Senate Environment and Public Works Committee
Subcommittee on Clean Air, Climate Change and Nuclear Safety
Hearing on S. 485, the Clear Skies Act
June 5, 2003
Testimony of Larry S. Monroe
Program Manager of Pollution Control Research
My name is Larry S. Monroe and I am the Program Manager of Pollution Control Research for Southern Company. Southern Company is a super regional energy company serving customers in Alabama, Florida, Georgia, and Mississippi. Southern Company is the second largest user of coal in the utility industry with some 21,626 megawatts of coal-fired generating capacity. I hold a Ph.D. in Chemical Engineering from MIT, and have been involved in research on pollution control for coal-based power plants for over 20 years in university, not-for-profit research institute, and corporate settings. At Southern Company, I manage a research group that evaluates, develops, demonstrates, and troubleshoots technologies to control particulates, SO2, NOx, and hazardous air pollutants, including mercury, from fossil-fired power plants.
For the last 2 years, I have been engaged in the national effort to develop technologies to control mercury emissions from coal-fired power plants, resulting from EPA’s decision in December 2000 to develop Maximum Available Control Technology (MACT) mercury regulations for coal plants. I serve as the utility co-chairperson of the EPRI program tasked with developing and evaluating mercury control technologies. I have also directed Southern Company’s efforts, along with our partners including other utilities, EPRI, the Department of Energy, and the Environmental Protection Agency, in an attempt to develop cost-effective controls of utility mercury emissions.
I have been representing Southern Company and the industry on the Utility MACT Working Group, a subcommittee formed under the Clean Air Act Advisory Committee to provide advice to the Environmental Protection Agency. As a member of the MACT Working group, I have been intimately involved in the discussions with all of the stakeholders – including the environmental community, the state/local/tribal regulatory agencies, and the industry stakeholders – on the form of the regulation and its impacts on the industry and the price of electricity. As a part of this effort, I have been the leader of the industry stakeholders on advising EPA on our view of the performance and cost of the available mercury control technologies.
Working with EPRI, DOE, and EPA, Southern Company is one of the leading utilities in the national effort to develop mercury controls. We hosted the first full-scale power plant testing of mercury control ever performed in the United States, and are just starting a long-term follow-on test at the same site. Southern has also established a unique program to explore the fundamentals of mercury chemistry in coal power plant flue gas, partnering with EPA, TVA, EPRI, and several other utilities.
Today I am also testifying on behalf of the Edison Electric Institute (EEI). EEI is the association of U.S. shareholder-owned electric companies, international affiliates and industry associates worldwide. EEI’s U.S. members serve more than 90 percent of all customers served by the shareholder-owned segment of the industry, generate approximately three-quarters of all of the electricity generated by electric companies in the country, and serve about 70 percent of all ultimate customers in the nation.
State of Technology
The state of technology development for control of mercury emissions from coal-fired power plants is very much in its infancy. Some early efforts at measuring the mercury emissions from power plants were attempted in the mid-1990’s, but the sampling techniques used were not adequate, and much of that data is questionable. The mercury content in typical coal-fired power plant flue gas is very low, measured at the parts per trillion level. A good analogy that describes the low concentration of mercury in coal-fired power plant flue gas is to imagine a pipe, one foot in diameter, built from the earth to the moon. If this pipe, all 238,000 miles long, were to be filled with coal-fired power plant flue gas, and the mercury all magically brought to one end, it would only take up the first 18 inches of this pipe. If we compare the mercury in coal-fired power plant flue gas to the other criteria pollutants (e.g., particulates, NOx, and SO2) you find that the mercury is one million times less concentrated than those other species. The low concentrations of mercury, along with the propensity of mercury to react in the sampling equipment, contribute to the difficulties in accurately measuring and controlling mercury emissions at cost effective levels.
The state of knowledge of mercury chemistry and mercury emissions from power plants has been so scarce that, in 1999, the Environmental Protection Agency (EPA) required all power plants to sample their coal supply and test for mercury content, and required a selected number of power plants to sample for the different mercury species before and after the flue gas entered existing pollution control devices. Southern Company participated in that effort by tracking every coal to every one of our power plants and further by sampling two of our plants for mercury species and emissions. Unfortunately, this EPA Information Collection Request (ICR) database, while suffering from some flaws in data collection and power plant selection, remains the best publicly available database of mercury emissions, with and without controls, and of mercury chemistry for U.S. power plants.
There are currently no commercial technologies that are available for controlling mercury from coal-fired power plants. That is, there are no vendors that are offering process systems that are supported by guarantees from the vendor for mercury control performance under all the conditions that an ordinary power plant is expected to encounter over the course of normal operating conditions and timelines. Of course, there are vendors that will offer their best guess at how a particular technology will perform, but the risk of non-performance rests with the utility. The reliance on vendor warranties is standard practice within the utility industry, and the inability of the vendors to issue guarantees is indicative of the pre-commercial status of all mercury control technologies.
