EMVAP –Emissions Variability Processor...Status Report on Development and Testing of the EMVAP...
Transcript of EMVAP –Emissions Variability Processor...Status Report on Development and Testing of the EMVAP...
Status Report on Development and Testing of the EMVAP Systemand Testing of the EMVAP System
EMVAP – Emissions Variability Processor Bob Paine AECOMBob Paine, AECOM
10th EPA Modeling Conference, March 14, 2012
Other contributors:Other contributors:
David Heinold and Rich Hamel, AECOM
El di K i i d N h K EPRIEladio Knipping and Naresh Kumar, EPRI
1
Sponsored by
Outline of PresentationOutline of Presentation
• EPA Modeling Guideline proceduresEPA Modeling Guideline procedures
• Nature of variable emissions distributions
• Description of procedures for Monte Carlo analysisDescription of procedures for Monte Carlo analysis approach to handle variable emissions: EMVAP
• Evaluation of EMVAP with 3 AERMOD evaluationEvaluation of EMVAP with 3 AERMOD evaluation databases
• Sensitivity of results to number of simulated yearsy y
• Conclusions
2
EPA Appendix W Modeling Proceduresd l d (f h• Modeled emission rate input (for short‐term
averages) is:
E i i li it ti l l ti f t– Emission limit x operating level x operating factor
(lb/MMBtu) (MMBtu/hr) (hours/year)
– Max. emission x design x continuous
limit capacity operation
• Modeling continuous operation for intermittent sources or maximum emission limits for variable
i i i f ti l l femission sources is of concern, particularly for a probabilistic NAAQS 3
Emission Rate Variability• Large variation possible over the course of a year
• Intermittent sources (e.g., emergency backup engines Intermittent sources (e.g., emergency backup enginesor bypass stacks) present modeling challenges
• For these sources, assuming fixed peak 1‐hour , g pemissions on a continuous basis will result in unrealistic modeled results
• Better approach is to assume a prescribed distributionof emission rates
• EMVAP (Emissions Variability Processor), described below, uses this information to develop alternative
t i di t d l d li i fways to indicate modeled compliance using a range of emission rates instead of just one value
4
Example of Hourly Emissions Sequencep y q
Example Emission Cumulative Frequency DistributionFrequency Distribution
(g/s)
mission
rate
Em
% cumulative frequency over all days
Example Emission Cases for EMVAPExample Emission Cases for EMVAP1) <=140 g/s 100% of the time2) <=100 g/s 98% of the time
Peak observed emission rate 133 g/s
(g/s)
2) < 100 g/s 98% of the time3) <= 65 g/s 89% of the time4) <= 40 g/s 84% of the time5) <= 25 g/s 77% of the time6) Off 20 % f h i
= 133 g/s
mission
rate 6) Off 20 % of the time
Weighted average emissionrate = 43.4 g/s, less than 1/3
Em
g/ , /of the peak value
% cumulative frequency over all days7
Approach (“EMVAP”) for Multiple Allowable Emissions ModelingAllowable Emissions Modeling
• Create an emissions frequency distribution
• Model the source with unit emissions (up to 5 “real” years) – different runs maybe needed over a range of h t texhaust parameters
• Create many (e.g. 1,000) simulated annual realizations of conc with random number generator for emission rateconc. with random number generator for emission rate
• Randomly assign an emission rate multiplier for each hour using the source‐specific emissions distributionhour using the source‐specific emissions distribution
• Process summary statistics over each year/receptor
• Use post processing software to add concentrations for• Use post‐processing software to add concentrations for multiple sources plus background
8
Random Selection ProcessRandom Selection Process• In some cases, peak emissions occur in groups of hhours
• The form of the 1‐hour NO2 and SO2 standard i l l th hi h t t ti h iinvolves only the highest concentration hour in any given day
• Therefore it is likely conservative to distribute peak• Therefore, it is likely conservative to distribute peak emission rates randomly rather than in groups
• Use of a random selection process such as a Monte• Use of a random selection process, such as a Monte Carlo procedure, is appropriate
