United State* Industrial Environmental Research EPA-600/2-78-217
Environmental Protection Laboratory December 1978
Research Tn
Afl»ncy
riangle P»rk MC 27711
Research and Deveiopment
Combustion Modification
Effects on NOx Emissions
from Gas-, Oil-, and
Coal-Fired Utility Boilers
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development and application of en-
vironmental technology. Elimination of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
8. "Special" Reports
9. Miscellaneous Reports
This report has been assigned to the ENVIRONMENTAL PROTECTION TECH-
NOLOGY series. This series describes research performed to develop and dem-
onstrate instrumentation, equipment, and methodology to repair or prevent en-
vironmental degradation from point and non-point sources of pollution. This work
provides the new or improved technology required for the control and treatment
of pollution sources to meet environmental quality standards.
EPA REVIEW NOTICE
This report has been reviewed by the U.S. Environmental Protection Agency, and
approved for publication. Approval does not signify that the contents necessarily
reflect the views and policy of the Agency, nor does mention of trade names or
commercial products constitute endorsement or recommendation for use.
This document is available to the public through the National Technical Informa-
tion Service, Springfield, Virginia 22161.
-------
EPA-600/2-78-217
December 1978
Combustion Modification Effects
on NOx Emissions from Gas-,
Oil-, and Coal-Fired
Utility Boilers
by
Owen W. Dykema
The Aerospace Corporation
Energy and Resources Division
Los Angeles, California 90009
Grant No. R803283-03
Program Element No. 1AB014
EPA Project Officer. Robert E. Hall
Industrial Environmental Research Laboratory
Office of Energy, Minerals, and Industry
Research Triangle Park, NC 27711
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, DC 20460
-------
ABSTRACT
The report represents the conclusion of 4 years of analysis of large
quantities of emissions, operating conditions, and boiler configuration data
from full-scale multiple-burner, electric generating boilers firing natural
gas, oil, and coal fuels. The overall objective of the study was to develop
from this data: (1) further understanding of the effects of combustion modifi-
cations on combustion, and the resulting effects on NOX emissions; and (2)
directly applicable guidelines for the application of combustion modification
techniques for the control of NOx emissions in full-scale operating utility
boilers. The report includes: (1) discussion of modeling techniques used to
analyze the date; (2) conclusions relative to the sources of NOX within the
furnace; (3) guidelines for NOX reduction; and (4) an example application of
the guidelines. Boiler firing types include single-wall, opposed and tangential
configurations.
The report concludes that NOX emissions are generated, in varying degrees,
from conversion of fuel-bound nitrogen (the predominant source), hetero-
geneous combustion and mixing zone, second-stage mixing zone, and active
burner region. Maintaining very fuel-rich initial combustion conditions,
holding the initial peak combustion to <2050 K, and delaying fuel gasification
and mixing until the gas has been cooled somewhat should reduce NOX emis-
sions from all four main sources.
ii
-------
PREFACE
This study represents the conclusion of four years of
study of modification of combustion in full-scale, multiburner utility
boilers for the purpose of NOX reduction and of some of the associated
side effects. Previous work is reported in: (1) Analysis of Test
Data for NOX Control in Gas- and Oil-Fired Utility Boilers (EPA-
650/2-75-012, January 1975); (2) Analysis of Test Data for NOx Control
in Coal-Fired Boilers (EPA-600/2-76-274. October 1976); (3) Effects
of Combustion Modifications for NOX Control on Utility Boiler Efficiency
and Combustion Stability (EPA-600/2-77-190, September 1977); and
(4) in papers presented at the first and second EPA Stationary Source
Combustion symposia (EPA-600/2-76-152c, June 1976) and (EPA-
600/7-77-073b, July 1977), respectively. The data and the analytical
techniques reported in the earlier reports were used as the basis for,
and were extended in, the work reported herein.
This study, as well as the three previous studies, was
conducted for the U. S. Environmental Protection Agency, Combustion
Research Branch, Industrial Environmental Research Laboratory,
Research Triangle Park, North Carolina, during the third year of a
three-year continuing grant. (The first study was conducted under a
separate EPA Grant No. R-802366 for this same EPA office. ) The first
study concerned the effects of combustion modifications of NOX
emissions in natural gas- and oil-fired boilers. The second study ex-
tended the analysis to coal-fired boilers. The third study evaluated the
effects of combustion modifications (necessary for NOx control) on two
possible limiting side effects: (1) excessive loss of plant efficiency;
and (2) combustion instability.
This final study report contains: (1) a consolidation and
summary of the previous work, as modified by certain simplifications
and extensions resulting from improved understanding gained in this
work; and (2) a simple example calculation of the developed guidelines,
to minimize NOX emissions in an oil-fired tangential boiler.
111
-------
A brief introduction is contained in Section I. Con-
clusions and recommendations arev contained in Section II, and a brief
summary of this final study is contained in Section III. Modifications
made during this study to the analytical technique developed in the
previous studies are discussed in Section IV. Results of the data
analyses, using this modified NOx calculation technique, are presented
in Section V. Finally, the example application of the developed guide-
lines, to minimize NOx emissions in an oil-fired tangential boiler, is
shown in Section VL
IV
-------
CONTENTS
PREFACE iii
NOMENCLATURE ix
ACKNOWLEDGMENTS xi
I. INTRODUCTION 1
n. CONCLUSIONS AND RECOMMENDATIONS ... 3
2. 1 Conclusions 3
2. 2 Recommendations 5
2. 2. 1 Guidelines 5
2. 2. 2 Application of Guidelines 6
III. SUMMARY 8
3. 1 Background 8
3. 2 Results of the Current Study 10
3. 2. 1 Modifications to the Analysis . . 10
3. 2. 2 Assessment of the Validity
of the Calculation 11
3. 2. 5 General Guidelines 12
3. 2. 4 Example Application of
Guidelines 14
IV. MODIFICATIONS TO THE ANALYSIS 15
4. 1 Background 15
4. 2 Simplification 19
4. 3 Improvements 21
-------
4. 3. 1 Boiler Air-Fuel Ratios 21
4. 3. 2 The Final Mixing Zone 31
4. 3. 3 Gas Cooling Rates 33
4. 3. 4 Finite Rate Gasification
and Mixing 36
4. 3. 5 Some Considerations
Relative to Heterogeneous
Flames 40
V. RESULTS OF DATA ANALYSIS 46
5. 1 Results of Regression Analyses 47
5. 1. 1 Correlation Coefficients 47
5. 1. 2 Coefficients of the Terms .... 47
5. 2 Direct Data Analyses 52
5. 2. 1 Natural Gas Fuel 53
5. 2. 2 Oil Fuel 56
5.2.3 Coal Fuels 58
5. 3 Effects of Gasification and Mixing Rates. 61
5. 4 Guidelines 63
5. 4. 1 Natural Gas Fuel 66
5.4.2 Oil Fuel 69
5. 4. 3 Coal Fuel 71
VI. EXAMPLE APPLICATION OF NO,, REDUCTION
GUIDELINES 77
6. 1 Preliminary Calculations 77
6. 2 Modification for Minimum NO 79
.X.
6. 3 Side Effects 81
REFERENCES 84
VI
-------
FIGURES
1. Carbon Dioxide and Oxygen Concentrations Measured
in the Flue Gases: (a) Natural Gas-Fired Boilers,
(b) Oil-Fired Boilers, (c) Two Single-Wall Boilers
Firing a Nominal Coal, (d) A 360-MW Tangential Boiler
Firing a Nominal Coal, and (e) A 330-MW Tangential
Boiler Firing a High-Nitrogen Coal 25
2. Comparison of Calculated and Measured NOX
Emissions: (a) Natural Gas-, Opposed-Fired
Boilers and (b) Natural Gas-, Single-Wall-Fired
Boilers 54
3. Comparison of Calculated and Measured NOX
Emissions: (a) Oil, Opposed-Fired Boilers and
(b) Oil, Single-Wall-Fired Boilers 57
4. Comparison of Calculated and Measured NOX
Emissions: (a) Coal-Fired, Single-Wall Boilers
and (b) Coal-Fired, Tangential Boilers 60
5. Effects of Combustion Staging on NOX Emissions:
Natural Gas-, Opposed-Fired Boilers 68
6. Effects of Combustion Staging on NOX Emissions:
Oil, Opposed-Fired Boilers . 70
7. Effects of Combustion Staging on NOx Emissions:
(a) Two Single-Wall Boilers Firing a. Nominal Coal,
(b) 360-MW Tangential Boiler Firing a Nominal Coal,
and (c) 330-MW Tangential Boiler Firing a High-
Nitrogen Coal 73
8. Parametric Calculations to Determine the Necessary
Flue Gas Recirculation in the Example Boiler 82
Vll
-------
TABLES
1. Boilers in the Data Sample 22
2. Summary of the Total Data Sample 23
3. Results of Regression Analyses of NOX
Emissions Data 48
4. Example of the Possible Wide Variation in
Measured NO Under Essentially the Same
Operating Conditions 55
5. Effects of Oil Vaporization and Mixing
Parameters on the Thermal NO Generated
Ji
in the Active Burner Region 64
6. Values of Example Boiler-Specific Input
Variables 78
Vlll
-------
NOMENCLATURE
FORTRAN Notation
DCG = distance into the furnace, from the burner exit, for
complete gasification and mixing
FSS = fraction of the vaporized fuel involved in combustion
under stoichiometric conditions
SON = product of the oxygen and nitrogen concentrations, in
the combustion products, appropriate to the Zeldovich
NOX formation rate equation
1/2
H N
Arabic Notation
A, B = arbitrary intermediate constants, equation (1)
dNO = the increment NO formed in a given stream tube, ppm
X X
K = a proportionality constant relating radiant heat flux, to a
water wall, to the fourth power of the combustion gas
temperature, A ? 4.
J/cm2 - sec - K4 (Btu/ft - hr - °R )
Q = the heat flux to the water walls,
J/cm2 - sec (Btu/ft2 - hr)
R = the surface-to-volume (circumference-to-flow area)
ratio of the radiant section of furnace,
cm
-1
(ft2)
T = temperature, K (°R)
Z = an arbitrary variable defined by equation (7)
t = time, seconds
IX
-------
Greek Notation
€ = the effective emissivity of the combustion products,
° dimensionless
°" = the Stephan-Boltzmann constant,
5. 77 x 10"12 J/cm2 - sec-K4 (1. 74 x 10"9 Btu/ft2-hr-°R4)
Subscripts
f refers to the final conditions at the end of a stream tube
g refers to the combustion gases
i refers to the initial conditions at the beginning of a stream
tube
w refers to the water wall
x
-------
ACKNOWLEDGMENTS
Sincere appreciation is acknowledged for the guidance
and assistance provided by Mr. Robert E. Hall of the Combustion
Research Branch, Industrial Environmental Research Laboratory,
Research Triangle Park, North Carolina, who was the U. So Environ-
mental Protection Agency Project Officer during the conduct of this
study.
A special acknowledgment is also due, once again, to
the Los Angeles Department of Water and Power for its continued
cooperation, over the years of this study, in making available its
data from full-scale operating utility boilers firing natural gas and
oil fuels.
Acknowledgment is also due Mrs. Sandra M. Barnes
and Ms. Frances J0 Twillie of The Aerospace Corporation for their
assistance in computer programming and operation.
XI
-------
SECTION I
INTROQUCTION
This report contains the results of a four-year
combustion modification study. The purpose of the study was to
reduce NOx emissions from utility boilers. In general, this work
was neither a detailed analytical study from first principles (of
some aspects of the NOx control problem) nor a full-scale, cut-
and-try experimental program. Rather, it represents an effort to
bridge the gap between these two extremes. Detailed information
developed from first principles, where available, was supplemented
by empiricism, using laboratory and full-scale boiler data, to
develop an engineering calculation for nitrogen oxide (NOX)
emissions from full-scale, multiburner utility boilers firing natural
gas, oil, and coal fuels.
Initially, the NOX calculation was relatively uncertain.
Therefore, by means of regression analysis techniques, a large
amount of data (about 600 tests) were used to quantify the NOx cal-
culation coefficients. The data was obtained from special NOx test-
ing in full-scale, operating utility boilers. Guidelines for NOx
reduction were then developed by conducting parametric calculations
with the quantified NOX calculation expression.
Simplifications and improvements made in the current
study resulted in a direct, sufficiently accurate calculation. The
analytically developed expression could be used directly to calculate
the NOX emissions, without the need to quantify via regression
analysis. The large data sample was then used only to assess and
verify the calculation.
The resulting NOX calculation is not a simple equation,
in terms of the significant independent variables, which can be written
down here. The final equation used in the regression analyses consisted
of only three terms (NO]., NO2, and NO3) each with a constant coefficient,
plus a regression analysis constant. Each term represents the result
of a complex computer calculation of the NOX contribution from a
major source within the furnace (i. e. , the active burner region,
the second-stage mixing zone, and the conversion of fuel-bound nitrogen
to NOX). Since the objectives of this study were primarily to analyze
-------
data and to develop further understanding and guidelines for the
control of NOX, the computer program listing and user manual are
not included in this report.
In general, the direct calculation is adequate for natural
gas- and oil-fired boilers. It was necessary, however, to empirically
establish the coefficient of one of the three terms to obtain satisfactory
agreement with data from the coal-fired boilers. Guidelines for NOX
reduction in full-scale, multiburner boilers were then reevaluated in
the current study.
Early in the previous studies, it appeared that strongly
staged combustion, necessary as a prime combustion modification to
minimize NOX, might result in significant losses in plant efficiency
and/or flame and combustion instability. Later in those studies, a
separate analysis was conducted to evaluate these possibilities. It was
concluded that there is no evidence, at least within the range of the
available data, that staged combustion significantly affects plant efficiency;
however, flame and combustion instability could be more of a problem
under these conditions. An analytical method was developed to guide
modifications to assure stable combustion.
This report is based on the previous work but largely
reports the results of this final study, which was conducted over the
period July 1976 through December 1977.
-------
SECTION II
CONCLUSIONS AND RECOMMENDATIONS
The following conclusions and recommendations result
from the entire four-year effort to develop guidelines for nitrogen
oxide (NOX) emission control in utility boilers. These guidelines are
within the constraints of high plant efficiency, stable combustion, and
low emissions of other air pollutants. Many conclusions and recom-
mendations developed in the previous studies (referred to in Section I)
are confirmed or modified in the current study; others have been
developed in this study. Conclusions largely concern the sources of
NOx within a full-scale, multiburner boiler, as indicated by this analysis,
while recommendations largely concern guidelines to reduce total NOX
emissions from the boilers by reducing the contributions from the
various boiler sources.
All conclusions and recommendations are based on the
overall conclusion that total NOX emissions are lower when two distinct
and separate stages of combustion are involved (staged combustion)
than when air-only burners are distributed among the active burners in
the burner array (off-stoichiometric combustion). This is especially
true when fuels containing significant amounts of bound nitrogen are
used.
2. 1 CONCLUSIONS
NOX emissions appear to be generated, in varying
degrees, from four main sources in full-scale, multiburner boilers:
a. Conversion of Fuel-Bound Nitrogen. By far the
predominant source (nearly 100 percent) of NOX
emissions in tangential coal-fired boilers is the
conversion of fuel bound nitrogen. The same is
true, to a lesser degree, in other coal-fired
boilers, depending on the degree of combustion
staging. Somewhat less than half of the NOX
from low-nitrogen oil-fired boilers (and, of course,
none from natural gas-fired boilers) is from fuel-
bound nitrogen.
-------
b. Heterogeneous Combustion and Mixing Zone. Sig-
nificant contributions to the overall NOx emissions
can be thermally generated in the heterogeneous
combustion and mixing region just off the active
burners. This contribution is largely independent
of the degree of combustion staging (the overall
active burner air-fuel ratio). This is because air-
fuel ratios in this region range from zero (in the
fuel) to infinity (in the as-yet unmixed combustion
air), and some combustion is always occurring
under stoichiometric conditions at the highest com-
bustion product temperatures anywhere in the
boiler (a combination of the stoichiometric com-
bustion temperature rise and the absence of signif-
icant heat loss). The same is true with gaseous
fuels because the gaseous fuel and combustion air
mixing is not instantaneous. NOX contributions
from this region can be reduced only by controlling
the temperature-time history of reactants in these
near-stoichiometric flames.
c. Second-Stage Mixing Zone. Calculations indicate
that a significant amount of NOX can be formed
during the mixing of the second stage combustion
air with the products of fuel-rich combustion in the
first stage,, As in the heterogeneous combustion
region, this mixing is not instantaneous, and all of
the products from the first stage must shift from
fuel-rich through stoichiometric to fuel-lean con-
ditions as this mixing occurs, over a finite time
period. Unlike the heterogeneous region, however,
considerable cooling has already taken place in
the first stage before the second stage air is intro-
duced. Calculations indicate that less than about
80 ppm of the total NOX are contributed by this
final mixing zone.
d. The Active Burner Region. If the peak temperatures
in the heterogeneous combustion region are very
high (of the order of 2200 K (3500°F)), significant
NOX can be generated in the remaining completely
gasified and mixed regions of the active burner
region where there has not yet been sufficient cool-
ing to reduce the thermal NOX formation rate.
Normally, however, modifications necessary to
control NOX from the major sources discussed in
(a) and (b) will also reduce NOX formed in the
remaining parts of the active burner region to
negligible levels.
-------
Overall conclusions regarding the potential for NOX
reduction by combustion modifications are dependent on the guidelines
discussed in the following paragraphs.
2. 2 RECOMMENDATIONS
The following recommendations concern guidelines for
combustion modifications most effective in minimizing NOX formed
from the various sources discussed in Section 2. 1. They are based on
observations from the final direct NOX calculation program and not
from parametric calculations involving a NOX data correlation equation
quantified by regression analyses, as in the previous studies. As a
result, extrapolations beyond existing data are somewhat more depend-
able. In addition, further analytical and experimental studies are rec-
ommended.
2. 2. 1 Guidelines
The guidelines developed during the course of this study
for the most effective control of NOX are as follows:
a. Conversion of Fuel-Bound Nitrogen. With fuels
containing significant bound nitrogen, minimum
conversion can be obtained by maintaining very
fuel-rich initial combustion conditions until all of
the fuel has gasified, mixed with combustion air,
and the products have shifted to equilibrium under
the fuel-rich conditions. This can be accom-
plished under staged combustion conditions with
the burners operated at overall burner air-fuel
equivalence ratios of less than about 0. 7. Such
fuel-rich burner operation may create greater
problems of combustion instability and/or flame
liftoff; however, analytical and engineering design
techniques have been developed in the previous
studies to control these potential problems.
b. Heterogeneous Combustion and Mixing Zone. NOX
thermally generated in the heterogeneous combustion
region of the active burner can best be minimized
by controlling the initial peak (near stoichiometric)
combustion temperatures to less than about 2050 K
(3230 F) and, to some extent, by delaying fuel
gasification and mixing until some gas cooling has
transpired. Combustion temperatures can be mini-
mized, while maintaining high plant efficiency, by
transferring more heat to the steam cycle, prior to
entering the air preheater, and less to the combustion
air and/or by recirculating flue gases into the com-
bustion air in the windbox or in the burners, prior
to combustion. While both of the se approaches
-------
would involve considerable retrofit construction,
both have been applied, with success, in exist-
ing boilers.
c. Second-Stage Mixing Zone. Reductions in the
initial peak combustion temperature necessary to
minimize NOX thermally formed in the hetero-
geneous combustion region should effectively
prevent significant thermal NOX in the second
stage of combustion. Care must be taken, how-
ever, to avoid cooling the gases from the active
burner region to such low values that carbon mon-
oxide and any unburned hydrocarbons in the
products of fuel-rich combustion in the first stage
cannot be burned out to acceptable levels of
emissions of these pollutants. This optimum com-
promise can only be determined in the specific
boiler. As initial peak temperatures are reduced,
under the very fuel-rich staging, the primary
source of NOX emissions may shift from the heter-
ogeneous combustion region to this final mixing
zone.
d. The Active Burner Region. The very fuel-rich
combustion staging discussed in recommendation
(a), coupled with the reduced initial peak com-
bustion temperatures discussed in recommendation
(b), should be adequate to prevent significant NOX
formation in the remainder of the active burner
region.
2. 2. 2 Application of Guidelines
Some expected results and further implications in appli-
cation of these guidelines are as follows:
a. Potential NOX Reduction. Full application of the
above guidelines, according to the calculations of
this study, should reduce NOx emissions in levels
less than 100 ppm with natural gas, oil or coal
fuels. These calculations are supported by data
from full-scale operating utility boilers, in (only)
one case, with natural gas fuel, to NOx levels
well below 100 ppm. With oil and coal fuels, how-
ever, the calculations are similarly supported
only to about 200 ppm. With the latter fuels, then,
the predicted further reductions, to less than 100
ppm, represent extrapolations from existing data.
b. Uncertainties and Further Study. The remaining
uncertainties in these extrapolations include the
following:
-------
1. Effectiveness of modifications necessary
to maintain combustion stability and to avoid
flame liftoff
2. Possible overall incomplete burnout of
carbon monoxide and/or hydrocarbons
3. Further hydrocarbon combustion in the
second stage, with attendant increases in NOX
from that source.