The most promising two technologies for mercury control in power plants are co-control by flue gas desulphurization (FGD) processes and the use of activated carbon injection (ACI) processes. To understand the co-control of mercury by FGD processes and the possibility of increased mercury control by NOx control processes, namely selective catalytic reduction (SCR) systems, a basic understanding of mercury chemistry is needed. First, coal is no different than any other solid material dug from the earth’s crust when it comes to the mercury content. In other words, coal is not enriched in mercury compared to ordinary rocks. The mercury in coal is there mainly as a sulfide compound, at a concentration that averages 50 parts per billion by weight. These sulfur-mercury compounds are the most common form of mercury found in nature and they tend to be very stable solids, only dissolved by a mixture of strong acids. Most everyone is familiar with mercury, the metal that is a liquid at room temperature and used widely in thermometers and blood pressure instruments seen in a physician’s office.
It is not a surprise that a metal that is liquid at room temperature would boil at much lower temperatures than ordinary metals, and mercury boils at only 674°F. Similarly, when coal burns in a utility boiler, mercury in the coal vaporizes and produces the vapor of the metal in the high temperature zones of the flame. This form of mercury is commonly referred to as elemental mercury, meaning that it exists in a form that is not combined with any other element. It is also known as “mercury zero,” a reference to the chemist’s shorthand of referring to the electron state of a pure element as zero, or Hg0.
As the temperature of the coal flue gas is cooled by the process of making and superheating steam, the elemental mercury vapor can react with other elements to form compounds. Our best knowledge of mercury chemistry suggests that mercury vapor can react with either chlorine or oxygen to produce mercury chloride (HgCl2) or mercury oxide (HgO). Since the electronic state of the mercury atom is now “plus two,” this form is sometimes called “mercury two,” ionic mercury, or oxidized mercury. These are all equivalent terms that describe the chemical state of the mercury. Finally, either of these two forms of mercury, the elemental or the ionic, can attach to solid particles, either fly ash or partially burned coal particles, and is typically referred to as “particulate mercury,” which is a physical description of the mercury form. To summarize, we generally classify the mercury in coal flue gas as being one of three forms: elemental, ionic, or particulate.
The proportions of the three chemical forms of mercury have a great influence over the behavior of the mercury in the flue gas in pollution control processes. The particulate form of mercury is the easiest form to remove, with high efficiency capture being normal along with the coal ash in electrostatic precipitators (ESPs) or bag houses. Unfortunately, in most power plants, the fraction of mercury contained in the particulate form is only a minor amount of the total mercury.
Flue Gas Desulphurization (FGD)
The most common method to remove sulfur dioxide (SO2) from coal-fired power plant flue gas is a wet scrubber. This device is a large tower, where the flue gas enters the tower near the bottom and flows upward, exiting through the top. When the flue gas is flowing, hundreds of nozzles spray a mixture of powdered limestone and water. The flue gas essentially flows up through a rain storm of these limestone-water droplets. Since SO2 is an acid, it reacts with the alkaline limestone solids and is neutralized.
The acid and base chemistry is so fast that the performance of the wet scrubber is dependent on the mixing between the flue gas and the droplets. Therefore, it is necessary to use multiple, large pumps and a large number of nozzles to produce the small droplets needed. The combined limestone-SO2 product from the scrubber is typically calcium sulfate, better known as gypsum – the white powder found inside wallboard (also called sheetrock). Gypsum is a naturally-occurring compound, mined both for fertilizer and wallboard.
In this common FGD process, the wet limestone scrubber, the form of the mercury in the flue gas entering the scrubber appears to be the most important factor in the efficiency of mercury capture. The ionic form of mercury, that which has reacted with oxygen or chlorine, tends to be soluble in water and is therefore captured along with the SO2, while the elemental mercury, being insoluble in water, passes through most of these processes. Therefore, our best understanding of the co-control of mercury with SO2 control processes suggests that the efficiency of mercury capture by these processes is related to the amount of the mercury that has converted from the elemental form to the ionic form. Anything that would help convert the elemental mercury to the ionic form will presumably increase the overall mercury control in plants equipped with wet scrubbers. (NOx control processes using selective catalytic reduction systems appear under some circumstances, and with some coals, to increase the amount of ionic mercury, and this will be discussed later.)
The biggest influence on the eventual form of mercury in the flue gas, and the apparent subsequent capture efficiency, appears to be the chlorine content of the coal. Coals with higher chlorine levels, when burned in a power plant, produce flue gas that is typically higher in the ionic form, the form which is most easily captured in an SO2 scrubber system. In general, the domestic coals found east of the Mississippi River tend to be much higher in chlorine content than the coals found in the West.
More specifically, the rank of the coal tends to be a good predictor of chlorine content. Coal rank is an indicator of the age of the coal and there are four major classifications of coal rank, listed in the order of high rank (or older coal) to low rank (or younger coal): anthracite, bituminous, sub bituminous, and lignite. Most coal found in the Eastern U.S. is bituminous coal, although there are some lignite deposits found in the Alabama-Mississippi coastal plain. These lignite reserves are not important to the coal-fired utility industry, however. Conversely, most of the coal found in the Western U.S., including Texas, is either sub bituminous or lignite rank coal. The exception in the West is some bituminous coal found in Colorado extending into New Mexico. All of the coals in the Western U.S., including the Western bituminous coals, are characterized by low chlorine contents, while the bituminous coals in the Eastern U.S. have much higher chlorine contents. Therefore, the expected amount of ionic mercury and consequently the expected capture in a scrubber will be much higher for coals from the Eastern U.S. than from those in the Western U.S.
