In incinerators, as with any engine that burns organic fuel, startup often results in puffs of particulate and organic emissions due to incomplete combustion of the fuel, cold combustion temperatures, and other factors. Shutdown of an incinerator is also a time for nonsteady-state combustion, and an opportunity for larger-than-usual emissions. After startup, optimal incinerator operations are desired, but not always achieved. Instances in which the temperature and combustion efficiency fall are called upsets. These usually are a result of operator error, equipment failure, or when the BTU value of the waste rises or falls precipitously, due to moisture content or inherent incombustibility of the waste. Upsets can occur due to maladjustment of the overfire and underfire air distribution systems, improper grate speed, blockages in the waste charging chute, malfunction in the auxiliary burner, a combinations of these, and other combustion problems. Poor operator control, either of the furnace (e.g., permitting temperature to drop or oxygen to drop or increase too much) or of the stoking operation (e.g., permitting poorly mixed wastes to be fed into the furnace), can cause reduced combustion efficiency.

In this paper operational data generated during cold starts and upsets from six municipal waste combustors (MWCs) will be presented to explore each of the major operational causes of cold starts and upsets and to quantify the increased emissions caused by these conditions.

The types of emissions generated by various incinerator designs and operations during normal conditions usually falls within ranges bounded by incinerator emission regulations, and have been described previously by this author, and in a recent report of the National Research Council. Optimum combustion depends upon maintenance of temperatures in the range of 1800 to 2000^oF, adequate availability of oxygen for combustion (i.e., turbulence), and residence time of the flue gases at these temperature and oxygen conditions. Rapidly attaining uniform gas flow through the incinerator is also important in maintaining steady-state combustion conditions. But in incinerators, as with any engine that burns organic fuel, startup can result in puffs of particulate and organic emissions due to incomplete combustion of the fuel, cold combustion temperatures, and other factors.

After startup, during the course of normal incinerator operation, optimal incinerator operations are desired, but not always achieved. Instances in which the temperature and combustion efficiency fall are called upsets. Incomplete combustion is due to unstable gas flow due to inhomogeneous conditions. Due to the fact that the moisture and BTU content of the heterogeneous waste streams fed to incinerators are constantly changing, and that equipment can be maladjusted due to operator failure, or over time, the equipment can fail. Thus, upsets can occur due to a number of causes, some related to design and others related to operation and maintenance.

Some designs of incinerator are not suited to maintenance of optimal combustion conditions at all times. Batch charging of waste, common in small waste incinerators, is inherently discontinuous, can adversely affect the combustion process in the furnace, and contributes to greater emissions relative to the continuous feed method. In batch feeding as new refuse is charged, combustion rate is dampened by the sudden advent of cooler conditions; but inbetween refuse charging, the waste burns vigorously. Such cycles can occur every ten minutes. Continuous feed systems permit more optimal, steady-state combustion conditions and is the preferred alternative available today for charging waste into an incinerator. Because incinerators are designed to operate within a minimum and maximum waste load, too little waste necessitates periodic shut down and startup. Similarly, the more that a plant is started up and shut down (for maintenance, inadequate or varying waste stream volume, or whatever reason), the more uneven is the combustion, and the greater the potential for unwanted emissions.

The inherent heterogeneity of wastes fed to incinerators presents the operator with a challenging problem: keeping many combustion parameters optimized at all times and ensuring that the equipment is in good working order. For example, the changing nature of the waste stream results in a need to adjust underfire and/or overfire air injection. If too little air is injected, incomplete combustion can result,(i.e. too few oxygens for the carbons released from the waste stream). Too much air can result in cooling of the furnace temperature and lowered residence time for combustion gases at higher temperature, such that it is insufficient to destroy organic compounds that volatilize from the wastes on the grate. The latter condition can also result in entrainment of more particulate matter into flue gases. Malfunction or maladjustment of the auxiliary burner, used to heat up the furnace and keep it within an optimal temperature range, or an overly fast grate speed (which results in insufficient drying of wet waste) can cause lower flame temperature and decreased destruction of organic compounds in the flue gas. Upsets are also caused by blockages in the waste charging chute, holes in furnace walls, and combinations of these and those mentioned previously, resulting from operator error and inadequate adjustment or repair of stoker, grate, and furnace equipment. Generally, upsets are characterized by low combustion efficiency and/or high particulate generation either on a short-term, transient basis or on a continuing basis, usually marked by high carbon monoxide (CO) levels.

Incinerator design can also affect the degree to which pollutants are formed. Incinerators with a single combustion chamber do not have as much flexibility as dual or multiple-chambered units in managing all the combustion parameters to minimize different emissions at the same time. This is because in a multiple-hearth unit, there is more than one opportunity for combustion of all carbon atoms in the flue gases to take place at the appropriate temperature (i.e., increased residence time). In a single-chambered unit, to achieve the same combustion rate, much more care must be taken.

What with the number of possible operational problems and design configurations, predicting the timing and severity of upsets is next to impossible. If any of these operating conditions is present, products of incomplete combustion (PICs) will probably be created -- the worse the conditions, the worse the level of emissions produced. These PICs include CO, and a wide variety of organic compounds (e.g., dioxin and furan and their chemical precursors: benzenes, phenols and their chlorinated forms), many of which are toxic and/or carcinogenic at very low concentrations. As a result, the public living near incinerators is uncertain and sometimes anxious about organic emissions caused by upsets. In the following sections this problem will be discussed further, and data will be presented to characterize the operational causes and quantify the increased emissions caused by cold starts and upsets.

The following briefly explains some of the most important reasons for emissions caused by cold starts and upsets.