4. Effects of very fuel-rich first stage com-
bustion on water-wall tube life.
All of these uncertainties should be evaluated experimentally in full-
scale operating utility boilers. So that boiler safety can be ensured,
modifications designed to avoid flame liftoff should first be evaluated
with a full-scale burner firing into a laboratory furnace.
Much analytical work remains to be done to improve
understanding of the fundamental, coupled physical and chemical pro-
cesses involved in the generation and control of NOX emission from
full-scale, operating utility boilers. Such work would not only increase
the accuracy of current NOX control guidelines but would develop
greater confidence in the resulting extrapolations from existing data.
The subject studies have indicated that the most significant of these
areas is the probable role of NOX destruction mechanisms in the net
conversion of fuel-bound nitrogen to NOX in the regions of hetero-
geneous combustion of coal and oil.
-------
SECTION HI
SUMMARY
3. 1 BACKGROUND
This report contains the results of a four-year
combustion modification study. The purpose of the study was to reduce
NOX emissions from utility boilers. The general approach was to
derive an analytical calculation technique based, where possible, on
what appear to be well-established mechanisms for controlling NOX
formation in heterogeneous flames. Remaining gaps in knowledge were
bridged with empiricism, by use of laboratory data on the particular
process where available. Large samples of data from NOx testing in
full-scale operating boilers were then used either to quantify the result-
ing calculations through regression analyses or to verify and evaluate
the calculations.
The calculation technique was developed in a study
reported in Reference 1, specifically for natural gas- and oil-fired
boilers. The model of combustion and NOX formation developed as the
basis for the calculation divided the radiant section of the boiler into
104 series and parallel tank-and-tube type of mixing and reaction zones.
In each of these zones, the mixing between the separate flow streams
entering the zone and the resulting hydrocarbon-air reactions (shifting
equilibrium) were assumed to take place instantaneously (in the tank).
Thermal NOx formation occurred when the tank products (slug flow)
flowed to the next serial mixing zone. Thermal NOX formation was
calculated by means of a simple rate equation based on the Zeldovich
mechanism. NOx concentrations were always assumed to be low enough
(below equilibrium) that homogeneous gaseous NOx destruction mechan-
isms, at least those resulting from the Zeldovich mechanism, could be
neglected. In the initial study, NOX formation from conversion of fuel-
bound nitrogen was approximated as a simple constant fraction of the
concentration of nitrogen in the fuel. This latter approximation was used
primarily because: (1) very little was known, at the time, either about
the magnitude of NOx emissions which might be generated from this
source or about the conversion mechanism; and (2) natural gas contains
no fuel-bound nitrogen and the oil fuel involved in the related data sample
contained little nitrogen. Other approximations necessary to provide a
manageable calculation and that were appropriate to the engineering
nature of the overall calculation included: (1) the use of a constant
8
-------
(linear) combustion gas cooling rate (with time); and (2) the assumption
of instantaneous gasification and mixing of the fuel with the combustion
air at the exit of the active burners.
Application to coal-fired boilers of the model developed
in previous studies required a more accurate and variable calculation
of the NOX formed from conversion of the fuel-bound nitrogen. There
is still no recognized mechanism for this conversion, particularly in
heterogeneous flames. A semiempirical calculation, based on a simple,
global model and using available laboratory and some full-scale utility
and industrial boiler data (oil and coal), was developed. This calcu-
lation was subsequently shown to agree quite well with estimates of the
NOX resulting from conversion of fuel-bound nitrogen in the coal data
sample of this study. Results of the application to coal-fired boilers
are reported in Reference 2 and general results from the studies of
gas-, oil-, and coal-fired utility boilers were presented at the EPA
First Symposium on Stationary Source Combustion [ 3],
In general, the results of the studies of NOX control in
natural gas-, oil-, and coal-fired utility boilers indicated no limit to
NOX reduction by combustion modification that is inherent in that NOX
control technique. The developed NOX calculation does have zero
NOX as a solution within what are thought to be reasonable hardware
and operating conditions. NOX limits, then, were thought to lie in
some undesirable side effects which might appear during attempts to
reach those hardware and operating conditions. One possible effect
is the increased potential for corrosion from the very fuel-rich operation
of the active burners (necessary to minimize conversion of fuel-bound
nitrogen). This area must be investigated experimentally and is being
pursued by other agencies. Two other possible limiting side effects
are significant reductions in plant efficiency and combustion and flame
instability. These effects were studied in the third year. Results of
that study are reported in Reference 4 and were summarized in a paper
fresented at the EPA Second Stationary Source Combustion Symposium
5 ]. In general, results of that study indicated that (1) no significant
effect of staged combustion on plant efficiency could be observed within
the scatter of the data and (2) a potential exists for increased problems
with flame liftoff and combustion instability under the very fuel-rich
burner operation necessary to control NOX emissions from conversion
of fuel-bound nitrogen. The available data sample was not adequate to
verify the effects of reduced combustion air temperature and flue gas
recirculation in the active burners on either plant efficiency or com-
bustion stability. An analytical technique was developed to assure com-
bustion stability even under the very fuel-rich operating conditions.
During the study of coal-fired utility boilers [ 2] , an
ancillary effort was made to develop a reasonable approximation of
finite rate gasification and mixing, particularly with coal fuels. Such
a calculation was developed, but it was beyond the scope of that study
to incorporate this calculation in the overall NO,, model.
-------
3.2 RESULTS OF THE CURRENT STUDY
The current study, the last in this long-term effort,
consisted of three tasks: (1) modification of the NOX calculation tech-
nique to incorporate results of insight and knowledge gained in the
earlier studies; (2) evaluation and assessment of the validity of the new
NOX calculation; and (3) demonstration of the application of the result-
ing guidelines by suggesting modifications to an example boiler to mini-
mize NOX emissions within the constraints of high plant efficiency,
stable combustion, and low emissions of other air pollutants.
3. 2. 1 Modifications to the Analysis
One result of the earlier studies of natural gas-, oil-,
and coal-fired utility boilers was a conclusion that NOX emissions are
as low or, in most cases, lower when the boiler is operated with staged
combustion than with off-stoichiometric combustion. The two firing
configurations are defined as follows: (1) in staged combustion, all of
the combustion air not entering the furnace through the active burners
is introduced downstream of all the active burners (two distinct and
separate stages of combustion); and (2) in off-stoichiometric configu-
rations, air-only burners can be distributed anywhere in the active
burner matrix. In staged combustion, the air-fuel ratio is uniform
throughout the entire active burner region. With off-stoichiometric
configurations, the combustion products from an active burner can
immediately mix with the air fron an adjacent air-only burner. These
products can subsequently mix with the products from another active
burner, and so on. The mixed air-fuel ratio can cross back and forth
across stoichiometric, repeatedly, while still in the active burner
region.
It was decided, therefore, that off-stoichiometric con-
figurations were no longer of interest in this NOX reduction program.
Not only were all of the tests involving off-stoichiometric configurations
eliminated from the data sample, but the capability to calculate NOX
emissions in these very complex mixing cases was deleted from the
computer program. This action greatly simplified the computer pro-
gram and allowed incorporation of more detailed (and, therefore, more
complicated) and improved calculations in (1) calculating overall boiler
air-fuel ratios from measured O2 and CC>2 data, allowing for carbon
losses; (2) finite rate mixing of the second stage combustion air with the
products of combustion in the active burner region; (3) radiation cooling
of the combustion products proportional to the fourth power of the product
temperature; and (4) finite rate gasification and mixing of the fuel and
combustion air from the active burners. Some effort was also spent in
an (unsuccessful) attempt to improve the calculation of NOX generated
from conversion of the fuel-bound nitrogen. Although the above improve-
ments were incorporated into the overall NOX emissions calculation,
it is still an engineering analysis, and these improvements were incor-
porated as first-order approximations commensurate with the engineer-
ing nature of the remaining program.
10
-------
Perhaps the most significant result of limiting the pro-
gram to staged combustion configurations was that the number of terms
required to calculate the NOX emissions were greatly reduced and the
remaining terms have more direct and clear physical significance. This
allowed greater confidence in the direct calculation of NOx emissions,
without the use of regression analyses to quantify the coefficients of the
remaining terms. In this study, then, the regression analyses were used
primarily as one means with which to evaluate the validity of the NOX
calculations. ,More insight and information could be developed from
direct comparison, on a point-by-point basis, of the NOX calculations
with the measured data. In these comparisons, the NOX emissions
were calculated directly, with little dependence on the results of the
regression analyses. The calculation consisted of summing the NOX
contributions from just three sources within the furnace: (1) NOX
thermally generated in the active burner region (including that gener-
ated in the heterogeneous combustion region where finite rate gasification
and mixing occur); (2) NOX generated in the active burner region from
conversion of the fuel-bound nitrogen; and (3) NOX thermally generated
in the region of finite rate mixing of the second-stage air with the prod-
ucts of combustion in the active burner region (the final mixing zone).
In this calculation, where possible, the coefficients of these terms
were taken to be equal to 1. 0 (the theoretical value), and the constant in
the regression analysis was taken to be equal to zero.
3. 2. 2 Assessment of the Validity of the Calculation
The validity of the final, modified NOX emissions cal-
culation was assessed in two ways: (1) by examination of the results of
regression analyses; and (2) by direct comparison, on a point-by-point
basis, of the calculated and measured data.
In general, the results of these assessments showed that
the NOX calculations were in good agreement with the available meas-
ured data except that: (1) agreement with the data from the single-wall
boilers firing natural gas or oil fuels was very poor; and (2) with coal
fuels the calculation indicated that considerable NOX is thermally
generated in the active burner region, while the measured data, par-
ticularly for tangential boilers, indicate negligible amounts.
Direct calculations, using just these three terms, all
with term coefficients of 1. 0, and no constant showed good agreement
with the measured data from the opposed-fired boilers firing natural
gas and oil fuels. On the average, the NOX levels calculated for 92
tests with natural gas in the opposed fired boilers were higher than the
measured levels by only 7 ppm, but an average deviation of just under
100 ppm remained. Calculations for 30 tests in these same boilers
with the oil fuel were high by 50 ppm, but the average deviation was
only 44 ppm. Direct calculations for coal-fired tangential boilers,
neglecting all NOX calculated to be thermally generated in the active
burner region (a term coefficient of zero), also showed good agreement
11
-------
with that data. The data calculated in this manner for 62 tests in two
tangential boilers, firing somewhat different types of coal, were higher
than the measured, by only 26 ppm with an average deviation of 43 ppm.
This same type of calculation for the single-wall, coal-fired boilers,
however, showed that the calculation was low by a constant 127 ppm.
The direct calculation is considered adequate for boilers
firing natural gas and oil fuels. The poor data agreemement with data
from single-wall boilers firng these fuels is considered a result of in-
adequate input data (combustion air temperatures in the burners)
rather than of error in the subsequent calculation. Calculation of NOX
thermally generated in the active burner region needs further work.
A summary of the final appropriate term coefficients
and empirical constants are shown in the following table:
Firing
Type
Fuel
Opposed Gas
Oppo s ed Oil
Single-wall Coal
Tangential Coal
No. of Term Coefficients
Tests ABR FMZ FBN
92
30
34
62
1.0
1. 0
0.0
0.0
1. 0
1.0
1.0
1.0
NA
1.0
1.0
1.0
ppm
Empir-
ical
Const.
0.0
-50
+ 127
0.0
Avg
Devi-
ation
100
44
46
43
Active burner region ABR,
fuel-bound nitrogen FBN.
final mixing zone FMZ, and
3.2.3
General Guidelines
The primary objective of this and the preceding studies
of NOX control in utility boilers was to develop and verify simple guide-
lines which can be used, along with a judicious testing program, to
minimize NOx emissions within the bounds of high plant efficiency,
stable combustion, and acceptable levels of emission of other air pollu-
tants. The very fuel-rich burner operation necessary to achieve very
low levels of NOX emissions can result in an increased tendency for
combustion instability and flame liftoff. An analytical technique was
developed to provide stable combustion. Care would have to be taken in
burner design so that the flames in the burner exit are soundly anchored.
(See, also, Recommendations, Section 2. 2. 2. )
The primary effects of staged combustion, with a very
fuel-rich first stage, are to minimize NOX from conversion of fuel-
bound nitrogen and thermal NOX generated in the mixing and combustion
12
-------
of product gases in the active burner region. Active burner air-fuel
equivalence ratios of less than about 0.7 may be necessary to achieve
very low NOX contributions from these sources.
With such fuel-rich operation of the active burners and
the entire first stage of combustion, significant NOX may still be ther-
mally generated in two other regions of the furnace: (1) the subregion
of the active burner region where the fuel gasifies and the resulting
gases mix with the combustion air (heterogeneous combustion); and (2)
the region where the second-stage air is mixed with the products of
fuel-rich combustion coming from the active burner region (the first
stage). Calculations in this study consistently indicated that NOX from
the second source is usually quite small but cannot be neglected if very
low total NOX levels are sought.
NOX generated in the region of heterogeneous combustion
appears to be relatively independent of the burner air-fuel ratio because
the local air-fuel ratios are largely controlled by the relative rates of
gasification and mixing. Burner design changes which tend to slow the
rate of fuel gasification and decrease the rate of mixing of these gases
with the combustion air can substantially reduce the NOx generated in
the heterogeneous combustion region. A more direct approach, however,
(which also reduces NOx formed in the second stage) is to reduce the
peak combustion temperature in the heterogeneous combustion region.
This can be done by techniques such as (1) reduction of the combustion
air temperature, (2) dilution of the combustion air with flue gas recir-
culated into the windbox or burners, and (3) water sprayed into the com-
bustion air. The objective of all of these is to reduce the maximum
(near stoichiometric) flame temperature in the earliest regions of com-
bustion, before appreciable gas cooling has occurred. Calculations in
this program indicate that maximum initial flame temperatures (which
occur at the burner exit) less than about 2050 K (3230°F) minimize NOx
formation in the heterogeneous combustion region and virtually elimin-
ate any from the second-stage mixing region.
The data available to this study show trends which verify
the guidelines presented herein. Estimates of very low NOx levels
achievable with these guidelines are largely based on the calculation
developed in this and the previous study. The calculation was improved
in this study to the point where the empiricism involved is greatly
reduced. For example, regression analyses of large data samples are
no longer used to quantify the NOX calculation. Thus, conclusions
relative to minimum NOX levels achievable by combustion modifications
represent relatively small analytical extrapolations from existing data.
The NOX calculation shows zero NOX as a solution
within the range of what are thought to be acceptable utility boiler oper-
ating conditions. Limits on NOX reduction may arise as a result of
undesirable side effects such as combustion instability, flame liftoff,
excessive water-wall erosion/corrosion, or from unknown effects
13
-------
which are currently not a part of the NOX calculation. An example
of the latter might be incomplete hydrocarbon combustion in the very
fuel-rich first stage, followed by combustion in the oxidizing atmos-
phere of the second stage. These guidelines, however, are intended
to guide the way to minimum NOX> in conjunction with testing to reveal
these other possible limiting factors. Full-scale boiler testing and
long-term operation with natural gas and oil fuels have rather closely
approached the desired operation, with no significant evidence of un-
desirable operation.
3. 2. 4 Example Application of Guidelines
As an analytical example of the application of the guide-
lines discussed in Section 3. 2. 3, an existing 320 MW natural gas-fired
tangential boiler was selected for modification to oil-firing. The
guidelines were applied to achieve minimum NOX emissions.
Staged combustion was (analytically) achieved by using
the existing NOX ports and by operating the top two rows of burners
(8 of the 24) air-only. A burner air-fuel equivalence ratio of 0. 638 was
selected as a design point to minimize (theoretically to zero) conversion
of the fuel-bound nitrogen to NOx. The oil fuel used contained 0. 24 per-
cent nitrogen, by weight.
The combustion air temperatures, at full load, were
reduced by about 39 K (70°F) to account for the higher oil combustion
temperature. The combustion air and flue gas recirculation tempera-
tures were controlled, respectively, to 478 K (400°F) and 542 K
(515°F).
In addition, flue gases -were (analytically) recirculated
and mixed with the combustion air prior to mixing with the fuel. The
amount of recirculated flue gas was treated as a variable, and the
resulting NOX emissions were then calculated. This calculation
showed that,such recirculation of flue gas in amounts corresponding to
more than 15 percent of the total burner air (all combustion air except
NOX port air), in combination with the staged combustion and com-
bustion air temperature control, resulted in predicted total NOX
emissions of less than 20 ppm.
14
-------
SECTION IV
MODIFICATIONS TO THE ANALYSIS
4. 1 BACKGROUND
This section contains the results of a. four-year
combustion modification study. The purpose of the study was to reduce
NOX emissions from utility boilers. Earlier reports on the study are
listed as References 1 through 5. The first year of the study analyzed
these effects in natural gas- and oil-fired utility boilers [ 1 ] . The
second year extended that approach to coal-fired boilers [2, 3]. The
third year examined the implications of these combustion modifications
on overall plant efficiency (electrical output per unit of heat input with
the fuel) and on combustion and flame stability [4, 5] since these
requirements might represent limitations on reduction of NOX by com-
bustion modifications.
The entire study made use of a large amount of data
from special tests of the effects of certain combustion modifications on
NOX emissions from full-scale operating utility boilers. A total of
575 tests involving combustion of natural gas, oil, and four types of
coal fuels in single-wall, opposed, and tangential firing configurations
were included in the sample. Distribution of the data among the fuels
and firing configurations is shown in References 1 and 2.
It was recognized from the outset that little of the
necessary basic phenomena involved in the formation and destruction
of NOX in diffusion flames was well understood. Application of exist-
ing understanding to the simultaneous and coupled gasification, mixing,
reaction, and heat transfer processes occurring in a full-scale multi-
burner utility boiler is still an extremely complex problem. Many
researchers were and are studying certain aspects of this problem, but
none are capable of integrating all aspects into a single analysis capable
of direct and accurate prediction of NOX emissions from full-scale
utility boilers starting from first principles. It is highly unlikely that
such a solution will ever be achieved (nor is it necessary that it be
achieved). Understanding of all applications of combustion can be
thought of as islands of fundamental understanding connected by empir-
ical bridges into satisfactory, workable design analyses. For example,
the basic chemical kinetics of combustion are thought to be well
15
-------
understood only in the case of hydrogen-bromine and, perhaps,
hydrogen-oxygen flames. The kinetics of hydrocarbon-air flames, the
case of interest in fossil-fuel-fired boilers, are very poorly understood.
Thus, with fossil fuels, the process of empirical bridging and global
approximation begins even at the most fundamental level.
The approach taken throughout this study was to start
with what is known about some of the individual processes, develop
some simple models or estimates of some of the more poorly known or
complex phenomena, and form them into a single expression which
should at least include most of the known, significant phenomena and
have the proper analytical form. The large samples of full-scale boiler
data were then used, via regression analyses, to fill in some of the
remaining gaps and to quantify the analytical expression. Results of
the regression analyses could be used to comment on the accuracy of
some of the assumptions incorporated in the analytical expression and
to provide a semiempirical, quantitative expression. This expression
could be exercised, in a parametric fashion, to further elucidate the
effects of some of the design parameters on NOX emissions. This
whole process represents an orderly, engineering approach to the
development of useful guidelines to minimize NOx emissions in full-
scale utility boilers. It can also provide other information, substan-
tiated by data, for feedback to fundamental and applied research pro-
grams.
The first attempt at this process (1) involved a simple
Arrhenius-type rate equation for the formation of NOx (zero destruction
rate), based on the Zeldovich mechanism, to account for thermally
generated NOX. The question of the appropriate rate coefficient was
bypassed by setting up the final expression so that the rate coefficient
was part of the coefficient of most of the terms. The rate coefficient
was thus lumped in with many other unknowns, and the total was quan-
tified by subsequent regression analyses of the large data samples.
Considering the other unknowns represented in the coefficients of the
terms in the final expression and the current state of understanding of
these rate coefficients, this step was considered justified. Since the
fuels involved in this first study (natural gas and low nitrogen oil) con-
tained little bound nitrogen, the amount of fuel-bound nitrogen converted
to NOX was approximated by a simple, constant fraction of the nitrogen
in the oil (and none in the natural gas).
The available data from full-scale, operating utility
boilers included relatively accurate measures of fuel flow rates, but
combustion air flow rates either were not measured or were measured
by devices calibrated against calculations based on the measured fuel
flow rates and the air-fuel ratio as indicated by flue gas chemical
compositions. Total combustion air flow rates, then, were calculated
in this initial study directly from the measured oxygen and carbon
dioxide concentrations in the flue gas. Simple stoichiometry was used,
involving the assumptions of negligible carbon loss and negligible
carbon monoxide in the flue gases.
16
-------
The mixing, reaction, cooling, and NOX formation
in the radiant section of the boiler (the furnace) was approximated by
dividing the major flow and mixing paths into 109 zones and modeling
each of these zones by the so-called tank-and-tube approximation. At
the beginning of each zone (the tank) mixing of the different streams
entering the zone, with perhaps different chemical compositions and
temperatures, was assumed to be instantaneous. The resulting mixed
gases, at uniform chemical composition and initial temperature, were
then assumed to flow through the zone (the tube), forming NOX and
cooling. The total NOX emissions were then calculated by summing
the contributions from each of these zones. Because the natural gas
and oil fuels in the available data sample were fired only in single-
wall or opposed configurations, a general model of this tank-and-tube
scheme was set up for those configurations. The complexity of such
a model, of course, dictated a computer solution.
