Subject:  Technical Support Document  -  Oldcastle Trident Federal Implementation Plan Revision 
Date:  March 8, 2017
From:  Aaron Worstell, Environmental Engineer
To:  Docket EPA-RO8-OAR-2017-0851
      In 2012, EPA finalized the Regional Haze Federal Implementation Plan (FIP) for the State of Montana.  Among other things, the FIP established a Best Available Retrofit Technology (BART) NOx emission limit for the kiln at the Trident cement plant (previously owned by owned by Holcim, Inc.). The BART emission limit for the Trident kiln of 6.5 lb NOx/ton clinker (as a 30-day rolling average) reflected installation of Selective Non-Catalytic Reduction (SNCR).  Based on information available at the time, the emission limit was derived using a 50% reduction in the baseline NOx emissions.  
      In September of 2016, the new owner of the Trident facility, Oldcastle Materials Cement Holding, Inc., requested that EPA revise the emission limit due to their concerns that the emission limit could not be met on a continuous basis without unacceptable levels of ammonia slip, that may in turn negatively impact operations, unduly increase reagent costs, and create a localized detached  plume.  Accordingly, the purpose of this technical support document (TSD) is to evaluate the NOx control effectiveness, and thereby emission limit, of SNCR when applied to long kilns.  In particular, in this TSD, EPA considers new information concerning SNCR performance that was not available at the time the 2012 FIP was promulgated.
      As an initial matter, EPA recognizes that it is difficult to predict the control effectiveness of SNCR for long cement kilns for a few reasons.  First, whereas SNCR has been applied to a many industrial sources, and in particular to coal-fired utility and industrial boilers, the number of long cement kilns that have been retrofitted with SNCR is relatively small.  In fact, until recently SNCR was not considered technically feasible for long kilns because the appropriate temperature window is in the middle of the kiln, requiring that the reagent be injected into the rotating kiln.  Second, and particularly in comparison to pre-heater/pre-calciner kilns, there is inherent variability in the operation of long kilns that make injection of reagent at the optimal temperature window difficult.  Third, the SNCR performance data for long kilns that is available does not reflect a contemporaneous measurement of uncontrolled and controlled NOx emission rates because it is not possible to measure the uncontrolled NOx emission rate inside the kiln.  Instead, the uncontrolled NOx emission rate, measured at the kiln exhaust, is taken from a baseline period prior to the installation of SNCR.  Thus, it is difficult to prospectively estimate the control effectiveness of one long kiln from the operation of another long kiln already equipped with SNCR.  Collectively, these factors introduce uncertainty when predicting the control effectiveness of SNCR when applied to long kilns, which is a necessary step in setting the NOx emission limit. This uncertainty has been the impetus for the use of post-installation control technology demonstrations to set emission limits in association with consent decree enforcement actions for long kilns (as discussed below). Also, in this TSD, we evaluate the baseline emission rate from which the SNCR-based emission limit is calculated for the Trident kiln.
      As stated in their submittals to EPA, Oldcastle is committed to installing and operating the SNCR on its Trident kiln.   The construction of the SNCR is underway and will likely be integrated into plant operations beginning during a shutdown scheduled for April of 2017.   As such, EPA's consideration of Oldcastle's concerns and the resulting proposed FIP revision for the Trident kiln address only the appropriate emission limit associated with the operation of SNCR. Because EPA is not revisiting the question of what control technology represents BART, this TSD does not include an updated five-factor BART analysis.
 EPA's Evaluation of SNCR Control Effectiveness in the 2012 FIP
In the 2012 FIP, EPA determined that a 50% control effectiveness was an appropriate estimate for SNCR at long kilns, such as the Trident kiln.  This was largely based on the SNCR performance observed on the three Ash Grove Cement long wet kilns located in Midlothian, Texas.  Emissions data submitted by Ash Grove to the Texas Commission on Environmental Quality (TCEQ) showed that the Midlothian kilns achieved emission rates in the range of 1.6 to 2.9 lb NOx/ton of clinker from June through August 2008 when using SNCR.  EPA compared this to baseline emissions data for the same three-month period in 2006.  Table 1 summarizes the 2006 and 2008 emissions rates, and associated percent reductions, that EPA used in support of the 2012 FIP.  