Combustion-related emissions are sometimes related to insufficient temperature to destroy organic compounds that volatilize from the wastes on the grate when an incinerator starts up. Other causes of low temperature occur when wet waste is fed to the incinerator, or when excessive oxygen enters the furnace, or the operator does not adjust the grate speed to account for a change in waste BTU content. If the auxiliary burner is inoperative or poorly adjusted, temperatures can go outside the normal operating range.

Another cause of inefficient combustion of waste in an incinerator is insufficient oxygen present to complete combustion. This can be caused by operator error (e.g., the control room operator does not ensure that sufficient air is injected either below or above the grate), or the crane operator does not mix the wastes sufficiently, resulting in occasional increases in BTU content of the waste (which demands more oxygen to complete combustion), or the operator does not adjust the grate speed to account for a change in waste BTU content.

Yet another cause of upset conditions is excessive injection of air into the furnace, which cools combustion, decreasing the efficiency of PIC destruction. Again, operator error can be the cause, or a decrease in the BTU content of the waste caused by insufficient mixing of waste, or neglecting to repair holes in the sides of the furnace (which permits unlimited amounts of relatively cold air to enter the furnace).

Insufficient mixing of the volatilized organics in the furnace with the injected oxygen can occur when the velocity of the combustion gases through the furnace is too high to allow complete oxidation to occur. This condition can be caused by operator error and maintenance issues as discussed under Excessive Oxygen. Insufficient residence time results in a lower destruction efficiency for PICs because the organics are not exposed to the high furnace temperatures long enough to destroy as many of the PICs. This condition is caused by poor furnace design (lacking bull noses and arches), but it can also be caused by excessive oxygen racing through the furnace.

Even if the total quantity of oxygen introduced into the furnace is sufficient to oxidize the volatilized organics in the furnace gases, and even if the overfire (above grate) and underfire (below grate) jets are properly aligned to maximize mixing and turbulence, another possible cause for upset is introduction of too much underfire air or too much overfire air relative to the total. Too much underfire air can cool the combustion on the grate and entrain particulate matter into the flue gases. Too much overfire air will result in a low rate of oxidation of the organics in that part of the furnace. Too little underfire or overfire air also has detrimental effects insofar as combustion efficiency is concerned.

Insufficient Time. If any of these operating conditions is present, products of incomplete combustion (PICs) will probably be created -- the worse the conditions, and the shorter the length of time the combustion gases are exposed to optimal temperature and oxygen conditions, the worse the level of emissions produced, as will be shown. These PICs include CO, and trace, though not insignificant quantities, of a wide variety of organic compounds (e.g., dioxin and furan and their chemical precursors, chlorinated benzenes and phenols), many of which are toxic and/or carcinogenic at very low concentrations. Optimization of mixing requires both turbulence and proper ratio of PICs to oxygen. If PICs are present in too high a quantity for the amount of oxygen present, the result is only partial combustion of PICs. On the other hand, if too much oxygen or air is present, the result is typically cooling of the combustion gases, which retards combustion efficiency and reduces PIC destruction efficiency. Thus, to maximize mixing, the furnace must be designed to promote it, and the correct ratio of oxygen to PIC must be maintained as closely as possible at all times. Generally speaking, upset conditions are characterized by low combustion efficiency and/or high particulate generation either on a short-term, transient basis or on a continuing basis, usually marked by high CO levels.

Thus, to summarize, in order to avoid upsets and maximize combustion efficiency it is necessary to maintain the appropriate temperature, time and turbulence -- the three "T"s of good combustion in order to maintain steady-state combustion. If combustion is not optimal, or if it is unstable, products of incomplete combustion (PICs) will be created. Maintenance of optimal combustion conditions in a furnace is ideally done in such a manner that the gases rising from the grate mix thoroughly and continuously with injected air, the optimal temperature range is maintained at all times by burning fossil fuel in an auxiliary burner during start-up, shut-down and during upsets, and by designing the furnace so that adequate turbulence and residence time for the combustion gases of at least one or two seconds at these conditions is achieved. Thus, the frequency of upsets in an incinerator depends, to some degree, on the design of the furnace but mainly on the operating practices and level of maintenance. The frequency of startup and shutdown depends on the amount of waste requiring incineration, and on the frequency of shutdowns for maintenance (both scheduled and unscheduled).

The foregoing discussion illustrates the great importance of operator training and vigilance, as well as maintenance. EPA requires that a site-specific operating manual must be developed and updated annually, although there are no specific universal requirements for its content or review of its accuracy. Operators have several key roles in optimizing combustion efficiency and minimizing the emissions of PICs in incinerators. For example, it has been shown that to optimize combustion efficiency and stability, the incoming waste must be mixed well prior to charging it into the furnace in order to distribute the moisture and BTU content as evenly as possible. The grate speed must vary with the combustibility of the waste charged. The amount of air injected into the furnace must vary as the composition of the waste being burned at any point in time varies; otherwise, insufficient or excess air in the combustion zone to burn the waste passing through at that time will result in combustion upset and products of incomplete combustion. Operators must inspect and ensure the integrity of the furnace walls and the functionality of mixing, charging, and combustion equipment to avoid upsets due to equipment malfunction. In most incinerators such features as mixing and stoking of waste into the incinerator, grate speed, overfire and underfire air injection rates, and selection of the temperature setpoint for the auxiliary burner are entirely or partially controlled by plant personnel, particularly the control room operator, the crane (refuse mixing) operator, the shift supervisor, and chief facility operator. It is of utmost importance they be trained and certified to know the variety of factors that promote upsets, and to be vigilant at all times.