The resulting analytical expression used in the regression
analyses contained seven linear terms to describe the formation of
thermal NOX, plus a single term for fuel-bound nitrogen conversion
and a constant. Of the seven thermal NOX terms, five described
similar zones in the active burner region, one concerned the zone
where NOX port flow (if any) is added and one described the NOX gen-
erated in the final zone after all fuel and air have mixed and reacted
but the product gases are still sufficiently hot to form significant
While the final expression consisted only of the linear
sum of these nine terms, each term required a more or less complex
computer calculation. A great deal of this complexity resulted from
the need to calculate the NOX for test cases where air-only burners
were intermixed with active burners (fuel plus air). Such configurations
result in what is usually called "off -stoichio metric combustion". This
term is often used to differentiate from "staged combustion", wherein
all combustion air not entering the furnace through active burners
enters through air-only burners and/or NOX ports located above
(downstream of) the active burners. Besides the mechanical problem
of identifying air -only and active burners on either side of (adjacent
to), opposite, and below a given burner, this intermixing of air -only
and active burners often resulted in huge shifts in local air-fuel ratios
(sometimes between very fuel- rich and very fuel -lean) in the lower
parts of the furnace as flows from active burners mixed directly with
flows from air-only burners. In addition, cases where the average
bulk gas air -fuel ratio starts fuel-lean (air -only burners located low
in the burner array) and approaches the overall boiler air -fuel ratio
from the fuel-lean side of stoichiometric had to be accounted for as
well as the more conventional staged combustion case where the bulk
gas air -fuel ratio starts fuel -rich and approaches the overall boiler
air -fuel ratio from below (and crosses) stoichiometric.
As a result of these complexities and the large number
of terms required to describe the mixing processes for all cases, the
17
-------
the coefficients in the final expression were difficult to calculate
directly. It was largely left to the subsequent regression analyses to
quantify these coefficients. Further, since a data sample can be fit
in many ways by an expression containing a large number of terms,
it was also difficult to interpret the resulting coefficients of an indiv-
idual term with regard to the accuracy of the assumptions and the cal-
culation technique involved in that term. The resulting expression,
however, quantified by the regression analyses, did relatively accu-
rately fit the available data and was successfully used in parametric
calculations to evaluate the effects of single and multiple independent
variables (modifications) on NOX emissions.
Another effect of the complexity of those NOX calcu-
lations was to allow (and justify) only very simple expressions for some
of the other phenomena such as the gas cooling rates, the vaporization
rates of the oil, and the mixing of gaseous and vaporized fuel with the
combustion air. Combustion product cooling was simply taken as
linear with time, and the vaporization and mixing rates were taken as
infinite (instantaneous vaporization and mixing at the burner exit). Any
further complexity introduced in that program might have created a
totally computer-oriented study and obscured the engineering objectives
of this study, i. e. , to develop useful guidelines for combustion modi-
fications to reduce NOX emissions.
Application of this technique to coal-fired utility boilers
required (1) development of a more detailed calculation of the NOX
generated by conversion of fuel-bound nitrogen, (2) modification of the
tank-and-tube mixing and flow scheme to accommodate tangentially
fired boiler configurations, and (3) improvement in calculation of the
thermal NOX generated in the second stage in two-stage combustion
configurations. Each of these necessary modifications were again
complicated by the need to describe or calculate NOX formation in the
off-stoichiometric firing configurations (air-only burners mixed in
with active burners in the burner array).
The final effort in studying the effects of combustion
modifications for NOX control was to evaluate the possible effects of
these mo d if i cations on overall plant efficiency and on flame and com-
bustion stability [4],
In each of the earlier studies, new phenomena were
investigated, and techniques of analysis were developed, extended,
evaluated, and improved. This study was intended to incorporate all
improvements developed in the first three years of study, to improve,
where possible, those areas shown to be most uncertain and to draw
final guidelines for the control of NOX within the constraints of high
plant efficiency, low emissions of other air pollutants, and stable flames
and combustion. The question of whether tube wall corrosion is affected
by staged combustion operation was not addressed in this study. This
subject is being investigated in other EPA-sponsored work.
-------
4,2 SIMPLIFICATION
The major single objective of this study was to simplify
the calculation of NOX within the furnace of utility boilers so that
certain desirable improvements (and recomplications) could be accom-
plished within the scope of this final, limited study. The only signifi-
cant simplification which appeared reasonable was to eliminate the
enormous complexity created by the initially presumed need to provide
the capability to calculate and correlate NOX emissions in boilers mod-
ified for off-stoichiometric combustion. These are configurations where
air-only burners are mixed in with active burners in the burner array.
Some of the resulting analytical complications are discussed in Section
4.1.
Nearly all of the results and data from the studies of the
effects of the locations of air-only burners in natural gas-, oil-, and
(particularly) coal-fired boiler burner arrays showed that NOx emissions
are always lowest when all of the combustion air not entering the furnace
through the active burners is introduced above, or downstream of, all
of the active burner flows (two-stage combustion). This appears to
result from the need to hold the local air-fuel ratio low, while the fuels
gasify and the initial hydrocarbon reactions and the simultaneous fuel-
bound nitrogen conversion reactions are completed. Particularly with
coal fuels, which gasify more slowly and contain significant concen-
trations of fuel-bound nitrogen, early mixing of air flows from adjacent
air-only burners with the still gasifying and reacting flow from an
active burner can increase the overall conversion of fuel-bound nitrogen
to NOX. The same is true, but to a lesser extent, with oil fuels. There
are some indications that two-stage combustion of gaseous fuels (contain-
ing no hydrocarbon-bound nitrogen) with a fuel-lean first stage may
produce NOX emissions at least as low as conventional two-stage com-
bustion. Even in this case, however, off-stoichiometric configurations
still appear to generate more (thermal) NOX than either two-stage con-
figuration.
Thus, while off-stoichiometric combustion configurations
are academically interesting and certainly represent an interesting
analytical challenge, such configurations do not appear to represent use-
ful approaches to maximum NOX reduction. Therefore, all off-
stoichiometric configurations were eliminated from the draft sample,
and the NOX calculation program was restricted to two-stage config-
urations with a fuel-rich first stage.
The direct result of this simplification was to reduce
the entire active burner region (burner flows as well as bulk gases
flow) to a single, constant air-fuel ratio. This eliminated the need to
describe the relative burner locations in detail; to calculate the various
combinations and locations of active and air-only burners adjacent to,
opposite of, and above and below each other; to track the subsequent
19
-------
mixing in the active burner region from these various combinations of
air-fuel ratios; and to estimate the average, effective air-fuel ratios
in the region where the fuels are gasifying and mixing and the fuel-bound
nitrogen conversion is occurring. The minimum of five terms, pre-
viously required, to adequately calculate and correlate thermal NOx
formation in the active burner region could be reduced to a single term,
covering the NOX generated in the active burner region.
The NOX calculation and correlation equation was reduced
to only three terms, plus the necessary constant. One term describes
the NOX formed in the first stage. A second term describes the NOX
formed in the final mixing zone (or second stage), where all combustion
air not entering the furnace through the active burners enters through
air-only burners in the top rows of the burner array or through NOX
ports above the burner array and mixes with the products of combustion
in the first stage. The third term accounted for conversion of fuel-
bound nitrogen.
A major advantage of this simplification is that the first
two terms represent the thermal NOX formed under distinctly different
operating conditions (the first and second stages of combustion), while
the third term represents the NOX independently formed in the first
stage via the conversion of fuel-bound nitrogen. By observing NOx data
variations resulting from major variations in combustion staging or
operating conditions, it appeared more feasible to attempt direct calcu-
lation of thermal NOx in the two major regions and the fuel-bound nitro-
gen conversion and to compare these calculations with the measured
data. Certainly, the coefficients developed from the multiple regression
analyses of the data can be more meaningful with regard to evaluating
the accuracy of assumption and estimates involved in calculation of the
NOX terms in the correlation equation.
In the initial studies [l, 2] , for example, the calculated
value of a given term might represent the NOX generated in all of the
secondary mixing zones immediately downstream of active burners
located immediately above air-only burners. No NOX data are available
for direct comparison with the calculated values for such terms. When
off-stoichiometric burner configurations are in the data sample, it is
necessary to account for these and other complicated mixing zones. The
regression analysis is capable of quantifying the appropriate coefficients,
but it is only the sum of these quantified terms which can be directly
compared to measured NOX data,,
The first major improvement effected in this study, then,
was the decision to make every effort to calculate NOX directly with the
three variable terms and to use the measured data as well as the co-
efficients of these terms (resulting from regression analyses) to evaluate
the calculations and to develop further insight into the effects of certain
hardware and operating conditions on NOX formation in utility boilers.
20
-------
One problem introduced by the simplification of the
NOX calculation was that a great deal of data in some of the samples
used in the previous studies [1,2] resulted from tests of off-
stoichiometric firing configurations, and these data were deleted from
the sample used in this study. Table 1 identifies the boiler types used
in this study, and Table 2 shows the remaining data sample. In a pre-
vious study [ 1 ], 250 and 139 tests on natural gas- and oil-fired boilers,
respectively, were available. Thus, the decision to exclude off-
stoichiometric firing configurations reduced the available natural gas-
fired data sample by more than half but reduced the oil-fired data sample
by only about 29 percent. While the data sample available for the single-
wall, coal-fired boiler type (Table 1, boiler reference no. 5) was
reduced by about half, the data samples for the two tangential boilers
(nos. 6 and 7) were not affected at all. In addition, 52 tests on four
other coal-fired boilers were not studied because of the limited and
somewhat questionable data.
4. 3 IMPROVEMENTS
Major improvements which, within the scope of this
study, could be incorporated in the analysis as a result of the simpli-
fication of the NOX calculation and correlation equation included:
(1) more accurate calculation of the overall boiler air-fuel ratio from
measured flue gas compositions; (2) more detailed calculation of the
NOX formed in the final mixing zone, where all of the combustion air
not entering the furnace through the active burners is mixed (at some
finite rate) with the products of combustion in the active burner region;
(3) more realistic product gas cooling rates; and (4) allowance for finite
rate fuel gasification and mixing. Some consideration was also given
to some additional effects of the combustion process in turbulent,
heterogeneous diffusion flames on the NOX emissions. Each of these
improvements is briefly discussed in the following paragraphs.
4. 3. 1 Boiler Air-Fuel Ratios
In the study of the effects of combustion modifications,
for NOX reduction, on plant efficiency [ 4 ], considerably more atten-
tion -was paid to carbon losses. In the current study, carbon losses are
defined as that carbon in the fuel which cannot be accounted for in the
sum of the measured CO and CO2 in the flue gases.
In the plant efficiency study, these losses represented
direct combustion efficiency losses. Relatively accurate fuel flow rate
measurements were available only for the natural gas- and oil-fired
data. Therefore, plant efficiency was studied only with respect to the
boilers firing these fuels. Carbon losses were not excessively large,
and no significant effects of staged or off-stoichiometric firing config-
urations on these losses could be determined. A great deal of scatter
in the measured values of CO2 and, where data was available, in CO
was observed. This included a significant number of cases where the
measured CO2 levels were higher than theoretically possible for any
level of 03.
21
-------
TABLE 1. BOILERS IN THE DATA SAMPLE
tNJ
Boiler
Ref.
No.
1
2
3
4
5
6
7
Firing
Type
Single
wall
Single
wall
Opposed
Opposed
Single
wall
Tangential
Tangential
Rated
Load, MW
180
240
240
350
125
330
360
No. of
Burners
16
12
12
24
16
20
20
NOX
Ports
No
No
Yes
Yes
No
a
a
Fuels Used
Natural gas,
oil
Natural gas,
oil
Natural gas,
oil
Natural gas,
oil
Nominal coal
High nitrogen
coal
Nominal coal
a. Secondary air port design gives some degree of fixed, effective NOx
port air flow.
-------
TABLE 2. SUMMARY OF THE TOTAL DATA SAMPLE
Boiler
Ref.
No.
No. of Tests
Total
NO Ports Closed
X
0
2
4
8
NO Ports Open
x c
0
2
4
8C
Natural Gas (total tests = 107)
1
2
3
4
8
7
36
56
7
6
6
21
0
1
1
0
1
0
2
0
0
0
0
1
a
a
15
32
a
a
8
1
a
a
4
0
a
a
0
1
Oil (total tests = 49)
1
2
3
4
7
12
6
24
6
10
3
11
0
2
0
0
1
0
0
0
0
0
0
0
a
a
3
13
" a
a
0
0
a
a
0
0
a
a
0
0
Coal (total tests = 99)
5
6
7
37
22
40
17
b
b
11
b
b
9
b
b
0
b
b
a
8
19
a
0
0
a
14
21
a
0
0
a. No NOX ports installed.
Co Number of burners operated air-only,,
b. Some degree of effective NOX port flow.
-------
The measured levels of C>2 and CO2 were used in all
of the previous studies to calculate combustion air flow rates from
measured (or estimated, in the case of coal) fuel flow rates through a
calculation of the overall boiler air-fuel ratio. Simple stoichiometry
was used to derive an expression for the air-fuel ratio which was pre-
dominantly a function of the measured O2 levels. This was certainly
adequate for natural gas- and oil-fired boilers because the carbon
losses, except, perhaps, for O2 levels less than about three percent,
were negligibly small. Such is not always the case with coal fuels.
Figures l(a) through (e) show plots of the measured CO2
and Oz data (volume percent, dry) of this study from the tests of the
boilers fired with natural gas, oil, and two types of coal. The two lines
shown on the figures represent the theoretical CO2/O2 relations for
complete combustion of the carbon in the fuel to CO2 and for a 20 per-
cent carbon loss. The related air-fuel equivalence ratio for complete
combustion of all of the carbon to CO2 is shown across the top of the
figures. The data from boilers firing the nominal coal (type no. 3) is
shown separately for the single-wall boilers (Figure l(c)) and the tan-
gential type (Figure l(d)) because the CO2/O2 data are significantly
different between the two firing types»
• Figure l(a) shows an example of the large scatter in the
data, with 18 data points appearing well above the theoreticalfor com-
plete combustion of all carbon to CO2- Although some of that scatter
could be due to scatter in the C>2 measurements, at least four cases
show higher levels of CO2 than would be possible with any level of C>2.
The majority of the data, however, show a rather regular trend from
complete combustion of all of the carbon to CO2 at O2 levels above
about 4 percent to what would appear to represent as much as a 10 per-
cent loss with O2 levels around 1 percent.
This trend is in the right direction since problems with
excessive CO and smoke begin to appear with C>2 'levels of less than
about 3 percent. In all cases, .measured CO levels were low enough
that the carbon involved would be negligible compared to that in the CO2«
It is usually normal policy, in fact, for the boiler operator to adjust
operating conditions if measured CO exceeds some fixed (low) measured
level. A similar adjustment is usually made, however, only to avoid
clearly visible smoke from the stack.
Figure l(b) also shows some measured data above the
theoretical line for complete combustion of the oil fuel carbon to CO2-
In general, however, the data appear to follow a line which would again
appear to represent about a 10 percent carbon loss. A trend to zero
carbon loss with increasing excess air (O2) is not apparent.
If the data shown in Figures l(a) and (b) truly repre-
sented carbon losses, one would expect that such losses with coal fuels
would be much higher because of the large ash content of the fuel, the
24
-------
13
1.05
AIR-FUEL EQUIVALENCE RATIO
(for complete combustion)
1.10 1.15 1.20 1.30
5
o
\—
ex.
12
11
o
o
in
10
X
o
o
CO
o
O
O
0 COMPLETE
COMBUSTION
20% CARBON LOSS
2345
OXYGEN CONCENTRATION, vol %, dry
(a) NATURAL-GAS-FIRED BOILERS
Figure 1. Carbon dioxide and oxygen concentrations
measured in the flue gases. (Sheet 1 of 5. )
25
-------
1.05
16
I15
S 14
o
o
o
13
X
o
o
CO
o:
o
12
11
AIR-FUEL EQUIVALENCE RATIO
(for complete combustion)
1.10 1,15 1.20 1.30
COMPLETE
COMBUSTION
20% CARBON LOSS
2345
OXYGEN CONCENTRATION, vol %, dry
(b) OIL-FIRED BOILERS
Figure 1. Carbon dioxide and oxygen concentrations measured
in the flue gases. (Sheet 2 Of 5. )
26
-------
1.05
18
17
z 16
o
I
I—
LU
O
o
o
15
I 14
o
o
00
o
13
12
AIR-FUEL EQUI VALANCE RATIO
(for complete combustion)
1.10 1.15 1.20
1.30
\
O
o
COMPLETE
COMBUSTION
O
1
20% CARBON LOSS
O
I
O
2345
OXYGEN CONCENTRATION, vol %, dry
(c) TWO SINGLE-WALL BOILERS
FIRING A NOMINAL COAL
Figure 1. Carbon dioxide and oxygen concentrations measured
in the flue gases. (Sheet 3 of 5.)
27
-------
o
<
O
O
X
O
17
16
15
14
13
o
CO
Si 12
11
AIR-FUEL EQUIVALANCE RATIO
(for complete combustion)
1.15 1.20 L30 1.40
T
I
o
o
COMPLETE
COMBUSTION
— OO
O
O
20% CARBON
O
O
O
O
O
O O
O
3456
OXYGEN CONCENTRATION, vol %, dry
(d) 360-MW TANGENTIAL BOILER
FIRING A NOMINAL COAL
Figure 1. Carbon dioxide and oxygen concentrations measured
in the flue gases. (Sheet 4 of 5.)
28
-------
18
1.05
o
<
I—
UJ
O
o
o
16
15
X
o 14
O
o
CO
on
5 13
12
AIR-FUEL EQUIVALENCE RATIO
(for complete combustion)
1.10 1.15 1.20 1.30
O
O O
o
20% CARBON
LOSS
8
COMPLETE
COMBUSTION
O
O O
°0
O
o
o
o
o
2345
OXYGEN CONCENTRATION, vol %, dry
(e) 330-MW TANGENTIAL BOILER
FIRING A HIGH-NITROGEN COAL
Figure 1. Carbon dioxide and oxygen concentrations measured
in the flue gases.(Sheet 5 of 5.)
29
-------
slagging, and the high particulate emissions associated with coal
burning. Figure l(c) shows the COz/C>Z data from two boilers with
single firing walls firing a nominal type of coal containing 11 percent
ash. These data also show a considerable amount of scatter, but the
general trend could again be interpreted to indicate 10 to 15 percent
carbon loss.
Figure l(d) shows data from a tangential boiler firing the
same type of coal fired in the single-wall boilers. Here the data are
widely scattered and well below the 20 percent carbon loss line. Figure
l(e) shows data from a similar tangential boiler firing a coal type con-
taining higher concentrations of fuel-bound nitrogen but the same
fractions of ash and moisture. Again the data are widely scattered,
but in this case the data are about as much above the zero to 20 percent
carbon loss lines as below.
The causes of the large data scatter observed in Figures
l(a) through l(e) are not known. Carbon losses of as much as 10 percent
seem possible in coal-fired boilers but do not seem likely with natural
gas and oil fuels. It may be possible that operation at very low levels
of excess air, for special, short-term NOX emissions testing, could pro-
duce higher carbon losses and CO than is considered normal. The cal-
culated CO2/O2 relation for complete combustion shown in Figure l(c)
duplicates that indicated by Crawford [6] for this fuel.
It seems most likely, therefore, that the data scatter and
the very low CC>2 levels measured with the coal fuels are due to some
kind of stratification of the flue gases, relative to the sampling devices,
yielding lower than the actual average levels of one or all of the COz»
OZ, or CO. Some cases were actually noted in the data where different
sampling points yielded CO measurements differing by 1000 to 2000 ppm.
This, then, was the general assumption made for this study.
If carbon losses were significant in some tests, the major
effect on the calculations of this study would be in the total air flow rate
through the boiler. A measured level of QZ in the flue gases, with some
of the carbon lost to the combustion process, would indicate a lower
input air-fuel ratio than if the carbon combustion were complete.
With regard to calculation of the combustion temperature
rise, however, if the unburned carbon were assumed to pass through the
combustion process like inert ash, being heated in initial combustion
but then giving up this heat as the particles cool, then the carbon loss
could essentially be neglected. Most combustion calculations neglect
these effects. In any case, more detailed accounting of carbon losses
in the local combustion processes would require knowledge of the carbon
losses (or unburned carbon) at all points in the boiler.
Therefore, for these calculations, carbon losses as indi-
cated by the COz/Oz data were taken into account only in the calculation
30
-------
of the overall boiler air-fuel ratio and, hence, in the total combustion
air flow rate. Even here, the measured CO2 data were bounded. The
measured CO2 data were tested against upper and lower limits corres-
ponding to complete combustion and 10 percent carbon loss, for the
measured 03 level. If the measured CC>2 was higher or lower than
these limits they were assigned values equal to the nearest limit level.