  Table 1.  Ash Grove Midlothian Monthly NOx Emissions, June through August, 2006-2008 
 
            June through August 2006 Emission Rate (lb/ton clinker) 
            June through August 2008 Emission Rate (lb/ton clinker) 
                                  Percentage 
                                  Reduction 
                                     (%) 

June  
July  
August  
Average 
June  
July  
August  
Average 

Kiln 1 
                                     5.2 
                                     5.0 
                                     4.5 
                                     4.9 
                                     1.7 
                                     1.6 
                                     2.2 
                                     1.8 
                                     62.5 
Kiln 2 
                                     5.0 
                                     4.1 
                                     3.9 
                                     4.4 
                                     2.7 
                                     2.6 
                                     2.8 
                                     2.7 
                                     37.7 
Kiln 3 
                                     5.0 
                                     4.4 
                                     4.2 
                                     4.5 
                                     2.9 
                                     2.6 
                                     2.5 
                                     2.7 
                                     40.5 

When the control effectiveness values (percent reduction) on all three kilns were averaged together, EPA found that SNCR achieved a 47.5% reduction in NOx.  
		During the public comment period for the 2012 FIP, commenters questioned the usefulness of the Midlothian data in setting emission limits for other long kilns.  Oldcastle has repeated some of those concerns in its recent submittals to EPA that request a less stringent emission limit.  In particular, the previous commenters, and now Oldcastle, pointed to the fact the Midlothian NOx emission rates (in lb/ton clinker) in subsequent years (2009 and 2010) were much higher than in 2008.  In response to comments on the 2012 FIP, we suggested that these higher NOx emission rates indicated that SNCR was not utilized to the fullest extent in 2009 and 2010 and thus were not representative of the potential control efficiency of SNCR.  Based on information recently obtained from Ash Grove Cement, we have been able to confirm that SNCR was underutilized.in those two years.  In 2008, while the Midlothian kilns were not yet subject to a NOx emission limit associated with the operation of SNCR, Ash Grove operated SNCR on the three kilns in order to understand how the control technology would work in preparation for upcoming emission requirements.  Then, beginning in 2009, the Midlothian facility was required to comply with a facility-wide SIP emission limit of 4.41 tons NOx/day during the ozone season. Also, demand for cement was low during 2009 and 2010.  As a result, Ash Grove was often able to meet the facility-wide emission limit with limited use of SNCR because one or more of the kilns was idle.  For example, in 2009, SNCR was only operated for 131, 1,051, and 142 hours, respectively on Kilns 1, 2, and 3.   Subsequently, starting in March 2011, in accordance with a settlement agreement, the Midlothian kilns were individually required to comply with a 30-day rolling average emission limit of 3.6 lb/ton clinker at all times throughout the year.  Consequently, and despite higher demand for cement, NOx emissions (in lb/ton clinker) dropped significantly when compared to 2009 and 2010.  Therefore, the SNCR performance data for Midlothian considered by EPA during the 2012 FIP development (2006 - 2008 data) was reliable and remains informative in setting a BART emission limit for the Trident kiln.  Regardless, as noted below, the EPA is now in possession of additional SNCR performance data for long kilns obtained through consent decree control technology demonstrations.  This more recent SNCR performance data, along with earlier data from the Midlothian kilns, has been used to inform the SNCR performance expectations for the Trident kiln.      
 Recent Control Effectiveness Data for Long Kilns Obtained through Consent Decree Control Technology Demonstrations
      Next we consider recent information regarding control effectiveness for long kilns obtained through consent decree control technology demonstrations. Since promulgation of our 2012 FIP, SNCR has been installed on a number of wet or dry long kilns in association with consent decree enforcement actions. SNCR has been installed on six long kilns (two wet, four dry) owned by LaFarge North America Inc., and on an additional long wet kiln owned by the Ash Grove Cement Company.  The Ash Grove kiln is the Montana City kiln for which EPA had earlier established a BART emission limit of 8.0 lb NOx/ton clinker (30-day rolling average) in our 2012 FIP.  
		Each of the kilns subject to a consent decree was required to establish an SNCR-based emission limit through a control technology demonstration. The demonstrations were designed to establish the optimal performance of SNCR, and were carried out through a number of steps, including design report, baseline period, optimization period, and demonstration period.  The control effectiveness data for these kilns, along with the data from the 2012 FIP for the three Midlothian kilns, is summarized in the attachment to this TSD.  See Attachment 1, Summary of SNCR Performance Data for Long Cement Kilns.  The control effectiveness shown for the kilns subject to consent decrees is highly variable and ranges from 29% to 47%, with a mean of 40%.  This control effectiveness reflects the percent reduction in the NOx emissions between the baseline and demonstration periods.  As noted earlier, it does not reflect contemporaneous NOx measurements.  These values compare favorably to the range of reductions (3-month average) observed for three Midlothian kilns of 37.7% to 62.5%, although the latter are somewhat higher.