Poor furnace design can cause an inability to achieve stable, optimal combustion conditions. For example, older MWC designs involved furnaces in which the combustion gases quickly came off the grate and swiftly passed out of the furnace. More recent designs have modified the furnace walls with "bull noses" and arches that retard upward movement of the combustion gases, and therefore enhance combustion efficiency by increasing residence time and turbulence. Older grate systems did not allow for variation in grate speed with variation in changes in the waste stream. Older designs did not have sophisticated controls for optimizing the quantity and location of air injection or for maintaining furnace temperature.

Once the undesirable components of the waste stream have been removed, the waste is fed into the drying area adjacent to the furnace. The charging process is in batches (which causes combustion to be less stable because of the uneven nature of the process), or continuously by charging the waste down an inclined chute onto a moving grate. It is preferable to have the waste spread out evenly to maximize combustion, so designs with inclined chutes for continuous charging are preferable.

The batch method of waste charging can adversely affect the combustion process in the furnace and contribute to greater emissions relative to the continuous feed method. This is due to the fact that efficient combustion is promoted by maintenance of combustion conditions as close to steady state as possible, but batch feeding is a discontinuous process, causing discontinuous combustion. In batch feeding as new refuse is charged, incineration is dampened by the sudden advent of cooler conditions, but inbetween refuse charging, the waste burns vigorously. Such cycles can occur every ten minutes. Designs for continuous feed chutes more closely approximate optimal, steady-state combustion conditions. A study sponsored by the American Society of Mechanical Engineers on the relationship between chlorine in waste streams and dioxin emissions, indicates that since combustion control is limited in most batch mode MWI’s, these incinerators "can be expected to emit relatively high PCDD/F levels associated with incomplete combustion."(page 3-3) This would support the idea that batch-fed incinerators should not be encouraged because of their inherent design factors that prevent good combustion. Thus, the continuous feed chute is the preferred alternative available today for charging waste into an incinerator since it maximizes steady state conditions and combustion efficiency.

Paradoxically, the preferred method which permits the greatest degree of waste screening for unwanted pollutant precursors (the tipping floor) has usually not been used in plants where there are crane-fed continuous feed chutes. This may have been due to size considerations. However, it is not difficult to envision a tipping floor where waste is screened adjacent to a pit into which waste is pushed by a front-end loader, and from which waste is fed by crane into a continuous stoker.

The concept of designing for continuous feed to promote steady-state combustion also applies to the degree of continuity in plant operations on the whole. The more that a plant is started up and shut down (for maintenance, inadequate or varying waste stream volume, or whatever reason), the more uneven is the combustion, and the greater the potential for unwanted emissions. For example, very small-scale incinerators -- on the order of 1 - 50 tons per day -- such as apartment and hospital incinerators, may only operate for a few hours per day. Some small municipal incinerators which are rated to burn as much as 250 tons per day, are sometimes underfed (or were originally designed to be oversized) and operate on one or two shifts per day. In these situations, the time spent starting up and shutting down, producing great concentrations of emissions, may be as long or longer than the time spent at more optimal combustion conditions. Thus, to minimize overall emissions due to incomplete combustion, it is necessary to charge waste continuously, to operate the incinerator 24 hours per day, and to match the size of the incinerator's furnace capacity to the waste that actually will be charged over a 24-hour period.

Furnace configurations vary considerably depending on when they were designed or built. The older designs, built prior to the late 1980s, many of which are still in operation, do not generally permit nearly as efficient combustion as the newer designs. In these earlier plants the entry of primary and secondary air into the furnace is not well controlled, and the excess air rates can be several times the amount required for adequate combustion of the off gases. Prime examples of this type of furnace are the unregulated and uncontrolled flue-fed apartment building incinerators, which still exist in some cities, and older municipal incinerators where the furnace walls extend straight up from the grate area. This design encourages greater upward velocities off the grate area, which increases the entrainment of particulate matter and reduces the residence time for combustion gases at the appropriate temperature and oxygen conditions. Poor designs in older incinerators also permit a great quantity of combustion gases to flow quickly through the furnace, reducing the temperature and mixing. Such designs clearly reduce combustion efficiency considerably. Such antiquated designs also result in cooling of the gases, further upsetting combustion efficiency.

In addition to aiding in combustion efficiency, optimization of primary air, maximization of residence time and maximization of turbulence also serve to reduce the entrainment of particulate matter (fly ash) upwards off the grate and through the furnace. This reduction in particulate generation by the furnace often decreases the possibility of greater particulate emissions from the incinerator. In fact, in certain incinerators known as dual-chambered (controlled-air) incinerators, the primary chamber, in which combustion initially takes place, is purposely fed less air so that it is starved for oxygen to some degree. While this will result in production of large quantities of PICs in the primary chamber, increased oxygen is injected into the secondary chamber to destroy the PICs. This arrangement minimizes particulate entrainment and also minimizes the creation of NOx (oxides of nitrogen) while not sacrificing destruction of PICs or combustion efficiency.

Considering the heterogeneous nature of municipal solid waste, with some components highly combustible and others not, strict maintenance of at least a minimum temperature throughout the furnace is necessary at all times. In modern incinerators maintenance of temperature is aided by means of auxiliary burners which are designed to be used during system startup, shutdown and upsets. These devices are fed fossil fuels and are typically set to come on automatically when the furnace temperature falls below a predetermined threshold. This temperature is usually in excess of 1500^oF at the location of the auxiliary burner (several feet above the grates). Prior to the advent of this design, high combustion efficiency was difficult to stabilize. Thus, in order to maintain adequate temperature in the furnace during startup, shutdown, and upsets, an auxiliary burner is a necessary design feature.