For calculation of the air-fuel equivalence ratio and hence the equil-
ibrium temperature rise, complete combustion of all carbon in the fuel
to CO2 was assumed (only the measured O2 level was used).
Although the data scatter in all samples is large, account-
ing for possible carbon losses as high as 10 percent, at least in determin-
ing the total boiler air flow rate, is considered an improvement over the
usual process of neglecting carbon losses altogether. This makes little
difference in the natural gas- and oil-fired data, but the coal data show
much stronger evidence of carbon losses. Of course, the large scatter
attributed to the CO2 data could also represent equally large scatter in
the Q£ measurements, possibly because of the same stratification prob-
lems, but no independent check on O2 measurements was available.
Obviously, large scatter in the 02 data will reflect directly in large
scatter in calculated NOX levels.
4. 3. 2 The Final Mixing Zone
In the study of NOX emissions from coal-fired boilers
[ 2 ], it was observed that nearly all of the NOX emissions from coal-
fired boilers resulted from the conversion of fuel-bound nitrogen and
from thermal NOx generated in the final mixing zone. The final mixing
zone was defined as that region in the furnace where all of the combustion
air not entering the furnace through the active burners was mixed with
the products of (in some cases, fuel-rich) combustion in the active burner
region. This was particularly true if the data from off-stoichiometric
configurations were excluded. In that study, the calculation of NOX
from the final mixing zone was taken as a simple constant for a given
boiler and fuel. That constant was an empirical function only of the
peak temperature of the product gases entering that zone.
The data indicated that in staged combustion configurations,
where the first stage (active burner region) air-fuel ratio was below
stoichiometric, mixing of the remaining combustion air with the products
of fuel-rich combustion from the active burner region was a relatively
slow process; during this process the average air-fuel ratio in the zone
passed through stoichiometric. Modeling of the entire final mixing zone
as a single tank (instantaneous mixing)-and-tube missed the significant
quantities of NOX which could be generated while the air-fuel ratio was
in the region of stoichiometric, where temperatures and the NOx for-
mation rate (exponential in temperature) are highest.
Of course, there is considerable cooling of the com-
bustion gases within the active burner region before these gases reach
31
-------
the final mixing zone. These first stage product gases, in all cases,
represent more than about three-quarters of the total flow through the
boiler and contain all of the fuel. Even though there is further re-
action between the first stage products and the remaining excess air in
the final mixing zone, the first stage gases could have cooled sufficiently
that negligible NOX would be formed in this final mixing zone. Thus
the initial combustion air temperature, the temperature rise due to com-
bustion in the first stage, and the degree of cooling of the first stage
product gases enroute to the final mixing zone critically determine the
amount of NOX formed in the final mixing zone. Because the NOX
formation rate is exponential in temperature, accurate calculation of
the gas temperature up to the final mixing zone is important.
In the study of coal-fired boilers [2] , the limited scope
allowed only an empirical estimate of the effects of finite rate mixing
in this zone to be developed. This provided at least a first-order
correction for the single tank-and-tube model of this zone. In this study,
the final mixing zone was divided into ten tank-and-tube subzones,
in each of which one-tenth of the excess air was mixed with the products
of the previous zone. In each of these subzones, the increment of air
was mixed instantaneously with the products of the previous subzone, and
a finite time was allowed for flow to the next subzone and for NOX gen-
eration. The total time for this mixing was taken as the time required
for flow of the combustion gases: (1) from the top level of the active
burners to the top level of air-only burners (if any); (2) from the top
level of air-only burners to the level of the open NOX ports (if any);
and (3) one-quarter of the furnace hydraulic diameter beyond the NOX
ports. In addition, the NOX subsequently formed in the fully mixed
product gases was included, to the point where the gases had cooled
sufficiently that the NOX formation rate became negligible.
Preliminary parametric calculations with this improved
description of the final mixing zone showed, as expected, that, with
constant overall boiler operating conditions, the NOX formed in the
final mixing zone increases to a maximum as the air-fuel ratio of the
first stage is decreased. This is because significant NOX is formed
only during the time period when the effective air-fuel ratio of the
product gases is near stoichiometrico With a very fuel-rich first stage,
negligible NOX is formed in the final mixing zone until sufficient ad-
ditional air has been mixed with these fuel-rich products to bring the
effective air-fuel ratio near stoichiometric. The highest first stage
air-fuel ratio where maximum NOX is formed in the final mixing zone,
and the subsequent total NOX formed in the whole final mixing zone, are
strongly dependent on the temperature of the fuel-rich products coming
from the first stage. Given sufficient time for cooling and/or a suf-
ficiently high rate of cooling, the thermal NOX formed when the final
excess air is mixed with the products of first stage combustion can be
reduced to negligible levels.
32
-------
There is some conjecture among researchers that, with
very fuel-rich first stage combustion of a fuel containing significant
concentrations of chemically bound nitrogen, some of the fuel, still con-
taining some of the fuel-bound nitrogen, may remain unburned through
the first stage. The fuel-bound nitrogen in this unburned fuel, then,
could be converted to NOX in the final mixing zone (during second stage
combustion). It is clear that, if the first stage were operated so fuel-
rich that no combustion could take place, the second stage would become
the first stage of combustion and all fuel-bound nitrogen conversion
would take place there. With stable combustion occurring in the first
stage, however, it is not clear just how fuel-rich the first stage would
have to be before significant fuel-bound nitrogen conversion begins to
occur in the second stage. Such conversion, however, if it does take
place in the second stage, might generate considerably more NOx be-
cause of the greater availability of oxygen in the second stage of com-
bustion.
Evidence of this phenomenon might be indicated by overall
NOX emission data which decrease monotonically with the first stage
air-fuel ratio until a minimum is reached. Further reduction in the
first stage air-fuel ratio, then, would show a rather rapid rise in total
NOX emissions. Some data (unpublished) exhibiting these character-
istics have been derived in laboratory experiments. It is not yet clear,
however, whether the rise in NOX levels at very low first stage air-fuel
ratios is due to conversion of fuel-bound nitrogen in the second stage or
to the increased formation of thermal NOX in the final mixing zone, as
discussed in this section. The data from full-scale boilers used in this
study do not indicate this phenomenon. Whether this is because these
boilers were never operated sufficiently fuel-rich in the first stage or
because the phenomena cannot exist in full-scale, operating boilers is
not known. In any case, no attempt was made in this study to allow for
any conversion of fuel-bound nitrogen in the final mixing zone.
4. 3. 3 Gas Cooling Rates
In the previous studies of NOX formation in utility boilers
[ 1, 2] , uncertainties in the complex flow, mixing and reaction processes
occurring in full-scale, multiburner boilers, particularly when operated
in off-stoichiometric configurations, indicated that nothing more complex
than a constant time-rate of cooling of the combustion products could be
justified. As a result of the large simplification afforded by eliminating
off-stoichiometric configurations from the analysis and the data sample,
it was considered that the next more accurate step could be taken in the
description of combustion product gas cooling. Since the great majority
of the heat transferred to the water walls in a typical utility boiler is
transferred by radiation, a cooling rate in some way proportional to the
fourth power of the local gas temperature seemed appropriate.
Again, the limited scope of this study and the remaining
degree (or lack) of sophistication in the rest of the analysis did not allow
33
-------
development of a comprehensive radiation cooling calculation or sub-
routine for these complex, full-scale boilers. A limited amount of
measured data on heat fluxes to the water walls of an opposed, natural
gas-fired boiler were available [ l] . These data showed measured
heat fluxes at six vertical locations in the boiler and measured gas tem-
peratures at one location, with the boiler operated at three load
conditions. From radiation heat transfer theory, variations in this
heat flux data should follow an expression in the fourth power of tem-
perature, such as
Q = K (AT 4 - BT 4) (1)
r v g i w ' v '
•
Plots of the heat flux data Q against the measured gas
temperatures Tg did indeed indicate an approximate fourth-power
dependence. Several attempts to fit the data directly with equation (1),
using reasonable estimates for the water-wall temperatures, resulted
in the conclusion that BT,^'*, must be negligibly small compared to
ATg . Therefore, all of the heat flux data was adjusted by adding a
small, constant heat flux (of 2 to 12 percent of the measured heat flux
data) such that a fit of the data with equation (1) yielded a value for this
second term which was exactly zero. This left the expression
Q = K T 4 (2)
r g
•
where Q is the heat/area-time, Tg is the temperature, and Kr was
empirically determined as
Q = Kr = 6 g<7 = . 374
-------
height, equation (2), was equated to a uniform heat loss from the com-
bustion gases as they crossed that increment of height. This yielded a
time rate of cooling given by the expression
= -0.04 R K T (4)
dt sv r g
where R is the surface-to-volume (circumference to flow area) ratio
of the furnace.
In each stream tube of the tank-and-tube mixing scheme,
the final temperature Tf was calculated from the initial temperature T^
from equation (5)
* 4
= ' + °' 08RsvKrTi \ <5>
where tf is the time required for flow through the stream tube.
Similarly, in each stream tube (at a constant air -fuel
ratio), equation (4) can be used to convert the Arhennius rate expression
for the formation of thermal NOx (based on the Zeldovich mechanism)
from a function of time to a function of temperature, and integrated over
the stream tube from the initial to the final temperature,, The resulting
expression for the increment of NOX (dNOx) formed in a given stream
tube is
= 5. 645
R K
sv r
e"Z (Z3 + 3Z2 + 6Z +6)
Zf
(6)
where Z = 67, 900/Tg (7)
and the term SON represents a product of the nitrogen/oxygen concen-
trations appropriate to the Zeldovich NOX formation rate equation,
1/2
[N2
Equations (5) and (6) combine the effects of time and
temperature on thermal NOx formation. If gas temperatures are very
high, the rate of cooling, equation (5), is very rapid, and the final tem-
perature can be much lower than the initial temperature, even if the
time required to flow through the stream tube is short. Large differences
in initial and final temperatures yield, equation (6), large increments of
NOX formed in the tube. The same amount of NOx could be formed in
the tube with a much lower initial gas temperature if the time to flow
through the tube is sufficiently long.
Equation (6) also shows the direct effect of the boiler
surf ace -to -volume ratio RSV and the rate of radiant heat transfer Kr
on thermal NOX formation. Small boilers, with larger surface -to -volume
35
-------
ratios, should generate less NOX emissions than the larger boilers.
Similarly, increasing the rate of radiant heat transfer (the effective, net
emissivity) will also reduce NOX emissions.
In the active burner region (because the large local mix-
ture ratio variations resulting from off-stoichiometric firing configu-
rations have been deleted), the mixing in the tank at the beginning of
each tank-and-tube section is between streams o'f similar composition;
however, the stream temperatures may be different. Once the initial,
mixed temperature is established, the temperature at the end of the tube
can be calculated from equation (5), and these two temperatures can be
used in equations (6) and (7) to calculate the increment of NOx generated
in. that stream tube. The total NOX formed in the first stage, then, is
the sum of these increments along each flow path through the active
burner region.
4. 3. 4 Finite Rate Gasification and Mixing
In the previous studies [ 1, 2] it was necessary to neglect
the effects on NOX formation of finite rate gasification of the oil and
coal fuels and of the mixing of natural gas and the gasified oil and coal
with combustion air. A preliminary model of such gasification and mix-
ing was developed during the course of those studies, the latter based on
a relaxation technique analogous to transient conductive heat transfer
analyses. The resulting calculation was too detailed and complex for the
NOX emissions analysis used at that time and, despite the simplifications
in this study, it is still too complex to be directly incorporated. There
are no inherent limitations to incorporation, but it was beyond the scope
of both the previous and this study.
That peripheral analysis was used generally, however,
to investigate the effects of finite rate gasification and mixing on local
air-fuel ratios and temperatures. The analysis was also used to develop
a simpler (again first order) means of accounting for these rates, com-
patible with the scope of this study as well as with the degree of sophisti-
cation of the rest of the NOX emissions analysis.
Results of these peripheral investigations, as well as
other considerations of heterogeneous diffusion flames, indicate a com-
bustion process wherein essentially pure fuel or fuel vapor is mixed
with essentially pure combustion air, with the local air-fuel ratios
ranging from zero to infinity. Initially, a very steep air-fuel ratio grad-
ient exists between the sources of unmixed, gaseous fuel and the surround-
ing combustion air. As gasification and mixing proceeds, all of the fuel
passes first through a stoichiometric air-fuel ratio region (the so-called
wrinkled flame front); the products of stoichiometric reactions are then
diluted with additional combustion air.
In cases where the total amount of combustion air avail-
able in the active burner region exceeds the stoichiometric air, the
36
-------
mixing process proceeds until all ol" the fuel is gasified. The region
of stoichiometric. combustion products then approaches the fuel source
more closely and finally disappears as the entire mixture approaches
the products of combustion at the overall (excess air) air-fuel ratio. In
cases where the overall average air-fuel ratio in the first stage is less
than stoichiometric, a point is reached where the amount of fuel gasified
exceeds the stoichiometric fuel for the available combustion air. At
this point, the region of stoichiometric combustion (the flame front)
expands away from the fuel source as the as yet unmixed air is over-
diluted with products of fuel-rich combustion. Again, the stoichiometric
region finally disappears as the entire mixture approaches the products
of combustion at the overall (excess fuel) air-fuel ratio.
Results of the preliminary mixing studies indicated that
a first order approximation of this finite rate gasification and mixing
process, for the purpose of NOX formation, could be developed by
defining three parallel and simultaneous zones within the flows from the
active burners: (1) a zone where the fuel either has not gasified or the
air-fuel ratio is so fuel-rich that no NOX could be formed; (2) a zone
where the average air-fuel ratio is near stoichiometric (the flame front);
and (3) the rest of the products not included in the first two zones.
A relatively simple calculation was set up, therefore,
composed of these three zones. The rate of gasification and mixing was
arbitrarily taken as a linear function of distance into the furnace. With
natural gas fuel, the distance for final, complete mixing is that required
for mixing only. The probable distance for complete mixing of gaseous
fuel with the combustion air was selected from considerations of the
analytical mixing results and empirical observations of flames in full-
scale utility boilers. With natural gas fuel, the mixing calculations in-
dicated that mixing should be complete within the core flow from the
active burner (less than two diameters from the burner exit). Obser-
vations of flames in coal-fired boilers indicate that the sum of the
distance required to gasify and mix these fuels is nearly the total distance
from the burner to the opposite wall of the furnace in single-wall boilers,
or to the centerline in opposed-fired boilers. Since impingement of
reacting coal particles or partially mixed gases on a water wall can have
very corrosive effects, it was finally assumed that both gasification and
mixing of the coal particles and the derived fuel vapors are (just)
complete at the opposite wall (or centerline). The appropriate distance
for complete vaporization and mixing of oil droplets and derived fuel
vapors lies between the distances established for the natural gas and
coal cases. Some further parametric variations were conducted with
oil fuels to better establish the appropriate distance.
In addition, because some fuel vapor must be available
and mixed with some air right at the burner exit to provide continuous
ignition and flame-holding, it was assumed that 20 percent of all fuels
was already gasified and mixed, in stoichiometric proportions, at the
37
-------
burner exit. In the cases of oil and coal fuels, this initial combustion
could be provided by the very fine oil droplets and coal particles in
the initial size distribution.
In each of the tank-and-tube zones describing the burner
flows in the overall NOX model, then, the quantity of fuel and air and
the average air-fuel ratio in the three subzones was determined by cal-
culating: (1) the amount of the fuel as yet ungasified (zero air-fuel ratio);
(2) a fixed fraction of the gasified fuel plus the related stoichiometric
proportion of combustion air (stoichiometric air-fuel ratio); and (3) all
of the rest of the combustion air and gasified fuel, in whatever air-fuel
ratio results.
The effective fraction of the gasified fuel in the subzone
at stoichiometric air-fuel ratio, of course, is not a commonly used or
well-established quantity. If mixing rates are very rapid compared to
gasification rates, the flame fronts might be expected to be very thin
and close to the gasifying fuel sources. Since natural gas is already
gasified, the flame fronts might be expected to be more wrinkled, broad-
er, and more diffuse. For this study, from considerations such as
this, the fraction of the gasified fuel which, at any instant in time, is
involved in reactions (including NOx formation) typical of stoichiometric
combustion was taken as 20 percent for all coal fuels and 30 percent for
natural gas. Again, this fraction for oil fuels should be intermediate
between those for gas and coal. Parametric calculations appeared to
indicate that 24. 5 percent was appropriate for oil.
The simple, finite rate gasification and mixing rate cal-
culation described in this section is capable of approximately modeling
the progress of gasification, stoichiometric combustion, and subsequent
dilution of gaseous, liquid and solid fuels to completion. Initially, 80
percent of the fuel is not yet gasified (or is not yet mixed with any air, in
the case of natural gas), 20 percent of the gasified fuel is mixed with
a stoichiometric proportion of air, and the rest of the air is unmixed with
any fuel. As the element of fuel and air proceeds into the furnace, more
fuel is gasified; more fuel is involved in stoichiometric combustion (up
to the limiting fractions discussed above); some products of stoichio-
metric combustion begin to be mixed with the remaining air; and the
air-fuel ratio in the latter zone begins to decrease.
In the case where the available combustion air is in excess
of the stoichiometric air, this process of gasification and mixing continues
until all of the fuel is gasified. Subsequent mixing then increases the
air-fuel ratio in the subzone, formerly at stoichiometric, and continues
to decrease that in the remaining zone until they both become equal to
the overall air-fuel ratio, at the (specified) end of the gasification and
mixing distance. In the case where the available combustion air is less
than the stoichiometric ratio to the fuel, this process of gasification and
mixing proceeds as in the previous case until the air-fuel ratio in the
third subzone decreases to stoichiometric. At that point, there is not
38
-------
sufficient air remaining to sustain a region of stoichiometric combustion,
and the air-fuel ratios in both zones decrease to the final, fuel-rich
ratio at the end of the gasification and mixing distance.
The continuous gasification and mixing approximation
described above was converted into finite steps and averaged over the
length of each of the stream tubes in the overall tank-and-tube model
of the burner flows. In effect, the series primary, secondary, adjacent,
and opposite tank-and-tube mixing zones [ 1 ] were each further sub-
divided into these three effectively parallel tank-and-tube mixing sections.
At the beginning of each of the series mixing zones, the gasification and
thermal and composition mixing appropriate to the three parallel sub-
zones were accomplished (instantaneously, in the tank) to establish the
initial conditions for flow through the three parallel subzones in the next
mixing zone in the series. Cooling and NOX formation were calculated
in each of these parallel subzones by the technique discussed in Section
4.3.3.
The somewhat complicated parallel-series gasification
and mixing scheme discussed in this section is important only in the
active burner flow streams in the lower part of the active burner region
of the furnace. Since little bulk gas flow has yet been generated, these
processes can proceed largely as idealized. In the upper parts of the
active burner region, the gas flow and mixing picture becomes increas-
ingly dominated by the bulk gas flow coming from the completed reactions
in the lower parts. Not only is the composition of this bulk gas flow
that of complete gasification, mixing, and combustion at the overall air-
fuel ratio of the first stage of combustion (in some cases quite fuel-rich),
but these gases have been in the furnace for some time and have cooled
considerably. Some cool, recirculated flue gases may even be mixed
in with these bulk gases. In this study, no attempt was made to further
model finite rate gasification and mixing of any of the fuels in the
presence of a significant cross-flow of bulk gases. Instead, as in the
previous studies, the fraction of the vertical flow area of the furnace
occupied by bulk gas flow at any burner level was linearly proportioned
to the amount of active burner flow already in the furnace. In the upper
burner levels, when the flow from an active burner intersects the bulk
gas flow, all gasification and mixing of the fuel and air within the burner
flow was assumed complete and mixed with the bulk gas flow. At all
burner levels, the finite rate gasification and mixing scheme discussed in
this section is carried out at least within the region of core flow (the
primary mixing zone) at the exit of the burner.
Although the finite gasification and mixing rate approxi-
mation discussed in this section (and used in all NOX emissions analyses
performed in this study) appears somewhat complicated and contains
two somewhat arbitrary constants, it does provide further insight into
the NOx formation process in full-scale, multiburner utility boilers.
Furthermore, no other technique appropriate to the scope of this study
is available. The two constants can be related to droplet and particle
39
-------
size distributions in the initial oil sprays and pulverized coal streams.
Large average droplet and particle sizes will extend the gasification
distance further into the furnace and may also decrease the thickness
of the stoichiometric flames surrounding the particles. Thus, incorpor-
ating this gasification and mixing calculation into the overall analysis
of NOX formation provides a qualitative means of investigating the
effects of initial particle size distributions on NOX emissions. Para-
metric variations of these constants, discussed in the Results, Section
V, of this report, show that increased particle sizes can reduce NOx
formation in the early regions of the burner flows but also can decrease
the cooling rate in these early regions, resulting in greater NOx for-
mation further downstream.
As often happens, however, improvements in accuracy
or detail in the description of one phenomenon open the door to questions
in others. In this case, the improvements in the description of finite
rate gasification and mixing introduce further questions concerning
the conversion of fuel-bound nitrogen to NOX in the stoichiometric
flame and the possible oxidation or reduction of this initial NOx in the
subsequent mixing of the products of stoichiometric combustion with the
remaining excess air or fuel. This question is discussed further in
Section 4. 3. 5. From the standpoint of the calculation used in this study,
finite rate gasification and mixing in no way affects the efficiency of
conversion of fuel-bound nitrogen to NOx as calculated with the model
developed and used in the previous studies [ 2 ].