Typical coal-fired power plant flue gas produced from combustion of the bituminous coals found in the Eastern U.S. would contain the following proportions of the mercury species: 60% ionic mercury, 38% elemental mercury, and 2% particulate mercury. The particulate mercury would be removed in the power plant’s electrostatic precipitator. We would expect the scrubber to remove 90 to 95% of the ionic mercury, and none of the elemental mercury. The overall mercury removal in this simple example would then be 56% (90% of the ionic and nearly 100% of the particulate mercury removed). This example is in good agreement with recent testing where, at three bituminous-fired power plants studied by EPRI, the FGD system removed 43 to 51% of the mercury.
However, most of the coals from the Western U.S. when used in a power plant produce much less ionic mercury, with typical estimates of: 25% ionic, 74% elemental, and less than 1% particulate. A scrubber on this power plant would then only be expected to remove 90% of the ionic and the electrostatic precipitator or bag house to remove nearly 100% of the particulate mercury. Therefore, the total mercury removal would be only 23.5%. The ICR database shows that power plants burning low rank coals ranged from near zero to 38% mercury capture without wet scrubbers, and 11 to 56% on those plants with scrubbers.
A problem with capturing mercury in wet FGD scrubbers has been discovered through analysis of the EPA Information Collection Request database. In some power plants that were tested for mercury species and also had wet SO2 scrubbers, the apparent high capture of ionic mercury was offset by an increase in the amount of elemental mercury as the flue gas moved through the scrubber. So, while the ionic mercury appeared to be captured at efficiencies approaching 95%, some of the ionic mercury, after being captured in the scrubber, was converted back to the elemental form, which evaporated from the scrubber and was then emitted as elemental mercury.
An example may help explain the effect. Say that, before the scrubber, there are 10 micrograms (one millionth of a gram or 2 billionth’s of a pound) of mercury in one cubic meter (about 35 cubic feet) of flue gas. Furthermore, let’s say that 60% of that is ionic and the balance is elemental, or 6 micrograms per cubic meter ionic and 4 micrograms per cubic meter of elemental mercury. In a power plant that shows this mercury release phenomena, we might see less than 0.1 microgram per cubic meter of ionic mercury at the stack exit, an apparent capture of 98.3% of the ionic mercury. But, we see the stack exit containing maybe 5.5 micrograms per cubic meter of elemental mercury, an increase of 37.5%.
The elemental mercury is not being captured but is actually increasing across the scrubber. When looking at the total mercury, the 10 micrograms per cubic meter at the scrubber inlet is reduced to only 5.6 micrograms per cubic meter (5.5 elemental and 0.1 ionic) at the stack, a total reduction of only 44%. The only logical explanation to explain these example numbers is that some of the captured ionic mercury is being re-released as elemental mercury. In this case, the ionic mercury is only being captured at 73%, when the re-released mercury is included.
This scrubber mercury re-release is not well understood at this point. An analysis by EPRI notes a correlation between an increase in the amount of fly ash captured in the scrubber and an increase in the mercury re-release. Further work by EPRI on a bench-scale scrubber shows that this phenomenon is transient, and it is not easy to predict when it will occur. Additionally, private testing by Southern Company at our DOE-sponsored flue gas scrubber at Georgia Power’s Plant Yates, south of Atlanta, has shown that this effect is present at some times, and not present at others. The significance of this effect is that the overall capture of mercury by a wet scrubber may be less over time than a short test period would indicate. Further research of this phenomenon is needed.
Most of the previous discussion assumes that the FGD process used is the wet limestone, forced-oxidation scrubber. Another process for SO2 control, used widely for low sulfur Western coals, is a lime-based spray dryer followed by a bag house that collects both the reacted lime along with all of the coal ash. The EPA Information Collection Request testing in 1999 indicates that this spray dryer-bag house FGD process may give very high mercury removals with bituminous coals. However, this is a rare application of this technology, and unfortunately is not widely applicable to all bituminous coal applications. The technology is only effective for SO2 control for low sulfur coals, is more expensive than the alternatives, and creates a large waste stream that has to be carefully handled for disposal. While this approach may be used in a few power plants burning Eastern bituminous coal for combined SO2 and mercury control, I do not expect it to be very widely selected because of these limitations.
Ironically, the best application of this FGD process is for Western coals, but there it appears to make the mercury control worse than just particulate control alone. That is, the use of a spray dryer-bag house system on most low rank coals (sub bituminous and lignite) is normally the best engineering and low-cost FGD solution for plants burning these coals for SO2 control, but the evidence suggests that it may worsen the mercury collection efficiency as compared to the use of a bag house alone. For example, EPA states that sub bituminous coal plants in the ICR database with only bag houses average 72% mercury control, while those with a bag house and a spray dryer for SO2 control average only 24% mercury removal.
