in the scrubber. This paper introduces a predictive capability for Hg retention in any Ca-based wet FGD, given Hg speciation at the FGD inlet, the flue gas composition, scrubber operating conditions, and the SO 2 capture efficiency. An equilibrium analysis accurately predicts Hg retention in a variety of full-scale wet FGDs with no parameter adjustments whatsoever. Based on this performance, the analysis is ultimately used to identify the major factors affecting Hg 0retention in wet FGDs. It also associates Hg re-emission with abrupt transitions among certain aqueous sulfur species. Whereas the current evaluations cover fully integrated, Ca-based FGDs, the analysis can be expanded to systems with pre-scrubbers or other scrubbing reagents, and to dry scrubbers. DATABASE Each dataset in this study was obtained on a full-scale, wet FGD using either limestone or Mg/lime reagents. Systems with pre-scrubbers to eliminate acid species were omitted at this stage. Various reagent delivery systems are represented, including spray or packed towers and natural and forced oxidation, as well as Mg/lime systems. But dual alkali, sodium carbonate, ammonia, and seawater scrubbers are relegated to future expansions. Each dataset must contain sufficient information to accurately estimate a flue gas composition. At a minimum, this comprises an ultimate analysis of the fuel and furnace stoichiometry although, in practice, measured compositions for O, NO, HCl, and SO 2X2 2 conversion efficiency across the FGD is required, are more reliable. The measured SO2 because SO capture is generally not equilibrated. Measured Hg vapor speciation across 2 the FGD is required to evaluate the predictions. Our evaluation database is arbitrarily divided into two groups, one reported in open literature and one assembled from EPA’s ICR database. In the literature database, six 1datasets were collected by the CONSOL testing team and the last one was collected by 3UND EERC. All Hg determinations were based on the OntarioHydro (OH) protocol. The dataset labeled as OH-LNB was taken on the system labeled as CNS-P6 after a retrofit with low-NO burners about a year after the original test series. All fuels were hv X bituminous coals with sulfur contents representing the highest levels in the American utility industry. The Cl-contents ranged from 0.04 to 0.17 wt. % and tended toward the upper end of the typical range for hv bituminous coals. Hg-contents varied from 0.09 to 0.22 ppm. The set of FGDs represents a fairly broad sample of Ca-based systems, including limestone systems with natural and forced oxidation, as well as three Mg/lime systems. Only the oxidation mode for the EERC1 system was unspecified. The pH values were typical, with higher values for the Mg/lime systems. Liquid-to-gas ratios ranged from 30 to 94, and the SO capture efficiencies (,) were well over 90 %, except for CNS-P2, 2SO2 CNS-P3, and CNS-P5. The Hg determinations also represent a broad range, with total inlet Hg concentrations from 6.5 to 17.9 ,g/dscm, and oxidized fractions (f) from Hg(II) 0.57 to 0.87. The Hg retention in the FGDs varied accordingly. Total Hg retentions (f) ranged from 38 to 83 %, while the retentions of Hg(II) (,) ranged from 78 to SCRBHgCl2 97 %. Our primary goal is to quantitatively interpret these variations. EPA’s ICR database contains 16 cases with Hg retentions across wet FGDs, but only ten were qualified into the evaluation database. One was omitted because the FGD used recycled soda ash liquor; two for omission of critical fuel properties; one for inconsistent furnace operating conditions; and two for badly scattered and erratic Hg retention data. All the accepted datasets used limestone slurry in spray, packed or SHU towers. All the Hg data were obtained with OH. The inlet Hg vapor concentrations ranged from 1.1 to 11.6 ,g/dscm, with 7.5 to 89 % Hg(II). Total Hg retentions ranged from -2.6 to 83 %, and the Hg(II) retentions ranged from 40 to 98 %. here are two main limitations in the ICR data: (1) The incomplete ultimate analyses are T insufficient to calculate the complete flue gas composition; and (2) The SO capture 2 efficiencies and FGD operating temperatures were not reported. Whereas the second omission is a source of significant uncertainty, the first was alleviated by estimating C/H/O/N