4. 3. 5 Some Considerations Relative to Heterogeneous Flames
The more detailed modeling of heterogeneous diffusion
flame processes discussed in the previous section opens the door to
further questions regarding: (1) the mechanism of the initial conversion
of fuel-bound nitrogen to NOX; and (2) the possible subsequent effects of
excess carbon, unburned fuel, and CO on this initially formed NOx.
The three-step process of combustion discussed in Section 4. 3. 4 relative
to finite rate gasification, mixing, and combustion in turbulent homo-
geneous (natural gas) and heterogeneous (oil and coal) diffusion flames
emphasizes that all of the fuel originates from a very fuel-rich source
(concentrated gaseous natural gas, liquid oil, or solid coal) and that,
over much of the gasification and mixing time, initial, local combustion
of the gasified and/or mixed fuel is under high-temperature, near-
stoichiometric conditions. The products of this stoichiometric com-
bustion are then mixed with the remaining air. Only after most or all
of the fuel has been gasified and/or mixed is there insufficient air or
fuel (overall fuel-rich or fuel-lean, respectively) to sustain this initial,
local stoichiometric combustion. The two questions raised when turbu-
lent heterogeneous diffusion flames are viewed in this manner involve:
(1) the air-fuel ratio conditions under which the fuel-bound nitrogen is
converted to NOX (not a question with natural gas fuels, which contain
no fuel-bound nitrogen); and (2) what happens to the very high levels of
40
-------
NOX formed in this stoichiometric region upon subsequent dilution
by the remaining fuel or air (overall fuel-rich or fuel-lean respect-
ively).
Many laboratory studies of the global effects of the overall
air-fuel ratio on the efficiency of conversion of fuel-bound nitrogen to
NOX, using homogeneous, premixed flames, show a clear decrease in
this conversion efficiency as the air-fuel ratio is decreased. Detailed
probing of this type of flame appears to indicate that reactions which
convert the fuel-bound nitrogen to NOX occur as fast and in approxi-
mately the same physical location in the flame as those which consume
the hydrocarbons in the fuel. One might assume, then, that in a heter-
ogeneous diffusion flame, where the initial hydrocarbon reactions gen-
erally occur under fuel-rich, near-stoichiometric conditions, the
efficiency of conversion of fuel-bound nitrogen to NOX would always be
similar in magnitude to conversion in a premixed gaseous flame under
near-stoichiometric conditions. Further, one might also expect that
the fuel-bound nitrogen conversion efficiency in a heterogeneous diffusion
flame should be largely independent of the overall air-fuel ratio. In-
stead, the fuel-bound nitrogen conversion efficiency observed in labor-
atory heterogeneous diffusion flames and in full-scale, multiburner
boilers appears to duplicate, in both magnitude and in variation with
the overall air-fuel ratio, that observed in laboratory premixed gaseous
flames.
It is not reasonable to postulate, for example, that com-
bustion of pulverized coal in a full-scale boiler proceeds like a premixed
gaseous flame. In the slow gasification and mixing processes surround-
ing heterogeneous combustion, it seems much more reasonable to
assume that the chemical reactions are much faster than these physical
processes and that the resulting combustion products subsequently main-
tain a reasonable shifting equilibrium as mixing proceeds.
Any view of the conversion of fuel-bound nitrogen in a
heterogeneous flame must recognize that, in the very fuel-rich region
between the fuel source (for example, a coal particle) and the surround-
ing flame, nitrogen can be present, both chemically bound in the gaseous
hydrocarbon fuel and as molecular nitrogen from the surrounding air.
Experimental evidence clearly shows, however, that oxidation of at
least some significant fraction of fuel-bound nitrogen is rapid in this
region, while oxidation of the nitrogen from the air is negligible. This
implies that, in the mechanism through which at least a portion of the
fuel-bound nitrogen is eventually oxidized to NO, the fuel-bound nitrogen
may pass through a series of C-H-N-O intermediates (such as HCN and
NCO) but may never appear as atomic or molecular nitrogen. In the
latter form, it would be indistinguishable from the nitrogen in the air.
In fact, that portion of the fuel-bound nitrogen which does not eventually
appear as NOX must have appeared very early as some stable nitrogen
compound, such as molecular nitrogen, because it was subsequently
able to survive passage through the surrounding hot, stoichiometric
41
-------
flame and, in some cases, even a hot oxygen-rich environment, with-
out appreciable oxidation to NO. The question, of course, concerns
why some of the fuel-bound nitrogen appears to follow the chain to NO,
while the rest does not.
The chemical form of the C-H-N-O compounds in the
fuel appears to have little effect on the fraction of nitrogen converted to
NO. In addition, in heterogeneous combustion, the initial oxidation of
the carbon and hydrogen in the C-H-N molecules must be largely com-
pleted in the region between the fuel source and the surrounding stoich-
iometric flame; therefore, any oxidation of the fuel-bound nitrogen
must also be completed in this region. These two observations suggest
that: (1) initially all of the fuel-bound nitrogen is oxidized to NO,
during the oxidation of carbon and hydrogen; and (2) some fraction of
this NO is subsequently reduced to other, stable forms of nitrogen
(such as Nz)« This latter step would result because the initial oxidation
reactions always occur in a fuel- rich (oxygen-limited) mixture, in a
heterogeneous flame.
Following this simple view of the process of conversion
of fuel-bound nitrogen, close to the fuel source (where little oxygen is
available), little of the C-H-N oxidation reactions could occur, and
little NO would be formed. As more oxygen is mixed with the fuel,
closer to the stoichiometric flame, more NO would be formed, during
the fast, highly exothermic hydrocarbon reactions, until a local air-fuel
ratio is reached where there is just sufficient oxygen available to oxidize
all cf the fuel to CO, H2O, and NO (called here the CO- stoichiometric
air -fuel ratio).
If no further oxygen were supplied, the slower shift
reaction
CO + H2O-*CO2 + H2 (8)
would begin shifting toward the water-gas equilibrium. This reaction
requires no further oxygen (nor frees any) and is very nearly thermo-
dynamically balanced. Essentially, all of the free oxygen is locked up
in the water-gas equilibrium and in the NO and
The mechanism by which the water-gas equilibrium is
achieved can essentially be described in two steps
CO + OH-*CO2 + H (9)
H + H2O->OH + H2 (10)
These are a pair of radical -shuttling reactions (H and OH). Reaction
(9) is fast, and its rate constant is quite insensitive to temperature.
However, at the relatively low temperatures of the very fuel-rich mix-
tures being discussed here, reaction (10) may be quite sluggish.
42
-------
During the shift from the initially high concentrations of CO and H2O
to the equilibrium mixtures (involving higher CO^ and H£ concentrations),
OH concentrations may be depleted and H concentrations increased by
reaction (10). This may result in very low, nonequilibrium levels of
OH and correspondingly high levels of H during the period when the
water-gas reaction is shifting to equilibrium.
One of the four reactions basic to the extended Zeldovich
(thermal) mechanism for NOx formation and destruction that is partic-
ularly important in fuel-rich mixtures is
H+NO^OH+N (11)
Clearly, if the OH levels are reduced and the H levels increased by the
dominant water-gas shift reaction, then reaction (11) will be forced to
the right, resulting in destruction of the NO.
In effect, the initially high CO concentrations require
oxygen in order to reach equilibrium CO2 levels; at the lower tempera-
tures where reaction (10) is sluggish, it may be easier to obtain the
oxygen from the NO via the forward reaction (11) rather than from the
water via reaction (10). The competition for oxygen, then, in an oxygen-
limited mixture, is one between the forward reactions represented in
reactions (10) and (11) (both endothermic) when reaction (8) is seeking
the water-gas equilibrium. This competition would ensure that, as long
as the combustion is locally oxygen-limited, only a fraction of the avail-
able fuel-bound nitrogen would end up as NO, while the remaining
fraction would be converted back to some other nitrogen compound such
as molecular nitrogen. The latter might be thought of essentially as
running the Zeldovich mechanism in reverse. Once this fractional con-
version of the fuel-bound nitrogen is accomplished, the NO could be
further reduced, but the fraction of the fuel-bound nitrogen already
converted to other stable nitrogen compounds could not be easily oxidized
to form more NO, except via the kinetically slow thermal mechanism.
In overall air-fuel mixtures containing excess air, how-
ever, the more volatile compounds will eventually be driven off, and
the gasification rate will become slow compared to the local gas mixing
rates. The char particle temperature may then rise, and further gasifi-
cation and direct char combustion could occur, in the more highly oxi-
dizing local conditions of excess air. Thus, under conditions of excess
air, additional fuel-bound nitrogen, otherwise trapped in the char,
could be converted to NO.
In overall air-fuel mixtures containing excess fuel, the
products of stoichiometric combustion (including the fraction of fuel-
bound nitrogen converted to NO but no C-H-N species) would at first be
diluted with excess air, but then, eventually, as gasification and mixing
near completion, they would be diluted with the products of fuel-rich
43
-------
combustion. At no time in the life of a fuel particle would the air-fuel
ratio at or very near the particle surface be higher than stoichiometric,
and the latter stages of char particle combustion would be under fuel-
rich conditions. Further, assuming that the water-gas equilibrium is
maintained as the mixture becomes more fuel-rich, more hydrogen
would be generated and the NO formed in the earlier stoichiometric
combustion could be further reduced, perhaps to other nitrogen com-
pounds such as molecular nitrogen.
In the case of staged combustion, with a fuel-rich first
stage, the processes described in the preceding paragraph would finally
again be reversed as the second-stage excess air is mixed into the
products of the fuel-rich first stage. The first-stage products would
contain: (1) the NO resulting from the initial conversion of the fuel-bound
nitrogen to NO and the subsequent partial reduction of that NO; (2) no
additional gas-phase fuel-bound nitrogen; and (3) some nitrogen still
chemically bound in the char (neglecting, for the moment, any NO
thermally formed in the first stage). Therefore, any additional fuel-
bound nitrcgen conversion which might occur in the second stage would
have to come from further gasification and/or direct oxidation of the
remaining char particles.
Much of the discussion in this section has been specu-
lative. Alternative mechanisms concerning fuel-bound nitrogen con-
version in heterogeneous flames, however, are not only lacking but
those available are also largely speculative. The vast majority of the
pertinent experiments and proposed mechanisms reported in the lit-
erature concern homogeneous, premixed flames. Similar pertinent
experiments with heterogeneous flames usually describe only the effects
of certain combustion conditions on the resulting overall fuel-bound
nitrogen conversion efficiencies. Some recent work, however, can be
interpreted as supporting the mechanism described in this section.
Aronowitz and Classman [ 7 ] , in studying the oxidation
of methanol, report that carbon dioxide does not begin to form sub-
stantially until the methanol is depleted. They suggest two reaction
steps, a first in which methanol forms carbon monoxide and a second
in which the carbon monoxide is subsequently oxidized to carbon dioxide.
Lewis 8 , in developing a model for entrained flow (particle) gasifiers,
assumed that the heterogeneous reactions generate carbon monoxide,
which is subsequently converted to carbon dioxide in the homogeneous
chemistry via the OH radical originating from water.
Song, Beer, and Sarofim [9] used a simplified scheme
for the conversion of nitrogen chemically bound in coal in which the
conversion to NOX of nitrogen in the volatiles and in the char are separ-
ate and independent processes. Their results show that the conversion
efficiency of both are functions of the fuel/oxygen equivalence ratio
and that both approach zero (for the volatiles) or become asymptotic
44
-------
to very low values (less than 15 percent) for the char nitrogen at fuel-
air equivalence ratios of about 1.7 to Z. 0. This is the approximate
ratio where just enough oxygen is available to initially oxidize all of
the carbon and hydrogen to carbon monoxide and water, with little
remaining for oxidation of the fuel-bound nitrogen.
Some effort was made during this study (unsuccessfully)
to further develop and quantify the serial process for heterogeneous
combustion and fuel bound-nitrogen conversion discussed in this section
(involving: (1) oxidation of carbon, hydrogen, and nitrogen to carbon
monoxide, water, and NO under fuel-rich to stoichiometric conditions;
(Z) water-gas shift reactions and partial reduction of the NO in overall
fuel-rich mixtures; and (3) further oxidation of the nitrogen in the
remaining char particles). Unfortunately, the considerable effort
required to develop and incorporate this mechanism was not within the
scope of this study. As a result, the model of bound nitrogen conversion
first proposed and used in earlier studies in this program [2] has been
little improved. It was used, essentially unchanged, throughout this
study.
45
-------
SECTION V
RESULTS OF DATA ANALYSES
In the previous studies [1,2] , regression analyses were
used to quantify the coefficients of the terms in a semiempirically
derived equation for the total NOX emissions (concentrations) from
full-scale utility boilers. Because of the complexity and the large
number of terms (nine) in the developed equation, it was difficult to
interpret the resulting coefficients of each term with regard to the accu-
racy of the assumptions and the calculation technique involved in each
term. The greatest benefit was derived from the parametric calcu-
lations concerning the effects of the broader combustion modifications
on the overall boiler NOX emissions; the entire nine-term expression
as quantified by the regression analyses was used.
In this study, as a result of the simplifications and im-
provements, the NOx calculation expression was reduced to three
terms, plus the constant necessary for the regression analyses. Each
of these terms was intended to represent a major source of NOX in the
furnace: (1) thermal NOx generated in the active burner region (the
first stage, in staged combustion); (2) thermal NOx formed in the final
mixing zone (the second stage, in staged combustion); and (3) fuel-
bound nitrogen conversion (assumed to occur entirely in the active
burner region). An attempt was made to directly calculate the NOx
contribution from each of these three sources, so that direct compar-
isons with available measured data could be made, where possible.
Significant barriers to direct correlation and comparison
of all available data, in one large sample, were expected because of
(1) unknown changes in effective gas emissivities between the three
fuels (natural gas, oil and coal) and (2) uncertainties in the gasification
and mixing rates appropriate to the three fuel states (gas, liquid and
solid). As a result, correlations and analyses were made separately
and will be discussed separately in this section for the three fuel
types.
46
-------
5. 1 RESULTS OF REGRESSION ANALYSES
Table 3 shows the results of regression analyses of
several samples of the available, appropriate NOX emission data. In
all cases, the data samples were small enough (Table 2) that direct
comparison of the calculated NOX levels with the measured NOX were
not only possible but more instructive (Section 5. 2). The results of the
regression analyses, therefore, were used more as an additional indi-
cation of the validity and accuracy of the NOX calculations. Unlike the
previous studies, parametric calculations with the quantified correlation
equation were not conducted.
Significant observations from the results of the regression
analyses fall into two categories: (1) the significance of the correlation
coefficients and (2) the significance of the empirically-derived coefficients
of the terms in the correlation equation.
5. 1. 1 Correlation Coefficients
Table 3 shows that the correlation coefficients for a given
fuel and boiler type were adequate to good, ranging from 0. 826 to 0. 915,
but that attempts to correlate data from several types of boilers, even
involving the same fuel type, resulted in rather poor correlation co-
efficients (0. 681 to 0. 786). This same result was observed in the pre-
vious studies, as shown in the last column of Table 3. For the natural
gas- and oil-fired boilers, the difficulty in correlating data from all
boiler types together in a single sample appears to be related more to
possible inaccuracies in some of the data obtained from the tests of the
single-wall boilers. This problem will be discussed further in Section
5.2.
Difficulties were also encountered in correlating single-
wall and tangential coal-fired boilers together in a single sample. As
shown in Table 3, separate correlations of the two coal-firing boiler
types are good, but when the data samples are combined the resulting
correlation coefficient is rather poor. There is no apparent trend in
the empirical values of the coefficients of the terms for the single-wall
and tangential configurations when correlated separately or as a total
coal-fired data sample.
5. 1. 2 Coefficients of the Terms
The best possible results of the regression analyses would
be for the empirical coefficients of all terms in the correlation equations
to be 1. 0 and the constant to be zero. This result would imply that the
calculated levels of NOX from the three sources (terms) were exactly
correct, both in slope (the sign of the coefficient) and in magnitude.
Such a result, of course, would require that the measured data be
exactly accurate as well. Large data scatter tends to reduce the mag-
nitude of the constant (leading eventually to a case where any coherent
variations are lost in the data scatter).
47
-------
TABLE 3. RESULTS OF REGRESSION ANALYSES
OF NOX EMISSIONS DATA
Sample
Boilers
Natural Gas
All boilers
Opposed-fired only
Oil
All boilers
Opposed-fired only
Coal
All boilers
Single -wall
Tangential
Empirical Coefficients for the
Terms Representing the NOX Sources
Correlation
Coefficient
0.786
0.826
0.762
0.875
0. 681
0. 894
0.915
Thermal NO.,
Ji.
Active
Burner
Region
0.289
0. 564
0.097
-0.419
0. 006
-0.054
0.001
Final
Mixing
Zone
8.43
4. 10
7.53
12.6
1.64
5.74
2. 17
Fuel-
Bound
Nitrogen
Conversion
NA
NA
0. 174
0. 160
0. 335
0. 802
0. 843
. Constant
162
114
213
302
268
254
2
Correlation
Coefficients,
Previous
Study
0. 837
0. 844
0. 734
0. 849
--
--
—
oo
-------
The coefficients for the two sources of thermal NOX for
natural gas-fired boilers are both positive, indicating that the calculated
trends are correct. The magnitude of the coefficients for the term
describing the active burner region and the magnitude of the constants
are reasonable. This is not too unexpected because the combustion gas
cooling rate was determined from data obtained with natural gas fired
in the largest of the four pairs of sister boilers in the natural gas data
sample. In addition, there are no problems associated with fuel gasifi-
cation rates or with fuel-bound nitrogen or sulfur. The coefficients for
the term describing NOX formed in the final mixing zone, however,
are much larger than one. This implies that either the gas cooling rate
used in the calculation is too large or the time spent by the gas enroute
to this final mixing zone is too large.
Study of the details of the NOX calculations in several of
the subzones within the active burner region shows that, with natural
gas fuels, the total NOX formed in the boiler is dominated by that formed
in the very early combustion zones, just off the burner exits and in the
lower burner levels, before appreciable mixing with bulk gases. These
are the regions of maximum combustion product temperatures. Because
of the exponential effect of temperature on thermal NOX formation, the
overall NOx calculation is very sensitive to the combustion air tempera-
ture and to the burner air-fuel ratio. Unfortunately, as discussed in
more detail in Section 5. 2, the actual combustion air temperatures at
the burner inlets were not directly measured but could only be inferred,
for this data sample, from rather limited measurements at a distant
location (at the exit of the preheater). This undoubtedly is a major cause
of the rather large scatter in the calculated NOX levels relative to the
measured NOX levels.
Similarly, NOX formed in the final mixing zone is again
exponentially sensitive to the combustion product temperature entering
that zone. In this case, accuracy of the initial gas temperature is
dependent on accurate measurement of the combustion air temperature
and on accurate calculation of the burner air-fuel ratios, the gas cooling
rates throughout the active burner region, and the quantity and tempera-
ture of any flue gas recirculation (from the furnace bottom). Calculated
contributions of NOx from the final mixing zone range over six orders
of magnitude for the various operating conditions in the data sample but
never exceed about 40 ppm. Because the NOX calculation is exponen-
tially related to the initial gas temperatures in that zone, small errors
in any of the input data or the subsequent calculation can cause large
variations in the calculated contribution of NOX from that region.
As indicated by the appropriate coefficients in Table 3,
for all fuels and boiler data samples, the regression analyses repeatedly
indicate that the calculated NOX formed in the final mixing zone, for one
or more of the possible reasons discussed above, is low by factors
ranging from 1. 6 to 12. 6. While this seems like a large error in the
49
-------
calculation, it is not large compared to the six orders of magnitude
of the range of the calculated contributions from this zone. The error
merely tends to highlight the problem of trying to accurately calculate
NO formation at relatively high temperatures in the light of the fact
that that rate is exponentially related to temperature. No effort was
made to further refine the cooling rate or any other part of the flow and
mixing calculations to reduce this error.
Another interesting observation from the empirically
determined term coefficients shown in Table 3 is that the coefficients
for the term describing NOX formed in the active burner region decrease
to essentially negligible levels while those related to the NOX formed
from fuel-bound nitrogen increase to acceptably high levels in the fuel
progression from natural gas, through oil, to coal. At the very least,
this variation, within the scatter of the data, reflects the increasing sig-
nificance in that progression of the fuel-bound nitrogen. These results
indicate that, with significant concentrations of elemental nitrogen chem-
ically bound in the fuel, variations in the measured data within the
data scatter can best be explained by looking primarily at the conversion
of this fuel nitrogen and neglecting variations in the thermal NOX
formed in the active burner region. This does not necessarily imply
that thermal NOX formed in the active burner region is negligible with
oil and coal fuels; it does imply that the amount of NOX formed in this
region is relatively independent of variations in operating conditions
(constant). In such a case, the regression analysis of the data could
eliminate the thermal NOX calculated to be formed in the active burner
region (by developing a very small coefficient for that term) and instead
account for that relatively constant NOX in the derived constant. This
appears to be the case with oil fuels and with coal fuels fired in the
single-wall boilers.