Various technologies are being investigated to attempt to further oxidize elemental mercury to ensure higher removal in a FGD system. Chemical injection, plasma discharges, and dedicated catalysts are all being tested and developed. These approaches are all under development, and only slow progress is being made.
Selective Catalytic and Non-Catalytic Reduction (SCR & SNCR) NOx Controls
One of the most intriguing possibilities is the ability of NOx control selective catalytic reduction (SCR) systems to enhance the amount of ionic mercury in the flue gas. A report on research done by a large German utility company in the early 1990’s claims that the catalyst used in a SCR system was effective in converting a high fraction of the elemental mercury to the ionic form, which was then captured in FGD equipment. The German claim was that the SCR catalyst changed the chlorine chemistry, making it more likely to convert elemental mercury to ionic mercury.
Based on this German research, EPA originally assumed that any power plant equipped with a SCR and FGD, burning any type of coal, would see: (1) almost all of the elemental mercury converted to ionic; (2) the ionic mercury captured in a scrubber in a high proportion; and (3) no mercury re-released from the FGD process – all adding up to an estimate of an overall 95% reduction in mercury emissions from those plants. A 95% mercury capture would require that the SCR catalyst be 97.5% effective in converting elemental to ionic mercury. Furthermore, the FGD system would have to be 97.5% effective in removing the ionic mercury – that is, not only does the scrubber have to perform at least as well on mercury as the SO2 (even though the mercury is one-millionth times as concentrated), but no re-release of mercury can occur. EPA’s assumptions were highly optimistic and recent power plant testing has shown these assumptions are not always true.
SCR catalyst degrades over time in its performance to reduce NOx, requiring replacement every three to five years. The catalytic activity is reduced by exposure to flue gas, either by poisoning of the catalyst active ingredient from the chemicals in the flue gas or by physical plugging of the catalyst surface by ash particles. It is not known, at present, how this catalyst deactivation affects its ability to oxidize mercury. The mercury oxidation of the catalyst could be reduced at the same rate as the NOx reduction, or it might be slower or faster. EPRI testing has only looked at two power plants and only in two ozone seasons (May 1 to September 30). So we have limited information, both in the number of plants tested and the time between tests. Therefore, any estimate of the long-term potential for co-benefits of SCR and FGD for mercury reductions must consider the possibility of catalyst aging and the subsequent potential loss in mercury oxidation.
For the lower rank coals, and particularly those found in the Western U.S., this SCR mercury oxidation does not appear to occur. Given the German claim of the effect being based on higher chlorine content, this is not much of a surprise. The low rank coals are typically low in chlorine, and to make matters worse, the ash of these coals is alkaline, so that whatever chlorine that is present, being an acid, is usually neutralized by the fly ash before it can ever reach the SCR catalyst. Testing in an EPRI program sponsored by utilities (including Southern Company) along with the Department of Energy (DOE) and the EPA has shown that mercury reduction in low rank coals do not seem to be helped by the addition of a SCR system. Since the majority of the mercury in the flue gases from these coals in the elemental state, the addition of any type of FGD system does not appear to control mercury emissions to any significant degree. In other words, for low rank coals (typically Western U.S. coals), we do see modest benefits on mercury control by adding wet FGD systems, but do not see any mercury co-benefits from adding an SCR to the power plants burning these coals. EPA has also seen the results of the testing, and we think that they have revised their assumptions about co-benefits for lignite and sub bituminous coal to reflect this new knowledge, that is, there are only modest mercury reductions based on co-benefits of NOx and SO2 reductions for these coals.
At the beginning of the MACT development process, EPA had assumed that selective non-catalytic reduction (SNCR) systems would contribute to increased mercury removal, and explicitly had assumptions about its performance in their models. SNCR uses ammonia injection at elevated temperatures (1900-2400°F) to reduce NOx without the use of a catalyst. Two years of testing have shown that this NOx reduction technology has no influence on mercury control in any plant with any coal rank. Finally, we think that the Agency has conceded this point and we hope that they no longer count SNCR as having any influence on mercury control.
Summarizing the current state of knowledge of controlling mercury via co-benefits of SO2 and NOx reductions, there are only a handful of power plants that have been tested for short time periods. Given this limited amount of data, we think that for bituminous coals the mercury reductions with a SCR and FGD will probably be between 80-90% for the best case, and that for sub bituminous and lignite coals the reduction will be a modest 20%. These estimates are optimistic taking into account the previous discussions of catalyst aging in SCR systems and mercury re-release for FGD systems, and are likely to be reduced even further in the future. We think that EPA is currently using an estimate of 90% for bituminous coals and something less than 90% for lignite and sub bituminous.