It is clear from the foregoing discussion why increased levels of emissions are generated during cold starts, upsets and shutdowns, but not exactly to what extent emissions are increased, or if such emissions are statistically significant. EPA has never required incinerator operators to record information (process conditions, emissions) continuously during startup, shutdown or upsets. Stack testing for dioxins is also never routinely conducted at these times, for compliance or other tests. Most of the studies available to the public have been done by governmental agencies, and the results, as pertain to off-normal combustion conditions, are summarized below.

The Peekskill incinerator in Westchester County, NY, consists of three mass-burn waterwall incinerators, each rated at 750 tons per day. Each has a transverse reciprocating grate made up of modular sections: the drying zone, two burning zones, and two finishing or burnout zones. Rates of underfire air and grate speed can be set for each zone. Overfire air is supplied through nozzles on the front and rear of each furnace.

A report from NYSERDA / NYSDEC conducted in the 1980s includes two test runs in which dioxin emissions were recorded during cold starts, as well as several under more normal operating conditions. This report was not written to examine cold starts in great detail. In fact, the report excluded the cold start runs from most of its analyses since analysis of variance (ANOVA) results showed that dioxin emissions during cold starts were statistically different (higher) than those emitted under normal operating conditions at a significance level of 0.0001 for both CDD and CDF. Run 4 was a normal cold start condition where the auxiliary gas burner was used to get the furnace up to normal operating temperature, the garbage was ignited, and the gas was turned off. This test sample was taken over a 65-minute period once the furnace was at "elevated temperature". Run 14 was research-oriented, in an attempt to determine if adding more natural gas than is usual would improve emissions during cold start. The report stated that the purpose of the cold start tests was observe the effect of feeding refuse to a furnace not at thermal equilibrium on CDD and CDF emissions. "These tests reflect continually changing operating conditions (non-steady state)."

The quantities of dioxin and furan generated during start up are striking. The results showed that the average CDD and CDF concentrations measured at the superheater exit (SE) during the first cold starts were two to three times higher than the average of the other 12 runs at steady operating conditions (124 ng/dscm of total CDD @7%O₂ vs. 11.4 to 84.3 ng/dscm for the normal runs). The temperature at the SE was 100 to 250^o less during the normal cold start (892^oF) than the other normal runs (995^oF – 1150^oF, with most over 1100^oF). Temperature was not recorded at the SE for run 14.

At the ESP inlet, the increase in dioxins from normal operations to normal cold start was between 18 and 51 times higher than normal conditions (7226 ng/dscm for the first cold start vs. a range of 43.8 to 209 ng/dscm for the other runs). This precipitous rise shows that secondary dioxin formation took place near the ESP inlet due to temperatures conducive to that process and the presence of dioxin precursors and catalysts. At the ESP outlet the increase was between 40 and 96 times, showing even more dioxin generation within the ESP. But what is of even greater interest is that the amount of secondary dioxin generated for the cold start run occurred at a rate between one and two orders of magnitude higher than for the normal combustion runs.

Why should the dioxin generation rate increase faster under cold start conditions than under steady-state conditions? There are two variables that different between the cold start and normal runs. The variable that likely distinguishes the normal cold start run from the others, and explains the result observed, is the furnace temperature, which is decidedly lower for the cold start. Since dioxin precursors (e.g., benzenes, phenols and chlorinated forms) are created in the furnace at the highest rate in the few hundred degrees below optimal furnace temperatures , the lower furnace temperature during cold start is likely to have caused a higher generation rate and a lower rate of destruction of dioxin precursors from within the furnace itself.

The other variable, the temperature at the ESP inlet, was 383^oF for the normal cold start run vs. 434 to 472^oF for the normal runs. Since research by Vogg and Steiglitz indicated the optimal temperature for secondary dioxin formation to be between 430^oF and 750^oF, peaking at 570^oF , it would seem that the normal runs would have greater secondary dioxin formation if temperature were the only variable. That there was nonetheless an enormous and larger secondary generation of dioxins during the cold start vs. the normal runs suggests that the greater generation rate during the cold start run of dioxin precursors (e.g., chlorobenzenes and chlorophenols), some of which require higher destruction temperatures (e.g., 1800^oF, vs. 1300^oF for dioxins), would seem to be the cause of the tremendously higher amount of dioxins in the flue gas further downstream. See Fig. 1.

Carbon Monoxide (CO), as one of the PICs that has been used as a surrogate for combustion efficiency and for dioxin/furan, was also measured for the various runs at Westchester. The mean CO concentration during the normal cold start was 180 ppmv at the superheater exit. The second, modified cold start had a mean of 114 ppm CO, and dioxin emissions nearly as high as the first, normal cold start. The CO levels for the 12 normal runs ranged from 6 ppm to 57 ppm at the superheater exit. The report indicated the mean CO level did not adequately characterize the range of CO experienced during the one-hour test. But EPA standards for existing MWCs specify averaging times of 4 hours for compliance with the CO emissions standard for four types of MWCs and 24 hours for compliance by four other types of MWCs, allowing much more opportunity for CO excursions to be masked by longer averaging times. (The CO standard for MWI’s for all plant types is 40 ppm, but over a 12-hour rolling average. Further, only one of the eight MWC plant types is required to meet a 50 ppmvv CO standard for existing plants. Averaging times are similar for new plants, and the range of CO emissions permitted is 50 to 150 ppmv, with only one of eight plant types being required to meet 50 ppmv.)

CO excursions are noteworthy in that they are associated with higher dioxin emissions, as shown in the next test program. A complicating factor is that every plant is allowed to exclude from Continuous Emissions Monitoring (CEMS) data 25% of the time the plant is operating from the reported data on a daily basis, and further exclude 10% of the calendar days per quarter. There is no stipulation regarding which data may be excluded, soit is possible (maybe likely) that the worst data will not be reported.