compositions from the calorific values with a normative analysis, as explained 4,5elsewhere. Subbituminous coals were burnt at three stations, and a lignite was burned at one station. Otherwise, all coals were hv bituminous. S-contents ranged from 0.75 to 4.1 daf wt. %; Cl-contents ranged from 90 to 2270 ppm; and Hg-contents ranged from 24 to 908 ppb. PREDICTING HG RETENTION IN WET FGDS 3 In the wake of EPA’s ICR database, several engineering correlations were reported to relate Hg retention in wet FGD units to the limited selection of coal properties and 6-8 No predominant factor or scrubber operating conditions included in that database. basis set is evident in these studies, and all the regressions for wet FGDs had correlation coefficients well below one-half. A potentially more fruitful approach based on equilibrium analyses of Hg species under wet scrubbing conditions demonstrated fair 9-11consistency with data from three full-scale scrubbers, but did not identify the primary chemical connections. 4,5Our preliminary regressions established direct connections among flue gas compositions, the extents of Hg oxidation at FGD inlets, and Hg retention efficiencies in FGDs. We regard them as clear signals that solution chemistry within the FGD determines Hg retention, and are therefore compelled to characterize this chemistry in greater detail. Whereas finite-rate kinetics are largely unknown for this reaction system, 9-11the reported thermochemical equilibrium analyses provide a firm foundation for quantitative applications, particularly regarding activity coefficients for the most relevant species. We first introduce a concise definition of the reaction system that retains all essential connections, yet avoids unnecessary adjustable parameters, then stage evaluations with both databases. Current thinking on SO absorption in FGDs is that several mass transport limitations 2 come into play, and that a variety of processing conditions must be specified to 12,13accurately predict the SO capture efficiency. We circumvent these complications by 2 regarding the SO capture efficiency as an input variable. This approach has the 2 important benefit of ensuring an accurate estimate for the incorporation of SO into the 2 scrubber solution, regardless of the accuracy of the scrubber model and without an undue burden on the input specifications. We also retain the assumption that only Hg(II) enters 0the scrubber solution, based on the very poor solubility of Hg. The equilibrium analysis requires the following input data: (1) Whether the FGD uses limestone slurry or Mg/lime. (2) The inlet gas composition as mass fractions for N, CO, O, SO, NO, CO, and 2222 HCl. The O and NO concentrations should be measured. 2 (3) The moisture content of the flue gas. (4) The total Hg concentration, in ,g/dcsm, and the fraction of Hg(II) at the FGD inlet, based on measured values. (5) The SO capture efficiency. 2 (6) The FGD operating temperature. Only the flowrates of Hg(II) and of the absorbed portion of the inlet SO are included in 2 the equilibrium calculations. The slurry phase is specified from the molar ratio of Ca to S, which was fixed at 1.05 for all cases. Gypsum is assumed to be the only solid product of SO absorption. For Mg/lime systems, the molar flow of MgSO was 5 % of the molar 23 flow of lime; this specification was shown to be inconsequential. 4 Equilibrium compositions were evaluated by minimizing the Gibbs free energy of the system with v.5.11 of HSC Chemistry for Windows (Outokumpu Research Oy, Pori, Finland), based on 9 vapor species, 44 aqueous species, and 7 independent solids. The vapor consists of only the standard flue gas species. The slurry or aqueous phase consists of dissolved species, cations, and anions representing the major intermediate compounds in FGD chemistry. Non-unity activity coefficients for the aqueous species were reported 9 based on Pitzer/Debye/Huckel calculations for very similar operating by Luckas et al., conditions. The shifts for the different compositions imposed in our calculations are expected to be minor compared to the deviations from unity in the reference values. The seven pure solids do not constitute a mixture but, rather, a collection of independent pure compounds. EVALUATIONS Separate evaluations are staged for the literature