Activated Carbon Injection
The second near-commercial technology for mercury control from coal-fired power plants is activated carbon injection (ACI). Activated carbon is a specially prepared product of coal or biomass that is able to adsorb many chemicals from gases or liquids. One of the primary uses of activated carbon is the treatment of drinking water. Water filtering systems sold for home use in home improvement stores are typically cartridge systems that include activated carbon as part of the filter. Activated carbon is being used currently to remove mercury from the flue gases from municipal, medical, and hazardous waste incinerators. In those applications, activated carbon can routinely collect over 90% of the mercury from the flue gas. However, the mercury concentrations in the stack after the activated carbon treatment in these incinerators are typically higher than that found in coal flue gas before treatment. That is, the amount of mercury in every cubic foot of incinerator stack gases after the control system using activated carbon is typically 5 to 10 times the amount in untreated coal flue gases from power plants. Another way to look at a comparison between incinerators and power plants is that most every power plant would meet the incinerator mercury regulations without any control technologies. Simply, incinerator mercury control by activated carbon stops where power plant flue gases begin. Therefore, it is not useful to use the experience of activated carbon in incinerators to inform the debate on its use in power plants.
The design of activated carbon injection for mercury control relies upon the existing equipment used to remove fly ash from the flue gas to also remove the added activated carbon. There are many side issues associated with the use of activated carbon in this mercury process approach, including contamination of the fly ash with carbon and interruption of the normal fly ash control by the added load of activated carbon. The injection ahead of electrostatic precipitators, which are in use by about 80% of the U.S. coal power plants, may require large amounts of activated carbon to achieve reasonable mercury control. The carbon will contaminate the fly ash making it unusable for recycling and may threaten the performance of the electrostatic precipitator for its intended use of removing fly ash. Injection of activated carbon in a bag house will not need as much activated carbon as an electrostatic precipitator, but will also contaminate the fly ash.
There have been only a handful of tests on the use of activated carbon to control mercury from coal-fired power plants. The very first test at full-scale in the United States was performed at a Southern Company power plant, Alabama Power’s E.C. Gaston Unit 3, located in Wilsonville, Alabama. This was the first in a series of four power plant tests in a sequence performed by ADA-Environmental Solutions of Littleton, Colorado. The test program was sponsored by DOE’s National Energy Technology Laboratory (NETL) with significant co-funding by participating utilities and vendors. All of these four sites are somewhat unique, and unfortunately do not well represent the nation’s power plant fleet.
Gaston Unit 3 is one of only four power plants in the U.S. that have an advanced particulate control system that consists of a small bag house installed downstream of the existing electrostatic precipitator. This arrangement, known as COHPACTM, is a patented EPRI invention. The activated carbon can be injected between the electrostatic precipitator and the bag house. The electrostatic precipitator collects over 95% of the fly ash, while the bag house collects the remainder of the ash and the activated carbon. This approach to activated carbon injection avoids contamination of the fly ash and does not jeopardize the operation of the electrostatic precipitator with additional carbon loading. The bag house is a large filter, which has hundreds of fabric bags that separate the solid ash and carbon from the flue gases, much like the paper bag in a household vacuum cleaner. Because the activated carbon can sit on the surface of the bags for several minutes and see a substantial amount of flue gas, it can effectively collect more mercury from the flue gas than injection into an electrostatic precipitator.
The activated carbon injection testing at Gaston, which burns an Eastern U.S. bituminous coal, ended with a seven-day test of mercury control, where the average mercury reduction over that time period was just under 80%, with a high of over 90% and a low of only 36%. This was a short-term test and probably does not reflect the ability of this system to always perform at this level. We found in this testing that the bag house at Gaston is not big enough to accommodate the amount of activated carbon needed to consistently achieve 90% mercury control for even just one week of testing. The testing was promising and DOE/NETL has funded a follow-on project that will test the mercury control at this location for one calendar year. This length of testing will allow a better estimate of the potential mercury control from this technology over the course of that one year. We are just starting this longer term testing, and the initial results were presented at an international pollution control conference sponsored by DOE, EPA, and EPRI just two weeks ago here in Washington. The initial results are not encouraging – we cannot repeat the performance of the seven-day test performed in 2001. The electrostatic precipitator ahead of the bag house at Gaston Unit 3 is not performing as well as it was during the earlier testing, and we cannot inject much activated carbon into this system without causing damage to the bag house. Two conclusions can be drawn from the first few weeks of operation of the long-term testing: (1) the bag house at this unit is simply not big enough to handle both the fly ash and carbon loading over all operating conditions, and (2) the 80% average mercury control seen in the earlier one week test cannot be sustained over the long term. It may be possible to achieve levels higher than 80% in other power plants with this configuration, assuming that the additional capital investment is made to build a large bag house. Again, this is a test at a power plant burning Eastern bituminous coal.