Fig. 1 Thermal Decomposition Characteristics of Selected Hazardous Hydrocarbons

The Pittsfield, MA facility consists of three 120 ton/day, two-stage, refractory lined incinerators with two waste heat boilers. Municipal solid waste (MSW) is burned under excess air conditions in the primary chamber; hot effluent gases pass into a secondary combustion chamber where any remaining uncombusted gases are burned.

Though no data were collected for startup or shutdown, the Pittsfield study included runs at different temperatures and oxygen levels to show how emissions varied when operating conditions were not optimized (i.e., upset conditions). The data showed that dioxin and furan emission in the flue gases were at a minimum when excess oxygen was between 9 and 11% in the hot zone. Total dioxin rose to over 50 ng/dscm corrected to 7% O₂ when excess oxygen was below 5% or above 12%. In fact, when the excess oxygen rose to over 11% the dioxins escalated quickly to over 100 ng/dscm and beyond.

In addition, a clear pattern was found with respect to furnace temperature impacts on dioxin. The optimal temperature window for Pittsfield, measured at the tertiary duct, some distance from primary combustion, was roughly between 1500 and 1650^oF. Dioxins increased to over 120 ng/dscm when the temperature fell below 1300^oF. The dioxin concentrations for the two low temperature runs (1300^oF and 1350^oF) were much higher than those for all other runs by a factor of more than four, and they were statistically different from emissions under normal operating conditions. The two low furnace temperature runs (1300^oF and 1350^oF) also produced CO levels that were more than a factor of 10 higher than the rest of the test runs, showing CO as a useful indicator in this case.

As important as the level of CO emissions in a MWC is, an equally important issue is the averaging time over which these emissions are evaluated. It is important to note, in this regard, that the ASME/NYSERDA Pittsfield combustion tests showed that CO levels above 100 ppm were associated with a greater certainty of higher dioxin levels. If new and existing incinerators are permitted to exceed this 100 ppm level routinely, by virtue of a 4- or 24-hour averaging time and a limit of 100 ppm or more, then the regulations are not designed to minimize dioxin emissions in these incinerators. The Pittsfield research demonstrates the importance of minimizing the number, intensity, and duration of CO spikes, and thus, of limiting the length of the averaging period for CO. Thus, to minimize the opportunity for formation of PICs, it is necessary to require an average limit for CO that would reflect a strict limitation on the frequency, intensity, and duration of excursions.

The Prince Edward Island (PEI) facility consists of three, two-stage incinerators, each rated at about 36 tons per day. The incinerator design uses controlled or "starved" air combustion (as contrasted with the excess air operations used at Pittsfield and Westchester). MSW is burned in the primary chamber, where a fraction of the total air needed for complete combustion is provided. The combustible gases enter the secondary chamber, where pre-heated air is added to complete combustion.

During testing the primary chamber temperature was maintained at a relatively constant 1292^oF within +100^oF, except for the low temperature test, where it was maintained at 1250^oF. The secondary chamber temperatures were kept at 1550^oF for two tests, at 1900^oF for the high temperature test, and 1350^oF for the low temperature test. The percent excess air also differed by about 40% between the tests involving normal and low secondary chamber temperatures. The PEI data showed a tendency for total dioxin concentrations to increase with increasing excess oxygen concentrations, which occurs in conjunction with lower furnace temperature. This is in agreement with the Pittsfield data. See Fig. 2.

Fig. 2. The influence of Excess air on CDD/CDF Emissions at Pittsfield, Quebec City and PEI

A comparative study of test results from Pittsfield, Westchester, and PEI in the above figure indicated that levels of dioxin and furan in the flue gas entering a pollution control device are affected by plant operating conditions if the conditions deviate sufficiently from normal operation. This study also indicated that furnace temperature can be used as a gross indicator of total dioxin and furan emissions, and that operating an incinerator at excess oxygen levels below about 5% may cause an increase in dioxin and furan emissions.

The Quebec City mass burn incinerator includes four incinerators/boilers rated at 227 tonnes per day each, each with a vibrating feeder-hopper, drying/burning/burnout grates, refractory lined lower burning zone, and waterwalled upper burning zone.

The second of two studies of this incinerator conducted by Environment Canada evaluated the combustor performance at a variety of operating conditions: low, medium and high load, percent excess air, furnace temperature, and primary/secondary air ratio. Some of these conditions were characterized as very poor (where primary/secondary air ratio was 90/10 and excess air was considered high (115%)), and poor (where furnace temperature was 850oC, excess air was very high at 130%, and primary/secondary air ratio was 60/40. Three other combinations, under low, design, and high load, were considered to be "good" operations.

Dioxin and furan emissions were measured for each of these test combinations, and statistical analysis of the data showed "a fairly strong correlation" between high excess air levels and dioxin/furan. This is shown in Fig. 2.

The best single parameter correlation (r² = 0.876) was a comparison of uncontrolled particulate matter entering the ESP versus dioxin/furan in the stack. Two other variables with extremely good single parameter fits were flue gas flow rate (r² = 0.771) and primary air flow rate (r² = 0.723). The rate of increase in dioxin/furan becomes exponential about around 125% excess air, indicating a move towards upset conditions.

Waste load also affected dioxin (see Fig. 3). More than 110% of optimal load resulted in a rapid increase in total CDD/CDF (from 100 to over 200 ng/dscm 7%O₂). Load less than 90% resulted in a more moderate increase in CDD/CDF (from 100 ng/dscm at 90% load to over 200 ng/dscm at 70% load).

Fig. 3. Effect of waste load on dioxin emissions at Quebec City

The Oswego facility consists of four, two-stage mass-burn units each rated at 50 tons/day. Batch loads are fed to the primary chamber (PC) and moved through the chamber along the stepped bottom by air-cooled transfer rams on a cycle of approximately seven to eight minutes. Combustible gases and entrained particulates exit the primary chamber to the secondary chamber (SC) where the flue gas is mixed with preheated secondary air to complete combustion.of unburned gases and particulate matter.