The term coefficients and the constant shown in Table 3
for the tangential boilers, however, indicate that the thermal NOX
formed in the active burner region should not only be considered constant
but also negligible. The derived coefficient for the term representing
that region is essentially zero (0. 001) and so is the derived constant
(2 ppm). In addition, the coefficients for the NOX from the final mixing
zone (2. 17) and from the conversion of fuel-bound nitrogen (0. 843) are
close enough to the expected 1. 0 that the resulting corrections, in terms
of ppm of NOX, are quite small (+13 ppm and -51 ppm, respectively).
Comparing the term coefficients derived for the single-
wall and the tangential boilers, then, seems to indicate that a significant,
approximately constant amount of NOX is generated in the active burner
region and in the burner flow and mixing region upstream of the bulk
gases in single-wall boilers; however, little NOX is generated in this
same region in tangential boilers. This result is qualitatively reason-
able because (1) the active burner flow streams travel into the furnace
much closer to the water walls and can take better advantage of direct
50
-------
radiative and convective heat transfer to quicly cool the products of
reaction and (2) the design of the burner arrays in a tangential boiler
inherently provides effective combustion staging and even with all
burners active the air-fuel ratio in the majority of the active burner
region in the furnace is significantly lower than that of the overall
boiler. These same characteristics may also account for the large
scatter in the measured CO2/O2 data discussed in Section 4. 3. !„
The agreement between the total calculated and measured
NOX data for the coal-fired boilers, particularly the tangential boilers,
is considered excellent if the NOX calculated for the initial combustion
region is taken as a constant (zero in tangential boilers). A level of
NOX formed in this region that is relatively independent of the operating
conditions is reasonable because of the slow rate of gasification of the
solid fuel. Over most of the total coal gasification and mixing time
(and distance into the furnace) the local air-fuel ratios are much more
dependent on the rate of gasification rather than on the overall average
burner air-fuel ratio. Only when gasification is nearly complete, and
the mixing with bulk gases may already have begun, does the overall
burner air-fuel ratio have an effect on NOX formation. Thus, the neg-
ligible variation of the NOX formed in the active burner region with
variations in burner operating conditions appears reasonable with coal
fuels and, to some extent, with oil fuels. The calculated magnitude of
that constant level of NOX formation with coal fuels, however, is still
much too high. Despite the extensive efforts, described in Section
4. 3. 4, to allow for very slow gasification and mixing rates, particularly
with the coal fuels, the calculation of the NOX formed in the early
regions of combustion of coal fuels results in NOX contributions which
are high. In single-wall boilers, this contribution is high by as much
as an order of magnitude. If this were a problem of inaccurate com-
bustion air temperature data (temperatures too high), as suspected
from other observations, one might expect the coefficients of the term
for the final mixing zone to be similarly low. In this case, however,
as discussed in Section 4. 3. 3, it is likely that the actual gas cooling
rate is much higher than that used in the calculation (derived from
natural gas-fired heat flux data), thereby compensating for erroneously
high initial temperatures. On the other hand, the very high calculated
levels for the NOX formed in the active burner region could also be an
indication that high levels of NOX are actually formed in the early com-
bustion of liquid and solid fuels, as calculated, but are subsequently
reduced by some NOX destruction mechanism not incorporated in the
calculation. Such a mechanism might be related to the mechanism of
fuel-bound nitrogen conversion, as discussed in Section 4. 3. 5, or it
could be an indication of the effect of fuel-bound sulfur (highest in coal)
on NOX formation, as postulated by Wendt [10 ] and others. There was
not sufficient time in this study to further investigate some of these
latter possibilities.
51
-------
Results of the regression analyses, then, shed some
light on calculation problems and general trends of NOx emissions
with the various fuels. More detailed results can be seen by direct
comparison of the calculated and measured NOX levels.
5.2 DIRECT DATA ANALYSES
While the observations from the correlations and the
empirically derived coefficients of the terms (and the constant) in the
NOX equation are instructive relative to the accuracy of the calculations
of NOx formed in the various regions of the furnace, the fact that the
contributions from each region to the final NOX levels are calculated
directly makes direct comparisons of calculated and measured NOx
levels perhaps even more instructive. Such direct comparisons were
not feasible in the earlier studies because (1) it was not considered
possible to express the contributions from each region in terms of NOX
in the flue gases, and (2) the larger number of terms (degrees of free-
dom) used in the earlier correlating equation made interpretation of
any differences difficult.
The correlation equation used in this study consists of
only three variable terms, representing the contribution of NOX
(expressed as ppm in the flue gases) thermally generated in the active
burner region and in the final mixing zone and from the conversion of
fuel-bound nitrogen, plus the constant necessity for regression analyses.
With natural gas fuels, the fuel-bound nitrogen term is always equal to
zero. The three variable terms represent real sources of NOX, but
the constant does not.
There is always a question concerning what NOX source
is represented in the constant or if the constant is merely an expression
of the data scatter. Following the discussion in Section 4. 3. 4, concern-
ing the finite rate gasification and mixing processes in the active burner
region, it is clear that the NOX generated in the active burner region
logically could consist of two parts: (1) one in which the combustion
conditions are largely controlled by the local gasification and mixing
rates, relatively independently of the overall burner operating conditions;
and (2) one in which the initial gasification and mixing processes are
essentially complete, and the combustion conditions are directly dependent
on the overall burner operating conditions. Since there is no gasification
and only a short mixing region with gaseous fuels, the NOX contribution
from the active burner region should be largely variable and dependent
on burner operating conditions. With solid fuels, however, the gasifi-
cation rate may be small compared to the cooling rate, and by the time
gasification is complete the product gas temperatures may be sufficiently
low that further NOX formation is small. In the latter case, the NOX
generated in the active burner region should be relatively independent of
the burner operating conditions. The case with liquid fuels should lie
somewhere between those of gaseous and solid fuels.
52
-------
All of these effects are included in the computer cal-
culation but are necessary to keep in mind when comparing the calcu-
lated NOX data with the measured and in interpreting any significant
differences.
5. 2. 1 Natural Gas Fuel
As mentioned, the NOX contributions from the active
burner region with gaseous fuels should be variable. Thus, despite
the term coefficients and the constant shown by the regression analyses,
the NOX calculated directly from the active burner region and the final
mixing zone, using coefficients of 1.0, should compare directly with
the measured data (i. e. , the constants derived by the regression analyses
have no physical meaning).
Figure 2(a) shows a comparison of the measured and cal-
culated NOX data from 92 tests in two sizes (four boilers) of opposed-
fired boilers with natural gas fuels (boiler nos. 3 and 4 in Table 2).
Perfect agreement would be represented by all data lying exactly on
the 45-degree line shown in the figure. The data show considerable
scatter, but the agreement is generally good. On the average, the cal-
culated NOX levels are higher than the measured, but by only 6. 6
ppm,, The average difference (deviation) between the measured and cal-
culated data, however, is just under 100 ppm.
A large part of the data scatter is due to the seven tests
shown in Figure 2(a) where the calculated NOX level is above about
810 ppm. These seven points have an average error of 227 ppm, with
an average deviation of 267 ppm. Without them, the rest of the data
shows an average error of -11. 5 ppm and an average deviation of
86 ppm.
Some of the data scatter could be just that, random errors
in either or both the measured NOX levels or in the input data (partic-
ularly the combustion air temperature and the CO2/O2 data) from which
the calculated NOx levels were derived. The scatter could also be due
to the inability of the calculation to take into account some subtle changes
in operating conditions which have strong effects on the actual NOX
levels. Table 4 shows the data on operating conditions input to the NOX
calculation for the five tests for which NOX levels between 917 and 984
ppm were calculated. All of these data were from the same boiler, with
the same natural gas fuel and the same burner and NOX port configu-
rations, and none involve flue gas recirculation. None of the input oper-
ating conditions are significantly different among these five tests except
the measured C>2 levels, and even these variations show no regular
trend which could explain the observed variations in the measured NOX
levels. As a result, the NOX levels calculated from these input data
must be, and are, very nearly the same. These are all high load
(even above rated) tests, and the combustion air temperatures are all
53
-------
1000 ,-
MIXED SINGLE-,
TWO-STAGE
CONFIGURATIONS
o
l/>
l/l
600
400
200
ALL DATA WITH ALL
BURNERS ACTIVE,
N0x PORTS CLOSED
O
I i I i I
200
400
600
800
1000
E
a
1000
800
600
400
200
CALCULATED N0x EMISSIONS, ppm, dry
(a) NATURAL GAS-. OPPOSED-FIRED BOILERS
BURNER OPERATION
© ALL BURNERS ACTIVE
® STAGED COMBUSTION
200
400
600
1000
CALCULATED NO EMISSIONS, ppm, dry
Ibl NATURAL GAS-, SINGLE-WALL-FIRED BOILERS
Figure 2. Comparison of calculated and measured NO
emissions.
x
54
-------
TABLE 4. EXAMPLE OF THE POSSIBLE WIDE VARIATION IN MEASURED
NOX UNDER ESSENTIALLY THE SAME OPERATING CONDITIONS
in
Identical Conditions:
350 MW, opposed-fired boiler
Natural gas fuel
All burners active, NOX ports closed
No flue gas recirculation
Operating
Measured
NOX,
ppm
919
840
632
543
298
Average
Average
Deviation,
Conditions:
°2» %
2. 30
1. 35
1. 11
2. 01
1. 58
1. 670
23.2
co2, %
10.45
10.80
10.90
10. 52
10.70
10.674
1.4
Load, MW Fuel
Flow,
Ib/sec
350 37. 8
354 38.7
352 36.3
352 36.7
351 36.8
351.8 37.260
0. 3 2. 1
Combustion
Air
Temperature,
F
585
580
577
580
585
581.4
0. 5
Unit Conversion:
kg/sec = 0.435 (Ib/sec)
K = 5
460)/9
-------
very high. In addition, the C>2 levels are all relatively low, yielding
very high values for the combustion temperature rise. The combination
of these maximum temperatures yields maximum combustion product
temperatures. At these very high temperatures, the NOX formation
rates (being exponential in temperature) are not only very high but are
also extremely sensitive to small errors in those measurements which
lead to the combustion product temperatures (i. e. , combustion air
temperature, COZ/Oz levels, and the equilibrium combustion tempera-
ture calculation). Small errors in these measured values could be
responsible for the observed large errors in the calculated NOX levels
under these test conditions. Similarily, control of actual NOX levels
under these operating conditions is also difficult because of the large
sensitivity of NOX to small (but real) variations in these operating
conditions.
Figure 2(b) shows a similar comparison of measured
and calculated NOX data, from 15 additional tests with natural gas, in
this case, from two sizes (three boilers) of single-wall-fired boilers.
Here the agreement between calculated and measured data is very poor
for all but the two tests involving staged combustion. Of the data from
tests with all burners active, on the average the calculated NOX levels
are higher than the measured levels by about 430 ppm. As in the
previous study [l] , concerning these particular boilers, the measured
combustion air temperatures could be in error (too high). This is
partially substantiated by the fact that the two cases of staged combustion
(involving low combustion temperature rise resulting from the rich
air-fuel mixture) show good agreement between calculated and measured
NOX levels. Also, it will subsequently be seen that this error is almost
identical in the data from these same boilers when fired with oil fuels.
Attempts in this and the previous study to correlate data from these
boilers with that from the opposed-fired boilers have never been
successful. Once again, no explanation other than the possibility of
input data error could be found.
5. 2. 2 Oil Fuel
As mentioned in Section 4. 3. 4, NOX generated in the
active burner region with oil fuels could be partly generated in the sub-
region dominated by the slow gasification of the liquid fuel or in the
subregion more controlled by the burner operating conditions. Figure
3(a) shows a comparison of the measured NOX levels with those calcu-
lated from the sum of the thermal NOX contributions, from the active
burner region and the final mixing zone, and the conversion of the fuel-
bound nitrogen (with all term coefficients equal to one and with no
additional constant) from the 30 (appropriate) tests in the opposed-,
oil-fired boilers. Much of the data is in good agreement but there is a
tendency to calculate excessively high NOX levels at the lower end of
the range. On the average, the calculation is about 50 ppm too high.
The average deviation of the data about that corrected calculation
(+50 ppm) is 44 ppm.
56
-------
600 i—
500
I
400
= 300
200
100
100
1000
800
t/1
z
o
600
400
200
08 0° O
200
300
400
500
600
CALCULATED NO EMISSIONS, ppm, dry
(a) OIL-, OPPOSED-FIRED BOILERS
BURNER OPERATION
O ALL BURNERS ACTIVE
« STAGED COMBUSTION
I I I
200
400
600
800
1000
CALCULATED N0x EMISSIONS, ppm, dry
(b) OIL-, SINGLE-WALL-FIRED BOILERS
Figure 3. Comparison of calculated and measured NO
emissions.
x
57
-------
The apparent error in the calculation is not considered
significant at this point because the NOX contribution from the active
burner region is rather strongly affected by the relatively unknown
liquid vaporization rate and the assumption of the fraction of the vapor-
ized fuel burning under stoichiometric conditions during the vaporization
period. As discussed later in this report (Section 5. 3), the thermal
NOX calculated to be formed in the active burner region could be
reduced 50 ppm by, for example, increasing the distance for complete
vaporization by about 5 percent. No effort was made to further tune
the calculation.
Figure 3(b) again shows data from the two sizes (four
boilers) of the single-wall configuration, analogous to Figure 2(b) but
in this case firing oil fuel. Again, some of the calculated data agrees
rather well with the measured, but with another group of data the cal-
culation is high by about the same amount as with the natural gas fuel
(430 ppm). Again, these data showing the large error involve the higher
combustion air temperatures and lower air-fuel ratios, leading to maxi-
mum combustion product temperatures. The two tests involving staged
combustion also both involve high combustion air temperatures, but
the calculation shows a low NOX level because of the fuel-rich com-
bustion.
The cause of the large calculation error in some oil-
fired, single-wall boiler data, then, as in the case of natural gas fired
in these same boilers, could be errors in the measured combustion air
temperatures. These temperatures were actually measured at the out-
let of the air preheater, not in the burners, and were measured in only
a small fraction of the tests. If the actual combustion air temperatures
in the burners during these tests were lower by 5 to 30 K (10 to 60 F),
the calculated NOX would also show good agreement with the measured
NOX.
5. 2. 3 Coal Fuels
Comparison of measured NO data from the coal-fired
boilers with the calculated values, using term coefficients of 1. 0 and
no constant, immediately shows the problem with the calculated con-
tribution of NOX from the active burner region. While the total meas-
ured NOX in the flue gases range from 200 to 600 ppm and the contri-
butions from the final mixing zone and from the conversion of fuel-
bound nitrogen appear quite reasonable and normal, the calculated NOX
contributions from the active burner region range vary widely, from as
low as about 200 to more than 8000 ppm.
It seems quite possible, as with the oil fuel, that varying
the constants describing the distance into the furnace for complete gas-
ification and the fraction of the gasified fuel in a stoichiometric flame
could reduce these calculated contributions to acceptable levels. How-
ever, the regression analysis (Table 3) consistently indicated that the
58
-------
contributions from the variable portion of the active burner region
should be neglected (a derived coefficient for the active burner region
term essentially equal to zero) for all coal-fired boilers and, in the
case of the tangential boilers, that the contributions from the constant
portion of the active burner region should also be zero. The constant
portion in coal-fired, single-wall boilers appeared to be significant.
In the interest of time (in the study), therefore, no effort was made to
empirically adjust the finite rate gasification and mixing constants for
the coal fuels. Instead, the coefficient of the active burner region term
was simply taken to be equal to zero for all coal-fired boilers of all
configurations. The constant portion was also taken to be equal to zero
for the tangential boiler. The constant portion appropriate to the single-
wall-fired boilers was established directly from the resulting calculated
and measured data comparison.
This conclusion implies that: (1) in single-wall, coal-
fired boilers, essentially all of the thermal NOX generated in the active
burner region is generated, while the coal is gasifying, in a region where
the combustion and thermal NOX formation conditions are largely con-
trolled by the transient gasification and mixing processes rather than
by the active burner operating conditions (a not too surprising conclusion);
and (2) in tangential boilers, the thermal NOX generated in any portion
of the active burner region is negligible compared to NOX generated by
conversion of fuel-bound nitrogen (a somewhat surprising result).
The conclusion relative to the single-wall boilers is
somewhat different from that reached in the previous study [ 2 ]. Al-
though both studies identified a relatively constant level of thermal NOX»
independent of burner operating conditions, and this constant level was
almost exactly the same (125 ppm in the previous study and 127 ppm
here), the previous study indicated that this NOX was thought to be gen-
erated in the final mixing zone.
Figure 4(a) shows a comparison of the measured and
calculated NOX data for the two coal-fired single-wall boilers. A
constant level of 127 ppm as the total NOX contribution from the active
burner region best fits the data comparison (the constant of 260 ppm
determined by the regression analysis for these boilers undoubtedly
partially accounts for random data scatter). The figure again shows
reasonable agreement between measured NOX levels and those calcu-
lated using a coefficient of zero for the active burner region term,
1. 0 for the final mixing zone and fuel-bound nitrogen terms, and a con-
stant level (ascribed to the active burner region) of 127 ppm. There
appears to be a tendency for the calculation to be low at the low and
high ends of the NOX range. The average deviation for all of the data
is 45. 6 ppm. Neglecting the three tests at the very low calculated
level, the average deviation would be 43. 8 ppm.
59
-------
CALCULATION APPROXIMATES THERMAL
NOX FORMED IN THE ACTIVE BURNER
REGION WITH A CONSTANT • 127 ppm
E
a
in
z
o
l/l
600
500
400
o
z
Q
300
200
— ALL BURNERS ACTIVE
200
300
400
500
600
£>
•o
E"
600
500
:= 400
300
200
CALCULATED N0x EMISSIONS, ppm, dry
(a) COAL-FIRED, SINGLE-WALL BOILERS
CALCULATION NEGLECTS THERMAL NOX
FORMED IN THE ACTIVE BURNER REGION
STAGED
COMBUSTION
(20% of
burners
air-only)
ALL BURNERS
ACTIVE
200
300
400
500
600
CALCULATED N0x EMISSIONS, ppm, dry
(b) COAL-FIRED TANGENTIAL BOILERS
Figure 4. Comparison of calculated and measured NOX
emissions.
60
-------
Figure 4(b) shows a comparison of the measured and
calculated NO data for the two tangential boilers firing two types of
coal (nominal and high nitrogen, 1. 3 and 1. 7 percent nitrogen, and 3. 1
and 1.7 percent sulfur, respectively). This calculation neglects all
NOX formed in the active burner region and considers only that NOX
from the conversion of fuel-bound nitrogen, and the thermal NOX formed
in the final mixing zone. However, that formed by conversion of the
fuel-bound nitrogen represents about 95 percent of the calculated NOX.
Agreement between the measured and calculated data
from the 330 MW boiler is excellent. On the average, the calculated
values are lower than the measured by only 9 ppm, with an average
deviation of 30 ppm. The calculated data for the 360 MW boiler, how-
ever, is higher than the measured data by 44 ppm. For all tests, in
both boilers (62 tests), the calculated levels are higher than the meas-
ured by 26 ppm, and the average deviation is 43 ppm.
Since the calculated NOX levels are largely due to con-
version of fuel-bound nitrogen and this source varies only with the
burner air-fuel ratio for a given coal, the two groups of data shown in
Figure 4(b) represent: (1) those tests with all burners active (the higher
group); and (2) those operating with staged combustion (20 percent, the
top row, of the burners operated air-only). ,
5.3 EFFECTS OF GASIFICATION AND MIXING RATES
As discussed in Section 4. 3. 4, the finite rate processes
of gasification of a. liquid (vaporization) or solid fuel have been approxi-
mated, in this study, by three parallel regions in the flow from an
active burner. These regions (not necessarily physically contiguous)
are assumed to contain the following: Region 1, all unvaporized fuel
(and no air); Region 2, the fraction FSS* of the total fuel vaporized up
to that point, plus a stoichiometric equivalent proportion of air; and
Region 3, all of the rest of the fuel and air coming from the active burner
not in the first two regions. The gasification and mixing rate is esti-
mated by selecting the distance DCG * from the burner exit to some
point in the furnace, over which the gasification and mixing is essentially
completed.
The fuel and air specie concentrations in each of these
three regions are related only to the fuel and air entering the furnace
from a particular burner. Therefore, if these gasification and mixing
processes are not complete before significant mixing with bulk gases
(products of complete combustion from burner levels below the partic-
ular burner) begins, then Region 3 cannot exist beyond this point, and
>
'FORTRAN notation used in the calculation of this study
to describe these terms.
61
-------
it is likely that the stoichiometric flame (Region 2) will be greatly
reduced in magnitude. As a result, the three-region gasification and
mixing process approximation is carried out only until this bulk gas
mixing begins.