and ICR databases, primarily because the omission of , from the ICR database significantly erodes the stringency. However, SO2 there are no adjustable parameters whatsoever in the evaluation of the literature database because all the necessary input data was specified in the test reports. The only exception is , for case EERC1, which was estimated as 95 %. All equilibria were evaluated at SO2 50:C. The equilibrium gas composition includes a nonzero amount of HgCl. This value 2IN, as 100(C - is used to assign the HgCl retention efficiency, ,HgCl2HgCl22EQIN4,5C )/C. The predicted total Hg retention is evaluated from ,. HgCl2HgCl2HgCl2 The predicted values for , and f are collected in Table 1 for the literature HgCl2SCRB database, along with the computed pH values. The predicted HgCl retention efficiencies 2 cover a broad range from 70 to 97 %, which is entirely responsible for the remarkable accuracy of the predicted total Hg retentions. The predictions match the retentions based on the measured Hg concentrations to within 5 %, even though the measured values span a range from 0.38 to 0.82. They are uniformly accurate across the entire range of f SCRB values. The standard deviation is only 0.033, and the correlation coefficient is 0.970. The estimated pH values are systematically lower by about 0.5 than the reported values 1for all cases, but otherwise consistent with the tendency for higher pH values in Mg/lime systems than in the limestone FGDs. These discrepancies may be due to uncertainties in the FGD temperatures (cf. Fig. 3, below), or to the comparison of a single nominal pH value for the entire system to a single-point measurement. The performance is less satisfactory with the ICR database, albeit with encouraging aspects. To compensate for the omission of ,, all equilibrium calculations for ICR SO2 cases were based on an assumed efficiency of 95 %. The predicted pH values are again reasonable for limestone slurries, except for two very low values for stations firing low-rank coals. The predicted values of , range from 37 to 98 % which, in turn, yield a HgCl2 range of f from 4 to 74 %. As seen in the parity plot in Fig. 1, the predicted Hg SCRB retentions are systematically low. They were correlated to the measured values by the dashed regression line through the origin with a slope of 0.786 in Fig. 1. But the correlation coefficient is only 0.736, and the std. dev. is 0.156. These values are 4,5comparable to the statistics for the flue-gas-based correlations. 5 Table 1. Predictions for the Literature Database. pH f, , SCRBSO2HgCl2 Pred. Mes’d. Literature Database CNS-P1 97 5.95 70.1 52 56 Predicted fCNS-P2 82 5.27 85.6 68 68 CNS-P3 87 5.45 96.7 85 82 CNS-P5 82 5.36 89.4 63 60 CNS-P6 96 5.98 76.0 54 59 OH-LNB 96 6.05 73.0 64 62 EERC1 95 6.39 69.8 40 38 Figure 1: Parity plot for total Hg retention in the ICR database with the dashed linear regression line for the entire database. Cases with FGDs commissioned since 1992 appear as (,). ICR Database80 60 ,% SCRB 40 20 0 020406080 Measured f,%SCRB 6 Notwithstanding, most of these discrepancies should be attributed to omissions in the input data, rather than inadequacies of the equilibrium analysis. The main omission is ,. Since the ICR database represents FGDs commissioned from 1979 through 1999, SO2 the assumed value is almost certainly too high for the older units. This explanation is corroborated by the much better evaluation with the FGDs commissioned since 1992, as seen in Fig. 1. Whereas the predictions for older units comprise all the cases with the largest discrepancies, all those for newer units are closer to the parity line. The regression line for these cases passes through the origin with a slope of 0.847, and has a correlation coefficient of 0.862 and a std. dev. of 0.101. This represents a marked improvement over the regressions with only flue gas compositions. 