The Oswego study, conducted by New York State agencies, compared dioxin and furan generation in groups of three runs under each of four conditions:

1. clean combustor, right after startup (Start of Campaign), at which time the secondary chamber (SC) was 1837 to 1875^oF,
2. dirty combustor, ready for maintenance shutdown (End of Campaign), at which the SC was 1817 to 1834^oF,
3. the Mid-Range Secondary Chamber Temperature ranged from 1738 to 1752^oF, and
4. the Low Secondary Chamber Temperature runs ranged from 1617 to 1634^oF. (The temperatures at the secondary chamber exit were lower from two to four hundred degrees (the low temperature SC exit was 1336^oF)).

At the SC exit, the low furnace temperature condition (1617^oF to 1634^oF furnace temperature, and 1336^oF at the SC exit) showed that total dioxin emissions ranged from 67.7 to 110.1 ng/dscm @ 7%O_2, averaging 84.5 ng/dscm @ 7%O₂,, increasing by a factor of six from the normal temperature condition (1738^oF to 1752^oF furnace temperature), an average dioxin emission of 13.6 ng/dscm. At the ESP Inlet, total dioxins at normal condition, 53.4 ng/dscm, increased by the same factor of six to an average of 289.9 ng/dscm for the low SC temperature. The means for the dioxin emissions for the low temperature runs were found to be statistically different and significantly correlated. Using one-way ANOVA, the chance that the dioxin means were not different for the different furnace conditions was 0.0049, measured at the SC exit, and even lower, 0.0001, measured at the ESP Inlet. Thus, the lower furnace temperature tested here was associated with a six-fold increase in dioxins.

Dioxin emissions were also correlated with CO emissions. Since CO measurements were taken continuously using a CEMS, different assumptions could be made about representation of CO as single value representations (SVRs). The correlation between total dioxins and CO at the SC exit that was most pronounced was at the 90^th percentile CO value, where r = 0.921, a nearly perfect positive correlation. In general the SVRs representing extreme values of CO (i.e., 90^th, 95^th and 99^th percentile) correlated most frequently with the dioxin and furan levels measured, indicating that it is frequency and duration of the highest values of CO that best predict changes in dioxin and furan flue gas concentrations at this facility. All relationships were significant at the 0.05 level. These data highlight the importance of avoiding CO spikes, and of accurately recording the amount of time during which elevated CO occurs, not simply averaging CO over long periods of time to mask such excursions.

The Mid-Connecticut facility burns Refuse-derived Fuel in 3 RDF-fired spreader-stoker boilers that process a total of 2000 tons per day of MSW. Four pneumatic distributors spread the RDF across the width of the combustion grate. Ten underfire air zones allow the operator to optimize combustion and to respond quickly to "piling" situations by manual adjustment of underfire air dampers. The overfire air system is equipped with four tangential assemblies, located in the furnace corners; each assembly includes three levels that are separately controlled to encourage vortex formation, providing longer residence times for combustion gases. Preheated combustion air enters the furnace forming a vortex, providing longer residence times for the combustion gases.

A major goal of the project was to determine generation of trace organics and metals in the furnace under different process operating conditions, not under "upset" conditions, per se. Nonetheless, the results are instructive in this analysis. Steam flow rate (an indicator of load) and combustion air flow rates / distributions were the primary independent variables defining operating conditions: "good", "poor", "very poor". However, as compared with the above studies, the variation in combustion conditions was smaller, as this study was focused on a more efficient range of operation than these other studies. Dioxins, furans, CO, total hydrocarbons, PCB, Cholorobenzenes, Chlorophenols, and PAHs were measured.

Multiple regression models showed the effect of various continuously monitored emission and process parameters on dioxin emissions (Prediction Models) and the effect of various combustion control measures on dioxin emissions (Control Models). The best prediction model showed that CO, NOx, moisture in the flue gas at the spray dryer inlet, and furnace temperature explained 93% of the variation in uncontrolled dioxin emissions, with CO explaining 79% of the variation by itself. The best control model showed that RDF moisture, rear wall overfire air, underfire air flow, and total air explained 67% of the variation in uncontrolled dioxin emissions.

Since CO was found to be such a strong predictor of dioxin emissions, the relationship was explored further. It was found that the percent of time the CO level was over 400 ppm was quite strongly correlated to the amount of uncontrolled dioxins generated, particularly when examining only those runs where there was poor combustion. "Poor combustion implies that greater amounts of organic material escape the combustor unburned. In the correlation between CO and PCDD/PCDF, use of only the poor combustion tests would improve R² from 0.70 to 0.95." The correlation between total hydrocarbons and dioxin/furan improved from a R² of 0.68 to 0.97 for the poor combustion tests. This means that for the poor combustion tests, 95% of the change in dioxin/furan values are explained by the change in CO emissions. This is consistent with the theory that during periods of incomplete combustion the amount of organic matter escaping the furnace strongly influences the formation of PCDD/PCDF. Thus, these data again provide support for the idea that CO is an important surrogate for dioxin, and that allowing longer averaging times for CO levels for compliance with standards will more likely result in higher dioxin/furan emissions since, under these conditions, more CO spikes can occur without exceeding standards.