		The kiln that is most comparable to the Oldcastle Trident kiln is the Ash Grove Montana City kiln because both are long wet kilns and operate in similar environments.  As such, it is reasonable to conclude that the Montana City kiln and the Oldcastle Trident kiln should be able to achieve comparable levels of NOx reduction per mole of uncontrolled NOx to injected reagent, i.e., at a given molar ratio (NOx:NH3).  During the baseline period of the control technology demonstration for Ash Grove Montana City, lasting approximately 6 months between March and August 2014, the kiln emitted NOx at a rate of 11.6 lb /ton clinker.  Following optimization of the SNCR system, the kiln emitted NOx at a rate of 7.0 lb/ton clinker over a period of approximately 10 months between July 2015 and April 2016.  Again, this reflects an emission reduction between the two periods of roughly 40% based on the use of SNCR.  Subsequently, as required by the consent decree, Ash Grove proposed, and EPA approved, a 30-day rolling average emission limit of 7.5 lb NOx/ton clinker, which is lower than the BART emission limit of 8.0 lb NOx/ton clinker. The 7.5 lb NOx/ton clinker emission limit was approved by the EPA on December 29, 2016.
	 	It is of particular importance that the SNCR installed at the Montana City kiln was ultimately optimized around ammonia slip.  The ammonia slip is the concentration of unreacted ammonia as measured at the kiln exhaust that is above the background concentration established during the baseline period. Initially, the optimization of the kiln proceeded "by injecting [increasing] set amounts of ammonia based on the estimated molar ratio of ammonia to the NOx emission rate identified during the baseline period."  However, this approach at times resulted in high levels of ammonia slip and an objectionable detached plume that was visible in the immediate vicinity of the facility.   In response, the optimization was then conducted based on an ammonia slip at the kiln exhaust of 10 ppm.  This too at times resulted in a detached plume.  Therefore, Ash Grove ultimately optimized the operation of SNCR around an ammonia slip of 5 ppm.  Ash Grove observed that "[r]educing ammonia slip from 10 ppm to 5 ppm did not significantly reduce the effectiveness of the SNCR system, as the average daily NOx emission rate during the 14-day period of 5 ppm ammonia slip was 6.4 lb/ton clinker and the maximum daily NOx emission rate was 7.3 lb/ton clinker."  The ammonia slip during the demonstration period that followed was then set to a target of 5 ppm, and Ash Grove demonstrated the ability to meet an emission limit of 7.5 lb/ton clinker (30-day rolling average) with this amount of ammonia slip.  This approach to optimization established that a control effectiveness of 40% can be reached while addressing the same concerns about excess ammonia that Oldcastle raised in relation to the Trident kiln.  
		In consideration of the entirety of the SNCR performance results for long kilns now available to EPA, and in particular that for the similar Ash Grove Montana City kiln, it is appropriate that the emission limit for the Trident kiln reflect a control effectiveness of 40%.  
 Baseline Emissions
In our 2012 FIP, EPA described the process of setting the BART emission limit for the Trident kiln as follows:
   In recalculating our proposed BART emission limit for NOx, we continue to rely on the estimate of baseline NOx emissions in lb/ton clinker provided in Holcim's 2012 submittal, cited in our proposal at 77 FR 24018, footnote 93.  That submittal listed a 99th percentile 30-day rolling average NOx emission rate of 12.6 lb/ton clinker, for the period 2008 - 2011. Applying a 50% reduction to the 99[th] percentile figure yields 6.3 lb/ton clinker. To allow for a sufficient margin of compliance for a 30-day rolling average limit that would apply at all times, including startup, shutdown and malfunction (as explained in our proposal at 77 FR 24018), we are setting the BART limit at 6.5 lb/ton clinker in our 2012 FIP.