With natural gas fuels, the burner flows are already
gasified, and mixing is always assumed to be complete, at any burner
level, within the burner core flow, before any significant mixing with
bulk gases begins. With coal fuels, however, the gasification process
takes so long that only at the lowest burner level, where no bulk gases
have been generated in a lower burner level, is gasification and mixing
complete before mixing with bulk gases begins. Therefore, with coal
fuels, attempts to control NOX by modifying coal particle sizes or
gaseous mixing rates (i. e. , burner design) will be most effective in
burners located low in the burner array. Attempts to carefully control
the gasification and mixing rates in the flow field from an active burner
located high in the burner array will tend to be disrupted by the gross
flow of bulk gases past the burner. On the other hand, burner design
changes can be effective at all burner levels with gaseous fuels because
the mixing is complete only a short distance into the furnace.
Of the two constants controlling the gasification and
mixing approximation in this calculation, the distance DCG can be
reasonably estimated from observations of full-scale flames in full-
scale boilers. Clearly, maximum liquid droplet or coal particle sizes
largely control DCG. The amount of combustion involved in a stoichio-
metric flame (FSS), however, is not easily estimated. It depends on
the dispersion of the particles in the burner flow and the large- and
small-scale turbulence in this flow, which, in turn, control the local
gaseous mixing rate relative to the gasification rate. If the particles
are widely dispersed and the local gaseous mixing rates are high com-
pared to the gasification rates, the stoichiometric flame will be thin
and close to the particles and the appropriate FSS will be small. If
the particles are closely grouped (in particular clouds) and the local
turbulence level and the gaseous mixing rates are low compared to the
gasification rates, the region of stoichiometric combustion may be
broad and may exist within and outside of the particle cloud. In such
cases, the appropriate FSS will be large.
The effects of varying DCG and FSS values on thermal
NOX generated in this region of gasification and mixing are not obvious.
By studying the effects of their variation on the various intermediate
steps in the NOX calculation, it appears that the primary effect of
distributed gasification and mixing is to provide more time at tempera-
tures where heat transfer (cooling) is appreciable but the NOX formation
is still low. While the temperatures and NOX formation rates in
Region 2, the stoichiometric flame, are always high, the existence of
this and Region 1 limits the amount of fuel species in Region 3. The
effective air-fuel ratio in Region 3 starts at infinity (no fuel) and slowly
62
-------
decreases as gasification and mixing proceed. Thus, although oxidizer-
rich, the temperature in this region rises relatively slowly. Over a
considerable fraction of the particle lifetime, the temperatures in
Region 3 are high enough to affect considerable cooling without gener-
ating appreciable NOX (cooling is proportional to the fourth power,
while NOX formation is an exponential function, of the combustion
product temperature)., With fuel-rich burner operation (staged com-
bustion), the effective air-fuel ratio in Region 3 must eventually pass
through stoichiometric, but by this time there usually has been sufficient
cooling to significantly reduce the peak NOX formation rate.
Some parametric calculations were made to investigate
the overall effect of variations in DCG and FSS. The opposed-, oil-
fired boilers were used for the calculation. Table 5 shows these results.
Thermal NOX contributions from the active burner region are tabulated,
with DCG and FSS as parameters. Calculations for two typical full-load
tests, with all burners active, are shown in the table.
Clearly, increasing the maximum oil droplet sizes in-
creases DCG, and the table shows that this results in minimum NOX.
Decreasing the gaseous mixing rate (relative to the gasification rate)
should increase FSS. The table shows that increasing FSS also mini-
mizes NOX. Both of these effects, however, are reduced under con-
ditions of staged combustion. The fact that the NOX shown in Table 5
is sometimes higher with NOX ports open than closed, for the same
DCG and FSS, is correct but would be offset by a lower conversion of
fuel-bound nitrogen with NOX ports open. At least within the range of
the variations in DCG and FSS tested, minimum NOX with oil fuels is
achieved with poor atomization (large droplets poorly distributed) in a
burner air flow of low turbulence.
Although not specifically run parametrically, the case
with coal fuels appears similar. Again, minimum NOx should result,
at least in single-wall-fired boilers, from operating conditions which
reduce the coal particle gasification rate and decrease the local gaseous
mixing rate. Study of NOX data from tangential boilers, however,
indicates that thermal NOX formed in the active burner region is usually
small. Variations in the gasification and mixing rates, therefore,
should have little effect on NOX in these boiler types.
Apparently, the general principles for NOx reduction
in the active burner region is simply to lengthen the total time to
complete gasification and mixing.
5. 4 GUIDELINES
The primary objective of this and the preceding studies
of NOx control by combustion modification was to develop guidelines
which can be used to guide design and design modifications to minimize
63
-------
TABLE 5. EFFECTS OF OIL VAPORIZATION AND MIXING
PARAMETERS ON THE THERMAL NOx GENERATED
IN THE ACTIVE BURNER REGION
Example Cases
350 MW opposed-fired boiler
Horizontal depth of furnace = 30 feet
Full-load operation, all burners active
NO Ports
X
02, %
C02, %
DCG, ft
5.5
6.5
7.5
7.5
7.5
FSS, %
0.25
0.25
C. 25
0.20
0. 15
Case 1
Closed
2.98
13.28
NOX,
Case 1
450
219
95
159
258
Case
Open
3.05
13.40
ppm (dry)
Case
328
253
160
209
243
2
2
a. DCG = the distance into the furnace for complete vaporization
and mixing
b. FSS = the fraction of the vaporized fuel involved in stoichio-
metric combustion over the majority of DCG
64
-------
NOX emissions from utility boilers, within the bounds of high plant
efficiency, stable combustion, and acceptable levels of emission of
other air pollutants. This and the previous studies made use of a
complex computer program and engineering model of NOX formation
in, and large samples of data from, full-scale, multiburner utilility
boilers to develop understanding of the major sources of NOX in such
boilers and to investigate the effects of variations in some of the major
hardware and operating conditions on this NOX. It is not the intent of
this or the previous studies simply to provide the final computer program.
Rather, the intent is and has been to glean from all of these studies
useful information or guidelines on appropriate modifications in hard-
ware and operating conditions which can be directly applied to full-
scale, multiburner boilers.
This and the previous studies concluded that staged com-
bustion can effect significant NOX reductions while maintaining high
plant efficiency and acceptable levels of emission of other air pollutants.
Staged combustion is defined as fuel-rich operation of the active burners
with the remaining combustion air introduced downstream of the active
burner region. Analytical techniques were developed to assure stable
combustion even with very fuel-rich operation of the active burners.
Although staged combustion can affect major reductions in NOX emissions,
in most cases other techniques are necessary to further reduce the
remaining thermally generated NOX emissions.
Although there is little substantiating data in the data
sample available to these studies, it appears that reduction of the com-
bustion air temperatures can also be very effective in reducing NOX
emissions. This effect is clearly shown by the analytical calculation.
Unless this temperature reduction is achieved by transferring more
heat to the steam cycle (reduced flue gas temperature into the air pre-
heater), however, significant plant efficiency losses can result (higher
sensible heat losses up the stack).
The data sample available to these studies also did not
include data on the effects of flue gas recirculation through the active
burners. Analytically, however, these studies indicate that such flue
gas recirculation could substantially reduce NOX with fuels and in
boilers where substantial fractions of the total NOX result from thermal
formation in the active burner region (i. e. , natural gas fuels, most
low nitrogen liquid fuels, and to some extent coal-fired, single-wall
boilers). Such flue gas recirculation also should not affect plant
efficiency or emissions of other air pollutants but could affect com-
bustion stability. It should have little effect on NOX from coal-fired
or, perhaps, on some oil-fired tangential boilers.
La the previous studies [ 1, 2], parametric calculations
were conducted and reported to show the effects of wide variations of
some of the major boiler hardware and operating conditions on the total
65
-------
NOX, using term coefficients and constants derived empirically from
regression analyses of all of the data. The data used in this study are
a subset of those used in the previous study in that all of the so-called
off-stoichiometric tests were eliminated. However, since both this and
the previous NOX calculations fit the same data, no specific parametric
calculations were conducted in this study. Rather, this study attempts
to explain more accurately why certain NOX variations occur when one
or more of the hardware or operating conditions are varied. This also
gives more credence to interpolations and extrapolations within and
beyond the available data because of the lesser degree of empiricism
involved in this study.
The two major differences from the previous study (in
the guidelines for reducing NOX discussed here) result primarily from
(1) the observation from the previous study (accepted at the outset of
this study) that NOX levels with staged combustion are equal to or less
than those achievable with off-stoichiometric configurations and (2) the
observation from this study that a relatively constant level of NOX
(relatively independent of variations in the degree of combustion staging)
is apparently thermally generated in that part of the active burner
region where the gasification and mixing rates rather than the burner
operating conditions control the local air-fuel ratio.
The latter conclusion is different from that reached in the
previous studies. This conclusion became apparent with the incorpor-
ation of an approximation of the finite rate gasification and mixing pro-
cesses during this study. Previously this constant level was ascribed
to the final mixing zone. This study indicates that the amount of NOX
generated in the final mixing zone in staged combustion is small compared
to that generated in the region controlled by the gasification and mixing
processes.
5. 4. 1 Natural Gas Fuel
Figure 2(a) compares NOX emissions levels measured
in the natural gas-, opposed-fired boiler data sample of this study
against those calculated by the technique of this study. The calculation
involved minimal empiricism (i. e. , although the NOX calculation
equation was that set up for data correlation, the term coefficients were
taken to be equal to 1. 0 and the correlation constant was taken to be
zero). Therefore, extrapolation outside of the range of the available
data has more meaning and involves less risk than the parametric cal-
culation of the previous studies, which used term coefficients and a
constant derived from empirical regression analysis.
With natural gas fuel, the analytical calculation shows
that NOX thermally generated in the active burner region dominates
the total NOX. No NOX is generated from conversion of fuel-bound
nitrogen. In all of the calculations with natural gas fuels, the NOx
thermally generated in the final mixing zone was calculated to be less
than 85 ppm, normally between zero and about 35 ppm. NOx from
66
-------
the final mixing zone was very sensitive to the combustion air tempera-
ture and only reached appreciable levels under full-load conditions.
Examination of the calculation for NOx from the final mixing zone shows
that a relatively small reduction in peak temperatures entering this
zone should effectively eliminate NOX from this zone. This could be
accomplished by (1) a small reduction in the peak combustion air temper-
ature, (2) use of some flue gas recirculation (entering from any point(s)
in the active burner region), or (3) delaying the introduction of the
second-stage air as long as possible. More rapid mixing of the second-
stage air could also reduce the NOX from this zone.
Despite the rapid mixing of the natural gas with the com-
bustion air near the burner exit, the calculation indicates that a small
region of combustion under stoichiometric conditions will still exist in
this mixing region under all overall burner air-fuel ratio operating con-
ditions. Operating the burner very fuel-rich will effectively eliminate
thermal NOX formation in all other subregions of the active burner
region. Using the staged-combustion technique alone, then, the cal-
culation indicates that as the burners are operated increasingly fuel-
rich the total NOX levels should decrease to about 180 to 200 ppm
(150 to 170 ppm from the active burner region and 30 to 40 ppm from
the final mixing zone, with about half of the combustion air diverted to
the second stage; the NOx levels then remain relatively constant for
even richer mixtures.
Figure 5 shows a plot of the NOX data measured in tests
of the four opposed-, natural gas-fired boilers (240 and 350 MW) under
full-load conditions. The NOX levels do indeed decrease with decreas-
ing burner air-fuel equivalence ratios, to a level of about 150 to 180
ppm, with about half of the stoichiometric air passing through the
active burners. The calculation also indicates that NOX levels under
conditions of overall boiler excess air greater than shown by the data
(with all burners active) are also greatly reduced from the peak levels;
however, this is not a solution of interest at this point.
It is unlikely that the burners could be operated with
much richer mixtures than those shown. It is of interest, however,
that these boilers were successfully operated at air-fuel equivalence
ratios as low as those shown. Furthermore, they have been continuously
operated, with natural gas, for several years at ratios of 0.55 to 0.67,
with no detrimental effects on the boilers and with high plant efficiency
and negligible emissions of hydrocarbon, smoke, or carbon monoxide.
Also, the possibility of flame liftoff and/or combustion
instability with very fuel-rich burner operation is significant. Since
most of the remaining NOX under such fuel-rich burner operation is
generated in the very early mixing and combustion region, just down-
stream of the burner exit, where stoichiometric combustion tempera-
tures are maximum, various methods of reducing this peak temperature
67
-------
FULL-LOAD OPERATION
1000
800
e
o.
°i 600
oo
O
00
00
S 400
X
O
200
0
O
00
O
&
oo
-
0
- Qqp
O OO °
- 0 o
CDO
i 1 i 1 . 1 i 1 i 1
0.4 0.6 0.8 1.0 1.2
BURNER AIR-FUEL EQUIVALANCE RATIO
1.4
Figure 5. Effects of combustion staging on NOX emissions:
natural gas-, opposed-fired boilers.
68
-------
are very effective in further reducing this residual NOX. As in con-
trolling NOX from the final mixing zone, small reductions in the
maximum full-load combustion air temperatures, using water sprays,
or flue gas recirculation (in this case directly into the burners or the
windbox) can significantly reduce this remaining NOX. More rapid
mixing of the gas with the combustion air should also be effective.
Introducing water vapor or recirculated flue gases into an already very
fuel-rich burner, however, could increase problems with flame liftoff
and/or combustion instability. The use of water sprays would result
in increased sensible heat losses up the stack, but the resulting efficiency
losses should be small.
5. 4. 2 Oil Fuel
With oil-fired boilers, the calculation indicates that NOX
generated from conversion of fuel-bound nitrogen is an appreciable
fraction of the total emissions levels, ranging from 22 to as high as
about 60 percent in the burner air-fuel ratio range near stoichiometric.
At this time, there is no known way to minimize NOX from this source
other than to operate with fuel-rich mixtures in the active burner region.
Fortunately, this will also reduce much of the thermal NOX formed in
the active burner region.
All of the remarks concerning NOX levels and reduction
techniques in the final mixing zone with natural gas fuels apply to oil
fuels as well. In the case of NOx thermally generated in the active
burner region with oil fuels, the problem is again similar to that with
natural gas. With burners operated very fuel-rich, to minimize con-
version of fuel-bound nitrogen to NOx, an appreciable concentration of
NOX can be thermally formed in the stoichiometric flames surrounding
the vaporizing fuel droplets. With the atomizers, oil fuel, and burners
used in the boilers of this study, the calculation indicates that about
200 ppm of NOX can be generated in this zone (depending on other boiler
operating conditions).
Thus, it is anticipated that as increasing fractions of
the combustion air are diverted to the second stage, the overall full-
load NOX levels should remain at least 200 ppm above NOx levels
generated from conversion of fuel-bound nitrogen until very fuel-rich
mixtures are reached. Figure 6 shows a plot of the NOX data available
to this study from full-load staged combustion tests of the four opposed-
fired boilers using oil fuel. As with the natural gas test data, the oil
data show the large quantities of NOX thermally formed throughout the
active burner region when its air-fuel equivalence ratio is near stoich-
iometric. For richer mixtures, the data appear to show a trend to
decrease to the expected 200 ppm (at full load), but the data does not
extend to sufficiently rich mixtures to verify this lower limit. The
dashed line shown in the figure is estimated from the previous para-
metric calculations (i.e., the calculation does show this limit).
69
-------
FULL-LOAD OPERATION
500
400
o_
CL
CO
o
CO
CO
300
200
100
0
O
/
/ o
o
\0
o\
\
\
0\
o
NOX FROM CONVERSION
OF FUEL-BOUND NITROGEN
0.7 0.8 0.9 1.0 1.1
BURNER AIR-FUEL EQUIVALENCE RATIO
Figure 6. Effects of combustion staging on NOX emissions:
oil, opposed-fired boilers.
70
-------
Table 5 shows that some reduction of this residual NOX
level could be achieved by reducing the degree of oil atomization and
the spray distribution. This will increase the distance DCG, at least,
and very likely increase the fraction FSS as well. Both of these trends
should reduce the residual NOX formed in the active burner region.
The table indicates that increasing DCG by about 15 percent could reduce
the residual NOX by nearly 40 percent. This could reduce the residual
NOX formed in the active burner region under very fuel-rich burner
operation to levels of the order of 120 ppm. Few data, other than the
effects shown by parametric variations of DCG and FSS in the calculation
of this study, are available to verify this trend.
As with the natural gas firings, the residual NOX is formed
in stoichiometric combustion in the vaporization and mixing region, and
it is very sensitive to the peak combustion temperatures. Again, the use
of combustion air temperature reduction, water sprays, and/or flue gas
recirculation (directly into this flame zone) should reduce this residual
NOX level.
Although the data shown in Figure 6 indicates no staged
combustion testing at air-fuel equivalence ratios less than 0. 85, these
boilers have now been operated for several years with (low sulfur) oil
fuels at ratios of 0. 55 to 0.7, with off-stoichiometric burner configu-
rations, again with no significant detrimental side effects.
5. 4. 3 Coal Fuels
General guidelines for maximum reduction of NOX from
coal-fired boilers are in some ways simpler than for the other fuel
types. NOx emissions are usually strongly dominated by conversion
of the fuel-bound nitrogen. Presently, no method of reduction of NOX
from this source other than staged combustion is apparent. For both oil
and coal, the gasification process is sufficiently slow that some of this
nitrogen conversion undoubtedly occurs well out into the furnace. In
order to assure minimum conversion of the fuel-bound nitrogen, then,
it appears highly desirable that the gasification and mixing processes be
completed under the minimum (overall burner) air-fuel ratio conditions.
This strongly suggests a staged combustion burner configuration rather
than off-stoichiometric configurations. In the latter case, early mixing
between adjacent active and air-only burners can raise the local air-
fuel ratios and allow some gasification and mixing under hot, oxidizing
conditions.
In the case of the coal-fired boilers, the related calcu-
lations are somewhat more empirical in that the NOX levels calculated
to be generated in the active burner region are all much too high. Al-
though the calculations used in the data comparisons shown in Figures
4(a) and (b) involve the direct calculation (term coefficients of 1. 0) of
NOX from fuel-bound nitrogen and the final mixing zone, the results of
71
-------
the regression analyses and the direct comparisons with measured
data were used to conclude that the coefficients applicable to the term
describing NOX thermally generated in the active burner region should
be set equal to zero. Also, with the single-wall boilers (only), the NOX
in the active burner region should be approximated by a constant level
(equal to 127 ppm).
The major source of NOX emissions from coal-fired
boilers is from conversion of the fuel-bound nitrogen, and the major
combustion modification technique for control is combustion staging.
Therefore, the effect of combustion staging on NOX from conversion of
fuel-bound nitrogen is of greatest interest in developing guidelines for
total NOX control. Figure 7(a) shows a typical plot of measured data
on overall NO emissions, and calculated curves on the NOX
emissions derived from conversion of the fuel-bound nitrogen in the
single-wall-fired boilers. Both are shown plotted against the burner
air-fuel equivalence ratio.
Combustion staging is accomplished in all of the coal-
fired boilers in the available data sample by cutting off the fuel flow to
some of the burners in the top burner level(s). Such combustion staging
has a different effect on NOX emissions than does varying the burner
air-fuel equivalence ratio by varying the overall boiler ratio with a
fixed burner configuration. Therefore, Figure 7(a) shows the calculated
NOX levels with four possible levels of combustion staging (62. 5 to 100
percent of the burners active) and the variation of these NOX levels
within each of these four staged combustion configurations resulting from
variations of the overall boiler air-fuel equivalence ratio between 1. 0
and 1.4. The measured data points (total NOX emissions) are coded to
correspond to the NOX levels calculated to result from, conversion of
fuel-bound nitrogen with the same staged combustion configuration. The
difference between the data and the related calculated curve is a measure
of the amount of thermally generated NOX.
The measured data show a general trend corresponding
to that shown by the calculation but are higher than the calculated levels
by about the constant 127 ppm discussed in Section 5. 2. 3. These data
are considered to verify the effects of staged combustion on NOX formed
from conversion of fuel-bound nitrogen but do not adequately explain the
high levels of thermal NOX apparently formed in these coal-fired boilers.
The trend for the thermally generated NOX to approach a very low (about
127 ppm) level with very fuel-rich burner operation seems to be reason-
ably established by most of the data, but the four data points at burner
equivalence ratios between 0. 6 and 0. 8 are in sharp disagreement. While
some reasonable conjectures concerning these four tests are possible,
none can be verified at this time.
Figure 7(b) shows a similar plot for the 360 MW tangen-
tial boiler in the data sample firing the same coal type as was fired in
72
-------
DATA
o
0
% OF BURNERS
ACTIVE
500
400
300
CO
o
00
to
^ 200
100
0
"BOILER
AIR-FUEL
EQUIVALfNCEA
RATIO = 1.0 v
FULL-LOAD OPERATION
O
O
%OF
BURNERS
ACTIVE
= 100
BOILER AIR-FUEL
EQUIVALENCE RATIO
= 1.4
0.6 0.8 1.0 1.2 1.4
BURNER AIR-FUEL EQUIVALENCE RATIO
(a) TWO SINGLE-WALL BOILERS FIRING A NOMINAL COAL
Figure 7. Effects of combustion staging on NOy emissions.
(Sheet 1 of 3.)