Examining the high load runs from a different perspective, the furnace temperature was 74^oC less (976 vs 1049^oC) for "poor" vs. "good" combustion conditions under high load. Under "poor" combustion conditions, total dioxin emissions at the spray dryer inlet were 317 ng/Sm³, but were 67 ng/Sm³ @ 12%CO₂ under "good" combustion conditions, a factor of almost five. The same relationship was true for furans, total hydrocarbons, PAHs, chlorobenzenes and chlorophenols. CO was 397 ppm for "poor" combustion and 116 ppm for "good" combustion under high load. These data show how important furnace temperature is to dioxin emissions under high load conditions.

An increase in underfire / overfire ratio was also found to be an indicator of increased PIC emissions. Under intermediate load, this ratio was 0.923 under "good" conditions and 1.632 under "very poor" conditions. This change resulted in an increase in CO from 93 to 903 ppm ng/Sm³ @ 12%CO_2., At this load total hydrocarbons increased from 2.5 to 52.4, PAH from 7,330 to 112,000, chlorophenols from 14,300 to 114,000, chlorobenezenes from 6,050 to 15,800, dioxins from 228 to 580, and furans from 579 to 1,280, all in ng/Sm³ @ CO₂.

In addition to the foregoing factors associated with combustion efficiency (i.e., adequate quantity of air, proper oxygen distribution above and below the grate, optimal circulation of oxygen, residence time of combustion gases at high temperature, and correct temperature distribution across the furnace, etc...) the reports show that CO level can be an indicator of combustion efficiency. In a 1989 policy document, USEPA stated that the first goal of good combustion practice (maximization of in-furnace destruction of trace organics) would be accomplished by optimizing waste feeding procedures, achieving adequate combustion temperatures, providing the proper amount and distribution of combustion air, and optimizing the mixing process. A failure in any one of these components would be accompanied by spikes or bulk increases in flue gas CO concentrations. Further, EPA states, "Failure to achieve the necessary temperatures and residence times will result in the escape of organics from the furnace, which will lead to elevated concentrations of CO in flue gases", "CO emissions typically increase when insufficient O₂ is available to complete combustion", and "Failure to distribute combustion air in the correct proportions to primary and secondary supplies can result in elevated organics and CO emissions". Aside from mentioning these factors, EPA did not include combustion requirements other than CO in its 1991 or 1995 MWC emissions regulations, and the requirements for CO permit higher emissions and longer averaging times than the above studies would suggest.

Looking further into EPA’s and other governments’ guidance on good combustion practice, EPA, in 1987 issued its first good combustion practice guidelines, which flatly stipulated a single CO limit, an O₂ limit, and other firm limitations that must be met by all plant combustor designs. This guidance, advising the states regarding good combustion practices, indicated that carbon monoxide emissions of 50 ppm for all plants, over a four-hour averaging time, along with 6-12% oxygen after combustion, among other requirements, were indicators of good combustion practice in a municipal waste combustor. EPA based these stipulations on information relating combustion efficiency and emissions of PICs to a number of incinerator performance criteria. Environment Canada also later adopted a 50 ppm CO standard in its standards for good combustion. Regulations issued by the Netherlands have been even more stringent at 44 ppm (corrected from 50 mg/Nm³).

EPA’s original recommendation for a CO limitation of 50 ppm during cold starts and other upset conditions seems warranted insofar as the Westchester cold start and other aforementioned upset data are concerned. Dioxin emissions in the thousands of nanograms (plus thousands more for furans) for both Westchester cold start runs correlated with a CO concentration of 180 ppmv during the normal cold start run and with 114 during the modified cold start. These results would call into question EPA’s later standards for CO which range from 50 ppm for certain new units to 250 ppm for certain older units.

Since temperature is an important factor governing operations, good combustion practices for all incinerators should require not only a minimum temperature of 1800^oF across the furnace to ensure destruction of PICs as was recommended by EPA in 1989, but also a maximum temperature of 2000^oF at fully mixed height to minimize formation of NOx. Further, since it is during times of startup, shutdown, and upsets that extremely high PIC emissions can occur, it is especially important for operators to be careful, ensuring that the auxiliary burner is maintained and set at the correct setpoint, and that the quantity/direction/location of oxygen supplied is correct at all times.

The data presented for these six very different plants show that it is precisely during startup, upset, and presumably shutdown, that the emissions of dioxins and other products of incomplete combustion and certain other pollutants are at their highest, and often many times higher than emissions during "normal" operations. These data analyses have a number of implications for incinerator regulations:

From a design standpoint, these analyses suggest that very small incinerators, that do not operate on a 24-hour per day basis due to insufficient waste, and which must start up and shut down sometimes on a daily basis or more, would not be capable of good combustion practice. Similarly, batch-fed incinerators are designed to require non-steady state operating practices (e.g., erratic temperature and oxygen regimes in the furnace due to repeated opening and closing of the charging door), and as such, would not be capable of good combustion practice. MWI’s, with intermittent charging design, typically have a 12-hour cycle, having two startups and shutdowns per day with poor combustion efficiency during those times. (Such units also have manual ash removal, which exposes workers to fugitive ash to a greater extent than with the more modern design involving continuous charging and automatic ash removal.)

These data also point out the likelihood that plants that don't have auxiliary burners or automatic combustion controls are prone to upset conditions and higher emissions of PICs. They also point out the benefit of using CEMS to measure combustion-related parameters (O₂, CO, furnace temperature, etc…), and of reporting it to the regulatory authorities and the public via a time-sensitive method (e.g., continuous telemetering to an internet website), so that problems can be recognized and rectified quickly.

The information on cold starts and upsets presented here would suggest that small plants, batch-fed and intermittent units, due to intractable aspects of their design, would be prone to cold starts and/or upsets, and might be candidates for phase-out, to be replaced by continuously operating units. It would behoove EPA to revisit its regulations regarding continued operation of such small incinerators.