0RE-EMISSION OF HG 0The available database conveys a fairly weak indication that Hg may be re-emitted from scrubbing solutions under some, but not all, operating conditions. For most cases in the 4,5ICR database, this difference is within ,5 %, which is within the measurement uncertainties. Notwithstanding the minor impact, additional equilibrium calculations were performed to identify the equilibrium conditions which would generate substantial 0amounts of Hg. All conditions were specified from case CNS-P2. The only variation 0which generated Hg was the inlet O concentration, and only for unrealistically low O 22 levels. There is an abrupt transition from dissolved HgCl and small amounts of HgCl 220. This vapor to essentially complete re-entrainment of all the original HgCl as Hg22-transition is accompanied by sharp reductions in the amounts of SO(aq) and 42--CaSO,2HO, and surges in the levels of aqueous SO, HSO, and SO. Such an 423320association among significant amounts of sulfite species and Hg vapor was observed in lab-scale characterization of mercury retention in wet FGDs, but for Na-based scrubbing 14solutions. Our calculations for limestone-based systems identify analogous behavior, but only for O levels that are much lower than the practical operating range. 2 PARAMETRIC SENSITIVITY Additional equilibrium calculations based on case CNS-P2 characterized variable ,, T, SO2 inlet concentrations of HCl, O, NO, HgCl, Ca/S levels, and slurry loadings. The 22 variations in ,, NO, HgCl, and Ca/S were inconsequential. The HgCl retentions SO222 varied from 72 to 97 % while pH stayed essentially uniform as the inlet HCl concentration was tripled to about 300 ppm. Even for 50 % increases in the HCl level, , ranged from 72.5 to 90 % which, in turn, changed f from 58 to 72 % for this HgCl2SCRB particular case. HCl variations exert the strongest impact identified so far, so the familiar impact of coal quality on extents of Hg oxidation will also be evident in Hg retention in FGDs. Flue gases from subbituminous coals have low HCl concentrations. So wet FGDs will retain less Hg from flue gas from subbituminous coals than from typical hv bituminous coals, even when the inlet levels of Hg(II) are the same. For example, the equilibrium analysis predicts , of 66 % and 86 % when typical subbituminous and hv bituminous coals HgCl2 were processed through the same FGD with the same inlet f values and , values. Hg(II)SO2 7 Another implication of the HCl dependence is that FGDs with prescrubbers to remove acids can be expected to retain less Hg than more modern, integrated designs. This effect alone may be mostly responsible for the poor predictions for older FGDs in the ICR database in Fig. 1. and pH. This effect is approximately as strong as Oxygen addition decreases both ,HgCl2 HCl additions, reducing , from 100 to 83 % for O additions up to twice the baseline HgCl22 level. Based on these results, one might have expected that HgCl retention depends on 2 the FGD oxidation mode, but no such influence was apparent in the evaluations with the literature dataset. Even so, the inlet O concentration into the FGD must be specified for 2 accurate equilibrium calculations, preferably with measured flue gas compositions. Higher FGD operating temperatures decrease , and increase pH. The temperature HgCl2 dependence of both variables is strong, so that heating the FGD by only 20:C reduces , from essentially 100 % to 70 %. Note also that the discrepancies of 0.5 in the HgCl2 predicted pH values in Table 1 are associated with temperature variations of only 5-7:C, so the assumed uniform temperature of 50:C in the calculations could be entirely responsible for these discrepancies. SUMMARY It is widely recognized that wet FGDs are effective at retaining oxidized Hg vapor species, and at significantly reducing smokestack Hg emissions. What was much less well established is the quantitative performance of FGDs of various configurations with a wide range of fuel quality and operating conditions. Fortunately, the simplest possible description of this chemical reaction system, based on thermochemical equilibrium, yields highly accurate predictions for total Hg retention. With the literature database, the predictions were as accurate as they possibly could be for both limestone and Mg/lime systems operating in both forced and natural oxidation mode. With the ICR database, the quantitative performance was almost as good for the most modern FGDs, which almost certainly conform to the very high SO absorption efficiencies assumed in the 2 calculations. The large discrepancies for older FGDs are tentatively attributed to the unspecified , and operating temperatures and to the possible elimination of HCl in SO2 