The innovative multi-pollutant control system installed at the AES Greenidge unit 4 coal plant started up in March 2007. What lessons can we draw from this project, which aims to demonstrate an economic emissions reduction retrofit solution for older power plants of small unit size?
AES Greenidge unit 4 (Figure 1) is a 107 MWe, 1953 vintage, tangentially-fired, balanced draft, reheat unit that fires pulverised eastern US bituminous coal as its primary fuel and can co-fire biomass (waste wood) at up to 10% of its heat input. It is representative of the 420 coal fired units with installed capacities in the range 50-300 MWe currently operating in the USA without SCR, FGD, or mercury control. These smaller units are a valuable part of the USA’s power generation infrastructure, contributing about 60 GWe in total. However, with ever tightening emissions requirements (eg CAIR, CAVR, possible mercury MACT standards and state regulations), pollution control retrofits will be required in many cases to allow these units to keep operating.
Unfortunately, the large capital and space requirements of conventional emissions control technologies – such as SCR and wet FGD – often make these options unattractive for the smaller unit, and a multi-pollutant approach (ie installation of a combination of technologies in an integrated and innovative way) is more appropriate. The Greenidge Multi-Pollutant Control Project – being carried out by a team consisting of CONSOL Energy Inc Research & Development, prime contractor, AES Greenidge LLC, owner, and Babcock Power Environmental Inc (BPEI), EPC contractor, seeks to demonstrate this, with funding from DoE and AES Greenidge.
The overall goal of the project is to show that the multi-pollutant control system being demonstrated, which had a capital cost of less than $350/kW and occupies less than 0.5 acre, can achieve full-load NOx emissions of = 0.10 lb/mmBtu, reduce SO2 and acid gas (SO3, HCl, and HF) emissions by = 95%, and reduce Hg emissions by = 90%, while the unit is firing 2-4% sulphur eastern US bituminous coal and co-firing up to 10% biomass. The system is designed to be applicable to a wide variety of coals, including high-sulphur, as over 80% of the 420 candidate units already mentioned are east of the Mississippi River. It also aims for maintenance requirements compatible with staffing levels at smaller plants, and for good turndown capabilities to cater to units that routinely cycle load in response to power demand.
A schematic of the multi-pollutant control process installed at AES Greenidge 4 is shown in Figure 2. It includes combustion modifications, a NOxOUT CASCADE® hybrid SNCR/SCR system (supplied by Fuel Tech), and a Turbosorp® circulating fluidised bed dry scrubbing system (supplied by BPEI and Austrian Energy and Environment), with baghouse ash recycling and activated carbon injection. The system design for Greenidge assumes 2.9%-sulphur bituminous coal and a baseline full-load NOx emission rate of ~0.30 lb/mmBtu prior to the installation of the new combustion modifications. Emissions control equipment before the project consisted of a separated overfire air (SOFA) system for NOx control and an ESP for particulate control, while the permitted SO2 emission rate of 3.8 lb/mmBtu was met by restricting the sulphur content of fuel.
NOx control is now accomplished using urea-based, in-furnace SNCR followed by a single-bed SCR reactor that is installed in a modified section of the ductwork (Figure 3) between the unit’s economiser and its two air heaters. The SCR process is fed exclusively by ammonia slip from the SNCR process. Static mixers located just upstream of the SCR are used to homogenise the velocity, temperature, and composition of the flue gas to promote optimal ammonia utilisation and NOx reduction across the relatively small SCR catalyst, which consists of a single layer that is ~1.3 m deep. Because the SCR reactor is able to consume ammonia slip (typically a limiting factor in SNCR design), the upstream SNCR system can operate at lower temperatures than a stand-alone SNCR system would, resulting in improved urea utilisation and greater NOx removal by the SNCR system, as well as sufficient NH3 slip to permit additional NOx reduction via SCR.
Because of its compact reactor design, the hybrid SNCR/SCR system avoids many of the capital costs associated with a conventional stand-alone SCR system.