At the request of EPA, Oldcastle submitted updated 30-day rolling average emissions data for the period of 2008 through 2016.  EPA has evaluated this data in order to determine whether the baseline value of 12.6 lb NOx/ton clinker used in our 2012 FIP remains a reasonable baseline for the purpose of setting the BART emission limit.  The 99[th] percentile 30-day rolling average from the 9-year period is 13.9 lb NOx/ton clinker.  In their, February 13, 2017 submittal to EPA, Oldcastle (through Bison Engineering, Inc.) proposed that this baseline value be used to calculate the BART emission limit.  However, this baseline value is the result of a short period of unusually high daily NOx emissions that occurred on various days between September and November of 2012.  Oldcastle stated that one likely cause of the high NOx emissions during this time period was the result of ash ring buildup inside the kiln. Oldcastle also noted that "ash ring build-up is a well-known problem that can develop in cement and lime kilns," and it can "disrupt normal kiln mixing and heat transfer and can degrade fuel efficiency, effects that would tend to increase NOx emissions on a per-ton of production basis."  Oldcastle also advocated that the high NOx emissions in late 2012 should be included when calculating the 99[th] percentile 30-day rolling average baseline emission rate because, though the emissions are atypical, they nonetheless represent operating conditions that may be anticipated in the future.  However, when compared to the emissions for the 9-year period as a whole, the emissions during late 2012 appear to reflect exceptional circumstances.  This is depicted graphically in the chart included in Attachment 2 to this TSD, showing that the emissions in late 2012 were far higher than any other period.  Indeed, in the 4-year period that followed, 2013-2016, the 99[th] percentile 30-day rolling average was been identical to that used in the 2012 FIP (i.e., 12.6 lb NOx/ton clinker).   In essence, the emissions in late 2012 represent an upset condition that should not be considered when calculating the baseline emissions.  Moreover, the original emissions data from 2008-2011, together with that for 2013-2016, yields 8 years of data; this is more than sufficient for establishing the baseline.  Thus, EPA concludes that an emission rate of 13.9 lb NOx/ton clinker does not represent a realistic emissions baseline.  
Moreover, immediately after the ash ring buildup, the daily emissions data shows that Oldcastle did not operate the kiln between November 27[th] and December 1[st] 2012.  Presumably, during this 5-day shutdown period, Oldcastle took corrective measures to remove the ash rings from the kiln, thereby returning the kiln to normal operation.  Emissions levels returned to typical levels immediately following the shutdown.  Also, background information shared by Oldcastle indicates that proper kiln operation and maintenance can help to prevent ash ring formation.  Thus, it is within Oldcastle's control to prevent ash ring formation, or at the very least, to promptly take corrective action when it does occur.  Given that compliance with the BART emission limit is assessed over a 30-day rolling period, Oldcastle is able to anticipate whether high short-term NOx emissions due to ash ring deposits may lead to non-compliance with the BART emission limit.  In such case, Oldcastle is able to take the appropriate operation and maintenance measures, and if necessary, shut down the kiln to remove the ash deposits  -  an action that they must inevitably take in any case to return to the kiln to efficient operation. 
      Finally, we note that the Oldcastle Trident and Ash Grove Montana City kilns have very similar NOx baseline emissions (pre-SNCR) when viewed as the 99[th] percentile 30-day rolling average. Baseline data collected for the Montana City kiln between March and August 2014 in association with the control technology demonstration shows that the 99[th] percentile 30-day rolling average emission rate was 12.8 lb NOx/ton clinker.  Though this baseline data was collected over a much shorter time than that for the Trident kiln, it is nearly equal to the value for Trident of 12.6 lb NOx/ton clinker. This is another indication that the two kilns should be able to achieve similar levels of controlled NOx emissions with SNCR.
 Conclusion
   In consideration of the entirety of the SNCR performance results for long kilns now available to EPA, and in particular, that for the similar Ash Grove Montana City kiln, it is appropriate that the emission limit for the Trident kiln reflect a control effectiveness of 40%.  In addition, in consideration of the 9-years of baseline data from 2008 through 2016, it is appropriate to retain the original baseline used in the 2012 FIP of 12.6 lb NOx/ton clinker (99[th] percentile 30-day rolling average).  Applying the 40% control effectiveness to this baseline emission rate yields a value of 7.6 lb NOx/ton clinker.  This compares very favorably to the emission limit of 7.5 lb NOx/ton clinker set through a control technology demonstration for Ash Grove Montana City, particularly given that two kilns have very similar baseline emissions (as 99[th] percentile 30-day rolling averages).  Accordingly, a revised emission limit of 7.6 lb NOx/ton clinker is recommended for the Trident kiln.

                                       
                                       