prescrubbers. The equilibrium calculations suggest that Hg retention is most sensitive to inlet HCl and O levels and FGD temperature; weakly dependent on ,; and insensitive 2SO2 to HgCl, NO, Ca/S ratio, and slurry dilution level in limestone FGDs, and to MgSO 23 levels in Mg/lime systems. Consequently, FGDs will retain less Hg from flue gases derived from subbituminous coals than from typical hv bituminous coals, even when the levels of Hg(II) at the FGD inlet are the same. Also, systems with prescrubbers to eliminate HCl probably retain less Hg than fully integrated FGDs. The analysis also 0predicts re-emission of Hg, but only for inlet O levels that are much lower than those in 2 practical applications. The predictions require only a proximate and ultimate analysis, expanded with Cl- and Hg-contents; a furnace stoichiometry or, preferably, measured O, NO, HCl, SO, and 22 moisture levels; the total inlet Hg concentration and the fraction of Hg(II); the FGD type 8 absorption (limestone or Mg/lime) and operating temperature; and the actual SO2 efficiency. ACKNOWLEDGEMENT This study was sponsored by the New Energy Development Organization’s (NEDO) Clean Coal Technology Center under the Toxic Metals Project, and administered by the Coal Research Laboratory, Idemitsu Kosan Company, Ltd. REFERENCES 1. DeVito, M. S.; Withum, J. A.; Statmick, R. M. “Flue gas measurements from coal- fired boilers equipped with wet scrubbers,” Int. J. of Environment and Pollution 2002 17(1/2), 126-142. 2. Evans, A. P.; Holmes, M. J; Redinger, K. E. Phase II Final Report, Rev. 1, Advanced Emissions Control Development Program, US DoE No. DE-FC22-94PC94251, McDermott Technology Inc. 1998. 3. Laudal, D. L., “JV Task 10 – Characterization and modeling of the forms of mercury from coal-fired power plants,” Final Report, US DoE No. DE-FC26-98FT40321, Univ. North Dakota EERC 2001. 4. Niksa, S.; Fujiwara, N., “The impact of wet FGD scrubbing on Hg emissions from “ Proc. U. S. EPA-DOE-EPRI Combined Power Plant Air coal-fired power stations , Pollutant Control Symp.: The Mega Symp., Paper No. 44, 2004, Washington, DC, Aug. 29-Sep. 1. 5. Niksa, S.; Fujiwara, N., “The impact of wet FGD scrubbing on Hg emissions from coal-fired power stations ,“ A&WMA J., to appear (2005). 6. Weilert, C. V.; Randall, D. W., “Analysis of ICR data for mercury removal from wet and dry FGD,” Proc. U. S. EPA-DOE-EPRI Combined Power Plant Air Pollutant Control Symp.: The Mega Symp. and AWMA Specialty Conf. on Mercury Emissions: Fate, Effects, and Control, 2001, Chicago, IL, Aug. 21-23. 7. Chu, P.; Behrens, G.; Laudal, D., “Estimating total and speciated mercury emissions from U. S. coal-fired power plants,” Proc. U. S. EPA-DOE-EPRI Combined Power Plant Air Pollutant Control Symp.: The Mega Symp. and AWMA Specialty Conf. on Mercury Emissions: Fate, Effects, and Control, 2001, Chicago, IL, Aug. 21-23. 8. Afonso, R.; Senior, C., “Assessment of mercury removal by existing air pollution control devices in full-scale power plants,” Proc. U. S. EPA-DOE-EPRI Combined Power Plant Air Pollutant Control Symp.: The Mega Symp. and AWMA Specialty Conf. on Mercury Emissions: Fate, Effects, and Control, 2001, Chicago, IL, Aug. 21-23. 9. Luckas, M.; Lucas, K.; Roth, H. “Computation of phase and chemical equilibria in flue-gas/water systems,” AIChE J. 1994 40(11), 1892-1900. 10. Krissmann, J.; Siddiqi, M. A.; Peters-Gerth, P.; Ripke, M.; Lucas, K. “A study of the thermodynamic behavior of mercury in a wet flue gas cleaning process,” Ind Eng. Chem. Res. 1998 37(7), 3288-94. 11. Sandelin, K.; Backman, R. “Trace elements in two pulverized coal-fired power stations,” Environ. Sci. Technol. 2001 35, 826-34. 9 absorption into limestone 12. Lancia, A.; Musmarra, D.; Pepe, F. “Modeling of SO2 suspensions,” Ind. Eng. Chem. Res. 1997 36(1), 197-203. 13. Kiil, S.; Michelsen, M. L.; Dam-Johanses, K. “Experimental investigation and modeling of a wet flue gas desulfurization plant,” Ind Eng. Chem. Res. 1998 37, 2792-2806. 14. Chang, J.; Ghorishi, S. B. “Why does flue gas elemental mercury concentration increase across a wet scrubber ?” Proc. U. S. EPA-DoE-EPRI Combined Power Plant Air Pollutant Control Symp.: The MEGA Symp., 2003, Washington, DC, May 19-22. KEY WORDS Mercury Emissions Mercury Speciation Mercury Retention Wet FGD Emissions Prediction 10