The NOx control system at AES Greenidge unit 4 also includes combustion modifications (low-NOx burners and SOFA) to achieve further reductions in NOx emissions and to improve the performance of the hybrid SNCR/SCR system. Hence, the system is designed to achieve the full-load NOx emission rate of = 0.10 lb/mmBtu by combining the combustion modifications, which are designed to produce NOx emissions of 0.25 lb/mmBtu, the SNCR, which is designed to reduce NOx by ~42% to 0.144 lb/mmBtu, and the SCR, which is designed to further reduce NOx by = 30%.
The SNCR system at AES Greenidge includes three zones of urea injection. At high generator loads, urea is injected into the mid- and low-temperature zones to maximise NOx removal and generate ammonia slip for the SCR reactor. At generator loads that produce economiser outlet temperatures below the minimum operating temperature for the SCR reactor, urea injection into the lowest-temperature zone is discontinued; however, urea continues to be injected into one or both of the mid- and high-temperature zones until the minimum SNCR operating temperature is reached, resulting in continued NOx removal (20-25%) via SNCR. Below the minimum SNCR operating temperature, NOx emissions continue to be controlled by the unit’s low-NOx combustion system.
Emissions of SO2 and other acid gases are reduced by = 95% in the Turbosorp® circulating fluidised bed dry scrubber system (Figure 4), which is installed downstream of the air heaters. In the Turbosorp® system, water and dry hydrated lime (Ca(OH)2), which is produced from pebble lime in an onsite hydrator installed as part of the project, are injected separately into a fluidised bed absorber. There, the flue gas is evaporatively cooled to within 45°F of its adiabatic saturation temperature and brought into intimate contact with the hydrated lime reagent in a fast fluidised bed.
The basic hydrated lime reacts with the acidic constituents of the flue gas (SO2, SO3, HCl, and HF) to form dry solid products (CaSO3, CaSO4, CaCl2, CaF2), which are separated from the flue gas in a new eight-compartment pulse jet baghouse. More than 95% of the collected solids are recycled to the absorber via air slides to maximise pollutant removal and lime utilisation.
A flue gas recycle system is also included to provide sufficient flue gas flow to maintain a fluidised bed in the absorber at low-load operation. A new booster fan, which was installed upstream of the unit’s existing induced-draft fans to overcome the pressure drop created by the installation of the in-duct SCR, fluidised bed absorber, and baghouse, provides the motive force for the flue gas recycle. The booster fan accounts for a majority of the multi-pollutant control system’s parasitic power requirement, which totals about 1.6% of the net electric output of the unit.
Because water and dry hydrated lime are injected separately into the Turbosorp® absorber vessel, the Ca(OH)2 injection rate is controlled solely by the SO2 loading in the flue gas and by the desired SO2 emission reduction, without being limited by the flue gas temperature or moisture content. As a result, the Turbosorp® system affords greater flexibility than a spray dryer for achieving deep emission reductions from a wide range of fuels, including high-sulphur coals. Moreover, the high solids recycle rate from the baghouse to the absorber vessel promotes efficient sorbent utilisation in the Turbosorp® system. Unlike wet FGD systems and spray dryers, the Turbosorp® system does not require slurry handling. This is expected to result in reduced maintenance requirements relative to the alternative.
In addition, unlike wet FGD, the Turbosorp® system does not produce a saturated flue gas, and therefore it is constructed from carbon steel and does not entail installation of a new corrosion-resistant stack. These factors, coupled with the mechanical simplicity of the Turbosorp® system relative to a wet FGD system, contribute to its comparatively lower capital costs.
The arrangement of the circulating fluidised bed dry scrubber, baghouse, and associated equipment is compact. As demonstrated in Figure 4, the various pieces of equipment are vertically tiered to permit gravity-assisted transport of solids where possible.