73
-------
DATA
O
% OF BURNERS
ACTIVE
100
80
FULL-LOAD OPERATION
500
400
300
.
O
1/T-
CO
200
X
100
0
BOILER AIR-FUEL
EQUIVAlfNCE
RATIO =
x/
O
O
O
O
%OF
BURNERS
ACTIVE
=100
BOILER AIR-FUEL
EQUIVALENCE RATIO
= 1.4
0.6 0.8 1.0 1.2 1.4
BURNER AIR-FUEL EQUIVALfNCE RATIO
(b) 360 AAW TANGENTIAL BOILER FIRING A NOMINAL COAL
Figure 7. Effects of combustion staging on NOx emissions.
(Sheet 2 of 3.)
74
-------
% OF BURNERS
DATA ACTIVE
FULL-LOAD OPERATION
500
400
E
Q_
oo
^.
O
CO
CO
c 300
200
100
0
O
100
80
O
O
BOILER AIR-FUEL
EQUIVALENCE
RATIO = 1.0
% OF
BURNERS
ACTIVE
= 100
BOILER AIR-FUEL
EQUIVALENCE RATIO
= 1.4
0.6 0.8 1.0 1.2 1.4
BURNER AIR-FUEL EQUIVALENCE RATIO
(c) 330-MW TANGENTIAL BOILER FIRING A HIGH-NITROGEN COAL
Figure 7. Effects of combustion staging on NO emissions.
(Sheet 3 of 3.)
75
-------
the single-wall boilers. As discussed in Section 5. 1, the tangential
boiler configuration appears to almost totally eliminate thermally
generated NOX throughout the boiler. The data shown in Figure 7(b)
tend to rather closely verify this, grouping about the appropriate cal-
culated staged combustion configuration curves. By extrapolating
these effects of staged combustion to a configuration in which only 60
percent of the burners are active, the calculation indicates that NOX
levels ranging from about zero to about 150 ppm could be achieved in
this boiler,,
Figure 7(c) shows a similar plot of the measured data
and the calculated levels with staged combustion in the 330 MW tangential
boiler in the data sample. Here, the data is even more directly in
agreement with the calculations. Again, the calculated extrapolation
indicates that NOX levels between zero and about 160 ppm (depending on
overall boiler excess air) should be possible with 40 percent staging of
the combustion air (60 percent of the burners active).
Figures 7 (a) through (c) show that staging about 40 per-
cent of the combustion air should reduce NOX emissions from con-
version of fuel-bound nitrogen to very low levels, particularly if the
boiler can be operated with low levels of excess air without generating
excessive emissions of carbon monoxides, unburned hydrocarbons, and
smoke. In tangential boilers, total NOX emissions should also be very
low with any level of combustion staging.
Single-wall, coal-fired boilers apparently thermally
generate considerable NOX emissions as well, possibly despite the
degree of combustion staging. This thermally generated NOX is
thought to originate in the subregion of the active burner region where
the coal gasification is taking place. Therefore, modifications to reduce
the combustion air temperature or the temperature rise due to com-
bustion (by dilution with recirculated flue gases, in the combustion air)
may be necessary, in conjunction with the combustion staging, to achieve
very low levels of total NOX emissions in face-fired boilers firing coal
fuels.
No data are available on long-term operation of coal-fired
boilers with as much as 40 percent of the combustion air diverted to a
second stage. Combustion instability may be a problem under such
operation but analytical techniques to provide stable combustion are
available [4 ]. The effects of such fuel-rich, reducing atmospheres
in the active burner region on the water walls (possible erosion/
corrosion) are not well-known but are being investigated in other research
programs.
76
-------
SECTION VI
EXAMPLE APPLICATION OF NOX REDUCTION GUIDELINES
The reasoning behind the data support and the probable
effects of the NOX reduction guidelines developed in this and the previous
studies [l, 2 ] are discussed in length in Section IV. An example appli-
cation is shown in this section simply to pull together the various obser-
vations and results into a practical demonstration. The example chosen
concerns modification of an existing boiler rather than the design of a
new boiler.
Only those modifications necessary to minimize NOX
will be discussed. It is clear that some of these modifications will
upset, to some degree (believed small), other aspects of boiler opera-
tion. These would have to be considered in the final, detailed boiler
modification. Nothing in this or the previous study justifies any attempt
here to comment on these other aspects of boiler design modifications
and operation.
No justification for the boiler chosen is offered. It is
simply considered a boiler, currently operated on natural gas fuel,
which might one day be converted to oil firing. It has not been fired
with oil fuel.
6. 1 PRELIMINARY CALCULATIONS
Hardware dimensions and certain operating condition
variables pertinent to this study are listed in Table 6. The table shows
data both as they exist when firing natural gas fuel and as they might be
modified for firing an oil fuel. The boiler NOX port and burner hard-
ware are essentially unchanged whether firing gas or oil except that a
smaller number of oil guns (24) than gas ports (48) are used. This
single change affects the (effective) vertical distance between the bur-
ners and between the top row of burners and the NO ports, the com-
bustion air flow area assigned to each burner (to each oil gun), and
the derived values for the furnace mixing zones and the combustion air
flow admittances.
Like all tangential boilers, even those without NOX ports,
the design of the burner stack, in each corner of the furnace,
77
-------
TABLE 6. VALUES OF EXAMPLE BOILER-SPECIFIC INPUT
VARIABLES
Variable
Boiler Geometry
Horizontal width, ft
Horizontal depth, ft
Number of firing walls
Burner Array
Total number of burners
Number of burners on a given level
Vertical distance between burners, ft
Horizontal width of a burner, ft
Burner air flow area, ft
NOX Ports
Number of ports
Vertical distance, top burner row to
ports, ft
Derived Mixing Zone Dimensions, ft
Primary and recirculation
Secondary
Adjacent/opposite
1 /2
Air Flow Admittances, lb,. -ft/ sec
Air-only burner
Active burners
NOX ports
Value
Gas
39
34
4a
48
4
1.51
2.50
3.78
4
12.2
4.39
9.07
18. 15
6.58
6.58
10. 86
Oil
39
34
4a
24
4
2.81
2.50
7.03
4
10.2
5.98
8.54
17.09
12.22
12.22
16.50
a. Four corners of a tangentially fired boiler
Units Conversion
meters = 0. 3048 x (feet)
Kilograms = 0.4535 x (pounds)
78
-------
effectively yields some builtin overfire air, or equivalent NOX ports.
This is because each fuel port is bracketed, above and below, by
approximately equal combustion air ports. For a vertical stacking
containing n fuel ports, therefore, there are n+1 combustion air ports.
The main effect of this is to reduce the amount of combustion air
associated with each fuel port to an n(n+l) fraction of the total. Thus,
the air-fuel ratio existing throughout the active burner region, even
with all burners active and no NOX ports, is l/(n+l) lower than the
overall boiler air-fuel ratio. The remaining l/(n+l) of the combustion
air is then admitted through the top combustion air port in an effective
second stage. In this example study, this effective overfire air is
lumped with the air admitted through the NOX ports. With the gaseous
fuel (12 fuel ports in a vertical stack), this additional overfire air
would represent 1/13 of the total burner air. With oil, the additional
overfire air would be 1/7 of the burner air. This is the reason that the
admittance to air flow through the effective, lumped NOX ports listed
in Table 6 is considerably larger with the oil fuel than with the gas.
The vertical distance between the top row of burners and the effective
NOX ports is different for this same reason.
There was insufficient data available for this study to
accurately calculate actual air flow admittances either for the NOX
ports or the burner air. The admittances listed in Table 6 are estimates.
Their ratios, however, were established to correspond to the estimated
air flow distribution between burners and NOX ports. The air flow
through all air ports in the burner stack was taken to be equal for all
ports, whether or not the associated fuel port was active, and the air
flow through the actual, designated NOX ports was taken to be 5 per-
cent of the total boiler air.
Not shown in Table 6, but necessary to the subsequent
NOX calculations, were data on the performance of the air-preheater.
These data showed that, at the 320 MW level used for this example
calculation, the temperature of the flue gases entering the air pre-
heater when firing natural gas fuel, is 636K (685 F) and that of the com-
bustion air leaving the preheater is 572K (570 F). No data were avail-
able with which to estimate the changes in the temperatures of the
combustion air between the air preheater exit and the burners. It was
estimated that a 28 to 56K (50 to 100°F) drop could occur.
6. 2 MODIFICATION FOR MINIMUM NOX
When firing this boiler with nitrogen-bearing oil fuels, a
significant amount of NOX emissions could be expected from conversion
of the fuel-bound nitrogen. All of the studies to date, however, indicate
that if the boiler is modified to operate with staged combustion such that
the air-fuel equivalence ratio in the active burners is of the order of
0. 6 to 0. 7, NOX from this source should be virtually eliminated, regard-
less of the nitrogen concentration in the fuel. Using the calculations
79
-------
of this study (a relatively small extrapolation from available data), a
design burner air-fuel equivalence ratio of 0. 638 was chosen. Con-
sidering the burner and NOX port admittances listed for oil in Table 6,
this burner air-fuel equivalence ratio could be achieved, with the boiler
operating with excess air equivalent to 3 percent oxygen, by operating
the top two levels of burners (8 burners) air-only (16 active burners)
and by restricting the NOX ports such that the admittance of an equiva-
lent NOX port (in one stack) is 15. 33, rather than the 16. 50 listed in
Table 6.
This low burner air-fuel ratio not only will minimize
NOX from conversion of fuel-bound nitrogen but will also virtually
eliminate any thermal NO formation in the active burner region except
that which might be formed in the subregion where fuel vaporization
takes place. In the equilibrium products of combustion from air-oil
mixtures with air-oil equivalence ratios less than about 0.7, both the
very low oxygen concentrations and the very low combustion tempera-
tures result in extremely low NOX formation rates by the thermal
mechanism.
This leaves the final mixing zone and the subregion of
fuel vaporization and mixing as the only remaining sources of significant
(thermal) NOX. Flue gas recirculation (into the combustion air) and/or
reduction in the combustion air temperature should minimize NOX from
both of these regions. The data from coal-fired tangential boilers in-
dicate that thermally generated NOX in any part of the active burner
region should be negligible. No similar data are available from oil-
fired tangential boilers, however, and, therefore, to be conservative,
the calculation appropriate to face-fired boilers was used in this example.
The question of the actual temperatures of the flue gases
and the combustion air in the burners of this example boiler is not
resolved. For this example calculation, it was simply assumed that
the temperatures of both the flue gases and the combustion air at the
burners will be reduced by 94K (170°F). This temperature reduction
was selected on the assumption that (1) up to 56K (100°F) reduction
could already be occurring as a result of heat losses in the ducting be-
tween the air preheater and the burners and (2) an additional 39K
(70°F) reduction is necessary to reduce the stoichiometric combustion
temperature of the oil fuel to that of natural gas. That part of the
94K (170°F) temperature reduction which is not accounted for in the
ducting losses should be accomplished by transferring more heat to the
steam cycle (perhaps in the economizer section) to avoid increasing
sensible heat losses up the stack. Suitable modifications in the pre-
heater operation and performance may be necessary. The resulting
temperatures of these gases at the burners, then, were taken to be
478K (400°F) for the combustion air and the recirculated flue gases,
respectively.
80
-------
With these temperatures, and the other burner, NO
port and boiler operation modifications, the necessary amount of flue
gas recirculation was treated as a parameter to calculate the remain-
ing total NOX emissions. The temperature reduction and the intro-
duction of even small amounts of recirculated flue gases virtually elim-
inate any NOX which might be thermally generated in the final mixing
zone. The very fuel-rich operation of the burners virtually eliminates
NOX generated from conversion of fuel-bound nitrogen or thermally
generated in any part of the active burner region, except in the sub-
region involving vaporization and mixing. The entire parametric cal-
culation involving the amount of flue gas recirculation, is reduced to
a calculation of the NOX thermally generated in the subregion of vapor-
ization and mixing.
Results of this calculation are shown in Figure 8. Also
shown are results of similar calculations for the existing configuration
firing natural gas fuel. The natural gas calculations are included to
show the strong effect of a reduction of 56K (100°F) in the gas tempera-
tures, as well as to provide a reference with which existing data can
be compared.
Figure 8 shows a strong reduction of the NOX emissions
with increasing flue gas recirculation. In the most stringent case, in-
volving a new oil-fired boiler located in the Los Angeles area, NOX
regulations for a 320 MW boiler require that NOX emissions be less
than about 37 ppm. To achieve this level with the oil-fired boiler con-
figuration and operating conditions described would require flue gas
recirculation into the burner air equal to 11. 8 percent of the burner air
(no flue gas recirculation into the NOX port air is required). This
would amount to about 11. Z percent of the total combustion air. The
same NOX level could be reached with natural gas fuel using flue gas
recirculation equal to 10 to 14 percent of the burner air. While data on
the NOX emissions from this boiler when fired with natural gas are not
currently available to this study, it appears that levels of the order of
35 ppm have been reached, but about 30 percent flue gas recirculation
was required.
6.3 SIDE EFFECTS
The example case of modification of an existing boiler
design for the purpose of NOX reduction results in operation with (1)
NOX ports open, (2) one-third (the top two levels) of the burners
operated air-only, (3) about 70 to 120°F reduction in combustion air
and recirculated flue gas temperatures, and (4) recirculation of flue
gases into the burner air in amounts equal to about 12 percent of the
burner air.
From the standpoint of plant efficiency, face-fired boilers
have been operated with oil fuel under such fuel-rich burner conditions
81
-------
TEMPERATURES, K(°F)
500
•a
E 400
a.
o.
t/f
O
«/> 300
LO
\
\
-\
CALCULATION FUEL
(D OIL
GAS
GAS
COMBUS-
TION AIR
478 (400)
572 (570)
517 (470)
FLUE
GASES
542 (515)
636 (685)
581 (585)
<
O
o
LU
5
GENERAL BOILER OPERATION
320-MW LOAD, 3% 02
NO PORTS OPEN
ONE THIRD OF BURNERS
OPERATED AIR-ONLY
FLUE GAS RECIRCULATION
INTO THE BURNERS
5 10 15 20
FLUE GAS RECIRCULATION, % OF BURNER AIR
Figure 8. Parametric calculations to determine the necessary
flue gas recirculation in the example boiler.
82
-------
with no observable reduction in overall plant efficiency. Some evidence
of combustion instability has been observed under such fuel-rich burner
conditions, but analytical techniques are available [4] to provide stable
combustion. Similarly, with adequate excess air in the second stage
(3 percent C>2 or greater), no problems with excessive emissions of
carbon monoxide, unburned hydrocarbons, or smoke are apparent.
Face-fired boilers have been operated for some time with very fuel-
rich first stage combustion of oil and natural gas fuels, with no apparent
detrimental effects to the water walls. Thus, the modification for the
very fuel-rich burner operation does not appear to create any new and
undesirable side effects.
The reduction in combustion air temperatures would
cause a small loss in plant efficiency if it were accomplished simply by
transferring less heat in the preheater and allowing hotter gases to
escape up the stack. If it is accomplished by transferring more heat to
the steam cycle as recommended, however, no efficiency losses are
expected.
The combination of very fuel-rich burner operation,
reduction in combustion air and recirculated flue gas temperatures at
the burners, and the introduction of recirculated flue gases into the
combustion air, together, could cause some flame liftoff and/or com-
bustion instability problems. There is little experience with this kind
of fuel-rich, lower combustion termperature burner operation. In
theory, the very fuel-rich operation should increase the tendency for
combustion instability, but the lower reactant temperatures and the
dilution of the heat release rate per unit volume flow should decrease
this tendency. All three of these modifications should make flame-
holding more difficult. Thus, possible problems in both combustion
instability and flame liftoff should be anticipated, at least until tests
indicate no problem.
83
-------
REFERENCES
1. O. Wo Dykema, Analysis of Test Data for NOX Control in
Gas- and Oil-Fired Utility Boilers, EPA-650/2-75-012
(NTIS PB 241918/AS), U. S. Environmental Protection Agency,
Research Triangle Park, North Carolina, January 1975.
2. O. W. Dykema, Analysis of Test Data for NOX Control in
Coal-Fired Utility Boilers, EPA-600/2-76-274 (NTIS PB
261066/AS), U. S. Environmental Protection Agency,
Research Triangle Park, North Carolina, October 1976.
3. O. W. Dykema and R. E. Hall, "Analysis of Gas-, Oil-
and Coal-Fired Utility Boiler Test Data, '' Proceedings of
the EPA Symposium on Stationary Source Combustion,
EPA-600/2-76-152c (NTIS PB 257146/AS), U. S. Environ-
mental Protection Agency, Research Triangle Park, North
Carolina, June 1976.
4. O. Wo Dykema, Effects of Combustion Modifications for NOX
Control on Utility Boiler Efficiency and Combustion Stability,
EPA-600/2-77-190, U. S. Environmental Protection Agency,
Research Triangle Park, North Carolina, September 1977.
5. O. W. Dykema, "Analysis of NOX Control in Stationary
Sources, " Proceedings of the Second Stationary Source Com-
bustion Symposium, V2 Utility and Large Industrial Boilers
EPA-600/7-77-037b, U. S. Environmental Protection
Agency, Research Triangle Park, North Carolina, July 1977.
6. A. R. Crawford, E. H. Manny and W. Bartok, Field Testing:
Application of Combustion Modifications to Control NOx
Emissions from Utility Boilers, EPA-650/2-74-066 (NTIS
PB 237344/AS), U. S. Environmental Protection Agency,
Research Triangle Park, North Carolina, June 1974.
7. D, Aronowitz and I. Glassman, "Some Results on the Oxidation
of Methanol, " presented at the fall technical meeting of the
Eastern Section of the Combustion Institute, East Hartford,
Connecticut, November 10-11, 1977.
84
-------
8. P. Lewis, et al. , "A Model for Entrained Flow Gasifiers, "
presented at the fall technical meeting of the Eastern Section
of the Combustion Institute, East Hartford, Connecticut,
November 10-11, 1977.
9. Y. H. Song, J, M. Beer, and A. F. Sarofim, "Fate of Fuel
Nitrogen During Pyrolysis and Oxidation, " MIT Industrial
Liaison Program, September 1977.
10. J. O. L. Wendt and J. M. Ekmann, "Effects of Fuel Sulfur
on Nitrogen Oxide Emissions, " Proceedings of the EPA
Symposium on Stationary Source Combustion, EPA-
600/2-76- 152a, (NTIS PB 257146/AS), U. S. Environmental
Protection Agency, Research Triangle Park, North
Carolina, June 1976.
85
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA-600/2-78-217
3. RECIPIENT'S ACCESSION-NO.
4. TITLE AND SUBTITLE
Combustion Modification Effects on NOx Emissions
from Gas-, Oil-, and Coal-Fired Utility Boilers
5. REPORT DATE
December 1978
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Owen W. Dyke ma
8. PERFORMING ORGANIZATION REPORT NO
9. PERFORMING ORGANIZATION NAME AND ADDRESS
The Aerospace Corporation
Energy and Resources Division
Los Angeles, California 90009
10. PROGRAM ELEMENT NO.
1AB014
11. CONTRACT/GRANT NO.
Grant R803283-03
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Task Final: 7/76 - 8/78
14. SPONSORING AGENCY CODE
EPA/600/13
15.SUPPLEMENTARY NOTES IERL_RTP project officer is Robert E. Hall, Mail Drop 65, 919/
541-2477. EPA-600/2-76-274 was an earlier report in this series.
16. ABSTRACTTjine rep0rt representsithe conclusion of 4 years of analysis of large quanti-
ties of emissions, operating conditions , and boiler configuration data from full-scale
multiple-burner, electric-generating boilers firing natural gas, oil, and coal fuels.
The overall objective of the study was to develop from this data: (1) further under-
standing of the effects of combustion modifications on combustion, and the resulting
effects on NOx emissions: and (2) directly applicable guidelines for the application
of combustion modification techniques for the control of NOx emissions in full-scale
operating utility boilers. The report includes: (1) discussion of modeling techniques
used to analyze the data: (2) conclusions relative to the sources of NOx within the
furnace; (3) guidelines for NOx reduction; and (4) an example application of the guide-
lines. Boiler firing types include single-wall, opposed and tangential configurations.
The report concludes that NOx emissions are generated, in varying degrees, from
conversion of fuel-bound nitrogen (the predominant source), heterogeneous combus-
tion and mixing zone, second-stage mixing zone, and active burner region. Main-
taining very fuel-rich initial combustion conditions, holding the initial peak combus-
tion temperature to <2050 K, and delaying fuel gasification and mixing until the gas
has been cooled somewhat should reduce NOx emissions from all four main sources.
7.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS
c. COS AT I Field/Group
Air Pollution
Combustion
Emission
Boilers
Nitrogen Oxides
Coal
Natural Gas
Fuel Oil
Electric Power
Plants
Mathematical
Models
Air Pollution Control
Stationary Sources
Utility Boilers
Combustion Modification
Staged Combustion
13B
2 IB
13A
07B
21D
10B
12A
8. DISTRIBUTION STATEMENT
Unlimited
19. SECURITY CLASS (ThisReport)
Unclassified
21. NO. OF PAGES
97
20. SECURITY CLASS (Thispage)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
86
------- |