The three other tests of full-scale mercury control using activated carbon in the joint industry-DOE project all involve the injection of activated carbon into the inlet of an electrostatic precipitator. The first electrostatic precipitator injection test was performed at Wisconsin Electric’s (now We Energies) Pleasant Prairie Power Plant, which burns a Western U.S. sub bituminous coal from the Powder River Basin in Wyoming and Montana. This unit has a large electrostatic precipitator that is likely to be able to handle the additional particle loading from the activated carbon. The test that occurred over one to two weeks was able to achieve a mercury control of between 60 and 70%, but notany higher, regardless of the amount of carbon injected into the system. The logical conclusion from the testing seems to indicate that there is a chemical limitation on the amount of mercury control from low rank coals like lignite and sub bituminous, and maybe for Western U.S. bituminous coals from Colorado and New Mexico. It appears that, similar to the SCR oxidation of mercury, the activated carbon needs sufficient chlorine in the flue gas to collect the mercury. Again, this result was over a very limited time span test and may not be repeatable over a yearlong period. Longer term testing of this approach in several power plants needs to be performed before any judgment of the mercury performance can be reliably made.
An additional consequence became clear during the test at We Energies’ Pleasant Prairie Power Plant. This site is able to sell all of the fly ash it produces for recycling into concrete. The activated carbon made the ash not usable for this purpose during the test period, but also contaminated the ash for about four weeks after carbon injection was discontinued. Southern Company declined a similar test at one of our sub bituminous coal plants, due to the expense of lost ash sales plus the added ash disposal costs.
The other two tests of activated carbon injection into electrostatic precipitators for mercury control were both performed in Massachusetts, at PG&E National Energy Group’s Salem Harbor and Brayton Point power plants. Salem Harbor is peculiar in that it produces a large fraction of unburned coal particles that persist into the electrostatic precipitator, possibly a result of the large amount of South American coal being burned there. This high level of carbon produced seems to remove a significant amount of mercury, with a baseline removal ranging from 87 to 94% with one coal, but dropping to 50 to 70% with a second coal, all even before activated carbon injection. The activated carbon injection was able to increase the mercury capture to over 90%. Of course, this testing has shown that a change of coal supply can dramatically change the mercury baseline performance and the subsequent increased capture by activated carbon injection.
Brayton Point is also a peculiar arrangement with two electrostatic precipitators in series. In the DOE test, activated carbon was injected between the two electrostatic precipitators, much like the injection between the ESP and bag house at the Gaston station. The baseline mercury removal, that is, the removal before activated carbon injection started, was 90.8%. This is very high as compared to historical data from that unit that recorded baseline mercury removals of 29 to 75%. The results in the ten days of testing suggest that, for short periods, the injection of activated carbon can increase the mercury removal from a baseline of 90.8% to 94.5% with the addition of activated carbon (10 pounds carbon injected for every million cubic feet of flue gas). Again, the short time of the test and the potential change in behavior with a change in coal supply makes it hard to extrapolate this performance much beyond the actual period of testing.
All of the electrostatic precipitator tests of activated carbon injection to date have involved relatively large, oversized equipment where the additional burden of collecting the injected activated carbon did not impact the operation, at least in the tests of under two weeks duration. For the same mercury collection efficiency as a COHPACTM bag house, the added carbon cost is substantial enough to justify the capital investment to build the bag house.
Another – potentially large – problem with this technology is that the supply of activated carbon is currently not sufficient to support any significant use for utility mercury control. I have publicly stated that, due to current uncertainties, Southern Company may use anywhere between 500 tons per year to 100,000 tons per year of activated carbon. The major U.S. manufacturer of activated carbon, Norit Americas, based in Atlanta, Georgia, have told us that they could supply an additional 20,000 tons per year with their existing capacity. Without long-term commitments from buyers, the activated carbon suppliers will very likely not make the needed investments to ensure that a large demand from the U.S. utility market could be met. In the 1970’s, the activated carbon industry built capacity in anticipation of clean water regulations and those investments resulted in a severe price decrease caused by oversupply, when the demand did not appear. The activated carbon suppliers are not likely to make the same speculative capital investments today. Add to this reluctance to invest ahead of demand the fact that it will likely take at least five years to design, finance, permit, and build activation carbon production facilities, and it becomes apparent that, if activated carbon injection becomes the technology of choice for power plant mercury control, the supply will not be available at the beginning.
There may be foreign supplies of activated carbon. As discussed at a recent conference, there may be about 50,000 to 60,000 tons per year available from a major European supplier. Also, China has started supplying activated carbon into the U.S. market, but initial experience with this material has shown quality control problems with its performance. All in all, there may be sufficient carbon available to supply a small part of the industry with today’s global supply, but there is not enough supply for any major use across the nation by the utility industry.
In early modeling efforts by EPA on the performance of activated carbon, the assumptions made about performance and the actual amount of activated carbon were grossly optimistic. The Agency used some estimates made by DOE in 1999, and the subsequent testing at full scale power plants has demonstrated that the performance is not as good as the earlier estimates. We think that the current set of performance and cost numbers offered by the Utility Air Regulatory Group in the MACT Working Group are the best estimate for mercury control processes using activated carbon.
In summary, the limited testing of activated carbon injection for power plant mercury control does not represent the average configuration of the U.S. power plant fleet, and the short-term tests that have taken place only represent what a well-controlled and well-managed test period performance could be – in other words, are likely to be close to the best case. Additional testing at the Southern Company plant has already shown that the earlier performance cannot be matched at this moment. Certainly additional testing, including long-term tests of at least eight months are needed to understand what the actual performance of activated carbon injection over longer times would be, with the wide variety of coals in use today. At this moment, the DOE/NETL is evaluating a number of proposals from utilities, vendors, and research contractors to test activated carbon for longer periods of time on a variety of plants, especially those that burn low rank coals.