The combination of combustion modifications, in-duct SCR, circulating fluidised bed dry scrubber, and baghouse achieves high mercury removal, even without mercury specific measures. This results from factors such as the conversion of elemental mercury (Hg0) to oxidised mercury (Hg2+) across the SCR catalyst, the removal of Hg2+ and SO3 (which can interfere with Hg adsorption on carbon particles) by moistened, basic Ca(OH)2 particles in the scrubber, and the removal of Hg2+ and possibly some Hg0 via adsorption onto carbon-containing fly ash and Ca(OH)2 at low temperatures in the baghouse, which facilitates contact between gaseous mercury and carbon or other sorbent contained in the dust cake that accumulates on its numerous filter bags. The combustion modifications also contribute to Hg removal by increasing the unburned carbon content of the fly ash, thereby improving its capacity for Hg capture.
Nevertheless, the Greenidge multi-pollutant control system also includes a mercury-specific activated carbon injection (ACI) system upstream of the Turbosorp® absorber vessel.
Economics and O&M
Table 1 summarises the estimated economic performance of the AES Greenidge multi-pollutant control system. The total EPC capital cost (excluding the ACI system, but including all other ancillary equipment) of $343/kW is about 40% less than the estimated cost would have been to retrofit the unit with conventional SCR and wet FGD systems. Costs for the ACI system are not shown in Table 1 because testing has shown that the ACI system is not needed to achieve the project’s Hg removal target. If included, the ACI system would add about $6/kW to the EPC capital cost. The total levelised costs for NOx and SO2, which include levelised capital and fixed and variable O&M costs, also cover control of mercury, acid gases, and particulates. These latter pollutants are removed as a co-benefit of the NOx and/or SO2 control systems at no incremental cost. The costs for urea in the NOx control system and for lime and waste disposal in the Turbosorp® system, which are the costs that figure in the unit’s dispatch calculations, are $626/ton of NOx and $241/ton of SO2, respectively. So, installation of the multi-pollutant control system has improved the unit’s dispatch economics relative to purchasing allowances.
The greatest problem encountered during the first year of operation of the multi-pollutant control system has been the accumulation of large particle ash (LPA) in the in-duct SCR catalyst, which began to occur soon after start-up. The project team worked throughout much of the first year to mitigate this problem.
The LPA, which consists of pieces of slag that in many cases are too large to pass through the honeycomb catalyst, becomes lodged in the catalyst channels and promotes subsequent accumulation and bridging of fly ash, eventually plugging a substantial portion of the catalyst (Figure 5a). This causes an increase in the pressure drop across the SCR reactor. At AES Greenidge, the pressure drop becomes sufficient to pose an implosion risk for downstream ductwork and the unit must be derated and/or taken offline for catalyst cleaning. Numerous outages were required for this purpose during the first year of commercial operation of the multi-pollutant control system. LPA accumulation in the SCR catalyst can also contribute to decreased NOx removal efficiency, increased ammonia slip, and increased catalyst erosion.
In conventional SCR installations, LPA can often be dealt with by installing a screen and hopper at a 90° bend in the ductwork upstream of the reactor. However, at AES Greenidge, the flue gas flow between the economiser and reactor is vertically downward, with no available 90° bends or hoppers. Therefore a solution has been developed consisting of a sloped screen in the ductwork between the economiser and the catalyst, intersecting the static mixers (Figure 5b). Eight vacuum ports were installed at the base of the screen to remove the collected LPA; soot blowers are located beneath the screen to help transport the LPA to the vacuum ports. The screen, vacuum ports, and two soot blowers were originally installed in May 2007. In September 2007, the two soot blowers were replaced with four rotary soot blowers, and a spring seal was installed to close the gap between the screen’s two sections, which bridge an expansion joint. A rake soot blower was also installed above the SCR catalyst to aid in resuspending accumulated fly ash. In late 2007, patches were installed to eliminate openings in several areas of the screen. These improvements have increased the time between derates and/or catalyst cleaning outages from less than one month to about three months. However, LPA particles that are large enough to plug the catalyst are still passing the screen. A smaller-pitch screen will be installed in May 2008; it is expected that this will significantly reduce the severity of the problem.