The data presented above suggest that frequent startup-shutdown cycles can possibly produce even more emissions than occasional upsets. EPA has demonstrated in its regulations a lack of interest, in knowing how often each incinerator is shut off and brought back online, and what the consequences of that might be on emissions. In addition to possibly abolishing batch-fed units designed for intermittent operation, the data point to the desirability of EPA to establishing strict limits on the number of startup-shutdown cycles per month on any incinerator.

The data also indicate a need for reporting on frequency, duration, and severity of start-ups, shutdowns, and upsets with CEMS data on combustion parameters collected for each upset and detailed stack sampling for dioxins/furans to determine the cause. If it is the case that one startup/shutdown cycle introduces many times the amount of emissions during one or two hour's time vs. "normal" operating conditions, these can overwhelm the quantity of "normal" emissions during longer time periods. Since quantified data characterizing all startups, shutdowns, and upsets have not been included in reporting requirements, the total emissions over time have been underestimated. If this is not rectified, future national estimates of PIC emissions will continue to be undercounted and estimates of environmental and health risk from incinerators, or incineration in general, will be underestimated.

The data also support the idea that CO is an important surrogate for dioxin (higher CO usually means higher dioxin emissions. But EPA has established averaging times for CO that are different for different types, ages and sizes of incinerators, such that the current range of MACT Standards for CO are currently 50 to 250 ppm. Permitting some plant designs to emit higher levels of CO than others is not only unfair to the better plants, but encourages the design and installation of inferior incinerator designs. The multiple CO limits and averaging times, depending on type of combustor, have not changed significantly from 1991. Longer averaging times of 4 to 24 hours for CO allow instantaneous emissions to reach higher levels (i.e. higher spikes) without detection than they would were shorter averaging times required, and ensures a greater length and severity of CO excursions than were a uniform shorter averaging time required. Longer averaging times also discourage incinerator operators from working to reduce excursions, since their severity is masked. Thus, shorter averaging times for CO (an hour or less, or every few minutes as is the case for opacity) are indicated. Based on the information presented above, a limitation of 50 ppm, the level EPA stipulated in its original guidance to states regarding good combustion practices for all plants and an averaging time of an hour or less, is supported as a level reflecting good combustion practice.

To further minimize upsets, future regulations could also codify earlier EPA guidance on combustion operations and require all plants to optimize oxygen (6 to 12%), temperature (1800-2000^oF), and other factors such as underfire/overfire ratio. Since there is still no numerical maximum flue gas temperature at the particulate matter control device inlet, designed to minimize emissions of dioxins (via enhanced cooling of flue gases), it would make sense to promulgate one. In order to keep the public and regulators informed about current emissions and combustion, regulations should require immediate reporting of CEMS data (e.g., via continuous telemetering to the internet), so that problems can be recognized and rectified quickly.

EPA should codify its earlier guidance on combustion operations and require all plants to optimize oxygen (6 to 12%), temperature (1800-2000^oF), and other factors such as underfire/overfire ratio on a plant-by-plant basis. All data and data summaries should be available in real time on the internet so that neighboring residents as well as regulatory authorities can be informed of plant performance in a timely manner.

While it might be argued that more modern plants with emissions controls will remove emissions generated as a result of cold starts, upsets and shutdowns, it cannot be guaranteed that these emissions control devices are 100% effective in removing PICs or that they are even operated optimally at all times. EPA’s MACT regulations do not require optimization of operations either of the emissions control devices or of the combustion process (e.g., by requiring limitations on temperature, oxygen, etc… for the purpose of minimizing emissions). Furthermore, though continuous emissions and process monitoring is required by regulation, every plant is allowed to exclude 25% of the time the plant is operating from the reported data on a daily basis, and further exclude 10% of the calendar days per quarter while the plant is operating. There is no stipulation in the standards and guidelines regarding which data may or may not be excluded, therefore one can assume that the worst data will not be reported. Also, in EPA’s regulations, emissions during startup, shutdown, and upsets are specifically excluded from being considered insofar as compliance is concerned.

If it is the case that one startup/shutdown cycle introduces several or many times the amount of emissions during one or two hour's time vs. "normal" operating conditions, these can overwhelm the quantity of "normal" emissions during longer time periods. Based on the analysis presented here, these data exclusions lead to a positively biased picture of plant performance over time. Further, as a result of the exclusions and since EPA does not require optimization, there is little incentive for operators to optimize either emissions control device removal efficiencies for PICs or combustor operations to minimize creation of PICs, even if improved performance were possible. To make this unavailability of data worse, exceedances need not be reported immediately. Also, if small MWC’s meet the emissions criteria for dioxin/furan, (as well as mercury, lead, and cadmium), they are then permitted to skip conducting performance tests two of every three years, and to submit simplified annual reports for these two years. The smallest plants, such as apartment incinerators, which have the least advanced designs and operations, are not covered at all by EPA’s regulations.

Current practice of allowing plants to avoid reporting requirements for years at a time does not encourage continued high level performance, and should be eliminated. Exclusion of the smallest plants invites construction of more small plants and allows localized sources of uncontrolled emissions (sometimes close to population receptors, as in apartment buildings). To get an accurate picture of plant performance at all times, there is a need to eliminate the various exclusion periods for compliance, and require collection and dissemination to the public of CEMS data recorded at all times at all plants, particularly at those times when emissions are likely to be highest (i.e., during startup, shutdown and upsets). In addition, plant performance would likely increase if the standards required optimization of combustion and emissions controls, establishment of strict limits on the number of startup-shutdown cycles per month, and requirements to report start up and shut down cycles for all units.

23 EPA Fact Sheets on Subpart Eb and Cb 40 CFR Part 60.

Keywords: incineration, combustion efficiency, upset and off-normal conditions, dioxin emissions