With sufficient capital investment to build a COHPACTM bag house large enough to handle both the fly ash and activated carbon, short-term performance of 90% mercury removal with bituminous coals may be possible, but, across the industry, an average removal of 80% is more likely to be achieved with today’s technology. This estimate is based on only one power plant, tested for only seven days, however. It appears that low rank coals, such as lignite and sub bituminous coals, may have a limit of 60-70% mercury removal, regardless of the amount of activated carbon used or whether a bag house has been installed. Again, only one power plant has been tested for less than two weeks to establish this estimate. Under certain circumstances, activated carbon injection into a large ESP may be able to get incremental mercury control, but only two power plants have been tested for less than two weeks. Finally, the supply of activated carbon is not sufficient today to accommodate a substantial demand from the utility sector and it may take five years to bring new activated carbon production facilities on line.
There are other technologies that show some promise in controlling mercury emissions from power plants, but they are all still research projects and are nowhere close to commercialization. Some of the multi-pollutant processes being developed do claim that mercury control is also removed along with SO2, particulates, and NOx. While this may be true, there are large questions about the costs, reliability, and long-term performance of these technologies. Most of these multi-pollutant processes make either fertilizer or acid chemical feedstocks from the NOx and SO2, and the ability to sell either of these waste streams in the future is questionable. The larger the penetration of these technologies into the utility market, the more of the byproducts that are produced, quickly over-saturating any potential market.
Possible future technologies that are being researched include capture of mercury by gold-plated surfaces, the use of chlorine addition to low rank coals to increase the mercury oxidation, injection of sulfur compounds to change the elemental and ionic mercury gases to solid sulfides that can be captured in the existing particulate control devices. Additionally, a large number of alternative sorbents to replace activated carbon, either with a less costly material cost or improved performance with less material injected, are under development. Unfortunately, we cannot predict whether these efforts will succeed, and we cannot base national energy policy on the hope that something is invented in time to produce the perceived needed level of mercury control.
Timing of Mercury Reductions
The timing of mercury reductions required, whether by regulations under a MACT provision or by a legislative process, needs to take under consideration both the state of knowledge about mercury control and the ability of the nation’s utility industry to install the required controls. Already, in the installation of NOx controls for the 2003 summer ozone season, we have experienced some labor shortages and tight supplies of steel, cranes, and auxiliary equipment such as fans, pumps, electric motors, switchgear, etc. If mercury control proceeds under a MACT regulation, every coal-fired power plant will have to meet the stated emissions requirements, and depending on the technologies being used, we expect shortages of steel, bag house bags, labor, and auxiliary equipment, not to mention the activated carbon supply issues discussed earlier. Southern Company estimates that the time required to install mercury controls under MACT would be at least seven years, and the time needed for the additional NOx and SO2 controls in Clear Skies would take probably eight to nine years.
Estimates of Benefits of Utility Mercury Reductions
EPRI and EPA are both engaged in research to attempt to predict the net effect on human health from reductions in emissions from U.S. coal-fired power plants. EPRI has just published their initial findings, and we think that EPA is working on similar model predictions. In the EPRI study, mercury deposition on the continental U.S. is predicted using a global mercury source and deposition model. The results indicate that the majority, around 70%, of the mercury falling on the U.S. is from sources outside the U.S. Additionally, this study predicts that U.S. utility emissions are estimated to contribute less than 8% of the mercury depositing in the U.S. This result is significant, because it indicates that reductions of mercury emissions from domestic utility sources will have a limited response on the amount of mercury depositing. In other words, since most of the mercury falling on the U.S. comes from overseas, controlling domestic utility emissions can have only a limited impact.
The EPRI study goes on to estimate the change in human exposure from significant reductions in utility mercury reductions. The only significant route of exposure to humans is through the consumption of large fish, captured in the wild. By estimating the change in U.S. deposition from reductions in utility emissions, the change in mercury in aquatic systems, and subsequently in fish, can be found. Taking the analysis one step further, EPRI has estimated the change in exposure to humans in the U.S. from utility mercury reductions.
The EPRI study looked at mercury reductions in a Clear Skies Act approach and in a mercury MACT regulation scenario. The results indicate under the Clear Skies approach, in the year 2020, mercury deposition in the continental U.S. would be reduced by an average of 1.5%, exposure of women of childbearing age to mercury would be reduced by 0.5%, and the fraction of the population above the reference dose for mercury would be reduced by only 0.064%. In the MACT approach, also for the year 2020, mercury deposition would be reduced by 1.2%, exposure of women of childbearing age to mercury would be reduced by 0.4%, and the fraction of the population above the reference dose would be reduced by 0.055%. Since U.S. utility emissions are only a small contributor to mercury in the environment, it is not surprising that significant reductions in those emissions will not greatly affect human exposure. One significant difference in the two approaches is that the present value incremental cost for mercury controls by 2020 is estimated to be about $6 billion for CSA and $19 billion for MACT.