Another problem – much less troublesome than the LPA – is that ammonia slip from the hybrid SNCR/SCR has been greater than expected, particularly when there is significant LPA accumulation in the SCR reactor. Thus far, the higher-than-expected ammonia slip has not significantly affected unit operability, but it will continue to be monitored.
Most of the O&M requirements associated with the Turbosorp® system have involved the lime hydration system, which is the most mechanically complex part of the process. The most common problem is plugging in the hydrated lime classification system. Adjustments have been made to reduce the accumulation of fines in the classifier, helping to lessen the severity of the problem. The plant has also encountered some problems with balls escaping from the ball mill and causing damage elsewhere in the system, as well as with freezing of lines and valves during periods of cold weather. On one occasion, the bucket elevator shaft failed, and on another, water was overfed to the hydrator. These problems, however, have usually been resolved without impacting the operation of the Turbosorp® scrubber. Plant personnel can continue to operate the scrubber while the hydrator is offline by using hydrated lime from their onsite inventory or by taking deliveries of hydrated lime. The onsite storage capacity for hydrated lime is being increased.
The major byproduct from the multi-pollutant control system is the product ash from the Turbosorp® system. This product ash is similar to spray dryer ash in that it is a mixture of CaSO3 and CaSO4 (including hydrates), fly ash, CaCO3, Ca(OH)2, CaO, CaCl2, CaF2, and inerts. It is a dry, free-flowing powder that contains about 1% moisture. For every ton of SO2 removed, about 3.2 tons of scrubber byproduct (excluding fly ash) are produced. This byproduct is currently landfilled but the project team is exploring potential reuses, which could include use in mine reclamation, use as a structural or flowable fill, or use in manufactured aggregate production. The product ash contains the Hg captured by the multi-pollutant control process. Mercury leaching tests have been performed on seven product ash samples and the amount of Hg in the leachate in all cases was <0.35 µg/l (the detection limit), which equates to <1.5% of the total Hg in the ash.
Emissions reduction performance
Guarantee testing of the multi-pollutant control system at AES Greenidge unit 4 was conducted in March and May 2007, shortly after start-up and commissioning of the system were completed. During the guarantee test periods, unit 4 operated at or near full load and fired eastern US bituminous coals containing 2.4-3.2% sulphur. Table 2 summarises the results (HF concentrations were below the detection limit at both the inlet and outlet of the Turbosorp® system, preventing the calculation of a removal efficiency for HF).
Although the NOx emisions performance target of 0.10 lb/mmBtu was demonstrated during short term guarantee testing, the plant has been unable to achieve this emission rate in the long term while also maintaining acceptable combustion characteristics, sufficiently high steam temperatures, and sufficiently low ammonia slip for routine operation. During the guarantee test period, the unit experienced flame attachments that damaged several burners, forcing plant personnel to reduce the aggressiveness of low-NOx firing. The NOx control problems have been exacerbated by the accumulation of LPA in the in-duct SCR reactor, as already noted, which contributes to decreased NOx removal efficiency and increased ammonia slip from the reactor.
The unit has thus generally operated with high-load NOx emissions of 0.10-0.15 lb/mmBtu since the guarantee testing period. Its permitted NOx emission rate is 0.15 lb/mmBtu for gross generator loads above 68 MW. The permitted emission rate increases to 0.28 lb/mmBtu when the gross generator load is between 53 and 68 MW and to 0.35 lb/mmBtu when the gross generator load is between 43 and 52 MW, consistent with the turndown strategy for the hybrid NOx control system.