There are no commercially available technologies for mercury controls for coal-fired power plants. There are systems in use in the waste incinerator industry, but the EPA requirements for mercury control for incinerators allow emitted concentrations to be five to ten times higher than uncontrolled coal power plant emissions. In an engineering sense, the low concentrations mean that you have to work that much harder to get each molecule of mercury. NOx and SO2 stack concentrations are one million times higher than mercury, so you have to work one million times harder to collect mercury as compared to either NOx or SO2.
There are two near-commercial mercury control technologies at present: co-control by FGD systems, with possible beneficial mercury chemical changes from SCR systems on plants burning bituminous coals, and the injection of activated carbon into existing or new particulate control devices, either ESPs or bag houses.
Plants burning bituminous coal from the Eastern U.S. which have installed SCR systems and wet scrubbers are likely to have between 80 and 90% mercury control in the beginning. There are large uncertainties about the potential adverse scrubber chemistry that could re-release captured mercury and also about the extent of SCR catalytic mercury oxidation over time, so it is likely that these estimates may decrease as we learn more.
For low rank coals such as sub bituminous and lignite (along with bituminous coal from the Western U.S.), the SCR systems do not appear to have any beneficial effects on mercury chemistry, probably due to the low chlorine content of the coals. Additionally, the addition of a wet FGD scrubber system may increase mercury control slightly, say by 20%, but the addition of a spray-dryer FGD system may even decrease the mercury removal as compared to the pre-FGD mercury removal performance.
Activated carbon tests to date have been short, less than two weeks, and have shown some promise, but also some difficulties. The only long-term test that is being performed is at Southern Company’s Plant Gaston, and the year long test is just beginning. The limited data from this one short test suggests that activated carbon injection into a COHPACTM bag house installed at a plant burning bituminous coal may be able to achieve short-term performance of 90% mercury removal, but an average across a year is more likely to be around 80%. We do not know what operation problems may occur after an extended period of activated carbon injection, but even at the beginning of the year long test, we are not able to match the previous short term performance.
Activated carbon injected into an electrostatic precipitator at a plant burning Powder River Basin sub bituminous coal has shown mercury removal of 60-70%, but only for a short test, and with serious consequences for ash sales and disposal. The chemistry of low rank coals like these may limit the final mercury removal that can be achieved with activated carbon. Again, based on this one power plant test for a short period, it is likely that a bag house and activated carbon injection would still only achieve 60-70% mercury removal on these coals.
Activated carbon supply is also an unanswered question. Activated carbon vendors have estimated the U.S. utility market may be between 500,000 and 1,500,000 tons per year. Between domestic supply and spare European capacity, there may be up to 150,000 tons per year available today. Without firm commitments, the suppliers are unwilling to make the investments to increase the supply, indicating that widespread use by the utility industry may create a worldwide shortage of activated carbon. Given that it takes roughly five years to bring a new activated carbon production facility on line, the prospects for widespread availability of activated carbon may be questionable.
In addition, the shortages encountered during the installation of NOx controls over the last several years have shown that shortages of labor, steel, cranes, and auxiliary equipment can occur, and installation of mercury controls under a MACT regulation or installation of more NOx and SO2 controls will surely cause even greater material and labor shortages. The only way to alleviate the shortages is to extend the required performance date to install the equipment. These shortages could spill over into other industries and cause price increases across the board.
There are other technologies under development for mercury control, but they are all very much still in a research stage. Various multi-pollutant processes are being touted, but they suffer from questions about performance, cost, and waste disposal issues. Other processes to specifically affect or capture mercury are also under development, but are at least eight to fifteen years away from deployment, if they work at all.
More tests and longer tests are needed to be able to reliably estimate performance and design the appropriate equipment and processes for mercury reductions in power plants with different equipment installed and burning different ranks of coal. The Department of Energy is currently evaluating a number of proposals from the utility industry, vendors, and research organizations to test a wide variety of plants and coals for mercury control, over a longer test period. The electric power industry, along with EPRI and equipment vendors, is engaged in a large, coordinated effort to develop and optimize cost-effective mercury emission reduction processes.
EPRI modeling suggests that U.S. utility emissions of mercury are only a small contributor to deposition of mercury in the continental U.S. Significant reductions of those emissions, either under a CSA or MACT approach, will only reduce deposition in the U.S. by 1.5%, and will only decrease exposures of the most sensitive population of women of childbearing age by 0.5% in 2020, as compared to 1999.
The utility industry does not have proven technologies to reduce mercury emissions, but we know that some reductions will occur as SO2 and NOx control systems are installed, either under Clear Skies or business-as-usual. The industry does not hold the position that mercury reductions should not occur, but asks that right timeline should be followed, one that considers the practical aspects of the cost and impact of making these reductions. Mercury emission reductions that are required before the technology has been fully developed will lead to significantly increased costs, to likely fuel switching from coal to natural gas, and to possible disruption of the nation’s energy supply.