Figure 6 shows average NOx emissions as a function of gross generator load during January 2008. A freshly cleaned catalyst layer was installed in the in-duct SCR reactor in late December 2007; hence, NOx emissions during January were minimally influenced by LPA in the catalyst. The overall average NOx emission rate (weighted by heat input) during the month was 0.15 lb/mmBtu, and the average NOx emission rate for gross generator loads above 68 MW was 0.14 lb/mmBtu. This NOx emission profile is typical of that observed at AES Greenidge Unit 4 during the first year of operation of the multi-pollutant control system.
During its first year of operation, the Turbosorp® system has consistently met or exceeded its performance target of = 95% SO2 removal. Figure 7 shows hourly measurements from January 2008.
On average, the Turbosorp® system reduced SO2 emissions from 3.69 lb/mmBtu to 0.14 lb/mmBtu during the month, resulting in a removal efficiency of 96.2%. Hourly average inlet SO2 rates ranged from 2.86 to 4.52 lb/mmBtu in January, and removal efficiencies greater than 99% were observed during a number of one-hour periods. The Turbosorp® system has shown an ability to achieve deep SO2 emission reductions with high-sulphur coals, consistent the design objective noted earlier. Tests conducted in October 2007 demonstrated 96% SO2 capture when the unit was firing coal with a sulphur content of 4.7 lb SO2/mmBtu, which is substantially greater than typical coal sulphur specifications for dry scrubbers.
The mercury removal performance of the system has also been commendable. Figure 8 summarises the results of mercury removal efficiency tests carried out between March and November 2007.
All 19 Hg measurements surpassed the 90% target. Moreover, 14 of the 19 tests were conducted without any activated carbon injection, and all of these exhibited > 90% removal efficiency.
The average coal-to-stack Hg removal efficiency measured during the 15 full-load tests shown in Figure 8 is = 96%, a more-than 94% reduction over the baseline Hg removal efficiency of 30% measured at AES Greenidge unit 4 in November 2004, prior to the installation of the multi-pollutant control system.
Table 3 summarises the SO3, HCl, and HF concentrations and removal efficiencies that have been measured at AES Greenidge unit 4 since the installation of the multi-pollutant control system.
As shown in Table 2, the Turbosorp® system exceeded its performance target for SO3 removal efficiency (= 95%) during guarantee testing on 2 May 2007, when the average SO3 concentration at the inlet to the Turbosorp® system was 26.5 ppmvd @ 3% O2, and the average SO3 removal efficiency was 97%. Since then SO3 removal efficiencies have varied considerably, owing largely to variations in SO3 concentrations at the Turbosorp® inlet, as illustrated in Table 3. Overall, however, installation of the Turbosorp® system has reduced the unit’s stack SO3 emissions to less than 1 ppmvd.
The Turbosorp® system demonstrated attainment of its performance target for HCl removal efficiency (= 95%) throughout its first year of operation.
HF concentrations at the stack have been below the method detection limit for all 14 tests performed to date. For five of these tests, the HF concentration measured at the inlet to the Turbosorp® system was also below the detection limit. The average HF concentration at the Turbosorp® inlet during the remaining nine tests was 1.45 ppmvd @ 3% O2, which is less than half the amount predicted from the coal fluorine analysis, and the average HF removal efficiency during these tests was > 86.9%. The disparity between the inlet HF concentration and coal fluorine content has not been resolved.
Finally, the particulate matter emission rate measured during 22 full-load tests between 28 March 2007 and 11 October 2007 was < 0.001 lb/mmBtu, which represents a more-than 98% reduction over the baseline full-load particulate matter emission rate of 0.063 lb/mmBtu measured in November 2004. The improvement in PM emissions has occurred in spite of the substantial increase in flue gas particulate loading brought about by the hydrated lime, reaction products, and high solid recycle rate in the Turbosorp® system. It results largely from the superior performance of the baghouse relative to the unit’s old ESP. In addition, the fluidised bed absorber is thought to promote agglomeration of fine particles, making them easier to capture in the downstream particulate collection equipment.