• But sources that operate in tandem can be treatedBut, sources that operate in tandem can be treated with the same sequence of random numbers
9
Purpose and Definition• The EMVAP system is a probabilistic post‐processor for AERMOD designed to more realistically model emission g ysources against short‐term NAAQS
• The EMVAP system consists of three modules + AERMOD:y
– EMDIST emissions analyzer : aids in determining emission inputs for AERMOD runs
– EMVAP probabilistic emission simulator: used to randomly generate modeled concentrations based on source emissions frequencies
– EMPOST post‐processor: takes EMVAP output and performs statistical analyses, generating modeled concentrations in the form of the NAAQS 10
EMDISTEMDIST
• Statistical Analysis to Aid in Emissions Input to y pEMVAP
• Required InputsTakes in up to 5 years of emissions data in either AERMOD– Takes in up to 5 years of emissions data in either AERMOD input file format or put into EMDIST format (described in read‐me file)
• Optional Inputs• Optional Inputs– If hourly stack exit temperature and velocity data are available, it is used in recommended emission case calculationscalculations
– EMDIST can also import hourly ambient temperature data and report corresponding emission rates and ambient temperature datatemperature data
11
EMVAP• Probabilistic Emissions Simulator Post‐processor for AERMOD (used after AERMOD is run for all cases)
• Required Inputs• Required Inputs– Number and list of years included in the analysis– The NAAQS standard for which the EMVAP results are to b dbe compared
– The number of Monte Carlo simulations to perform– File containing the receptors used in the AERMOD runs– Random number file used in generating emissions randomly each hour – sources can be linked with a common sequence of random numbers
– For each of up to 10 source data sets:• The load cases and names of the AERMOD result files • Emission cases and associated frequencies
12
EMPOSTEMPOST
• Results Determination and Reporting for EMVAPResults Determination and Reporting for EMVAP
• Required Inputs
– Number and list of years in the analysisNumber and list of years in the analysis
– Name of the file containing receptor coordinates.
The number of modeling iterations performed in– The number of modeling iterations performed in EMVAP and the names of the iteration files
– Definition of statistics to report if desiredDefinition of statistics to report, if desired
13
EMVAP EvaluationEMVAP Evaluation
• Selected 3 AERMOD Databases with variety of terrainSelected 3 AERMOD Databases with variety of terrain settings
• Ran AERMOD with both actual and constant peak p(allowable) hourly emissions – got 99th percentile peak daily 1‐hour max pre vs. obs
• Ran EMVAP to get the same result from median value over 1000 simulated years
• Expectation: EMVAP result would be between that of actual and allowable emissions
14
Evaluation DatabasesEvaluation Databases
• Lovett Generating Station – complex terrain (Hudson• Lovett Generating Station – complex terrain (Hudson River Valley)
– 1 full year test case 8 monitors1 full year test case, 8 monitors
• Clifty Creek Generating Station – Ohio River gorge
– 1 full year with 3 units with differing load profiles– 1 full year with 3 units with differing load profiles, 6 monitors
• Kincaid Power Station – flat corn fields of IllinoisKincaid Power Station flat corn fields of Illinois
– Partial year case, 1 stack, 28 monitors
15
Lovett
16
Frequency Distribution of SO2 Emissions at L tt 1988
300
350
Lovett, 1988
250
300
ons (g/s)
150
200
SO2Em
issio
50
100
Hou
rly
0
50
0 10 20 30 40 50 60 70 80 90
Percent of Emissions
17
Percent of Emissions
Lovett Generating Station – Exit Velocity vs. Emission Rate
40Residual Plot as a Function of Emissions and Exit Velocity
30
35
20
25
ty (m
/s)
10
15
Exit
Velo
cit
0
5
1.234 - 75 (n = 1053) 75 - 90 (n = 931) 90 - 140 (n = 2502) 140 - 220 (n = 2245) 220 - 255 (n = 448) 255 - 395.806 (n =
E
1.234 75 (n 1053) 75 90 (n 931) 90 140 (n 2502) 140 220 (n 2245) 220 255 (n 448) 255 395.806 (n 433)
SO2 Emission Cases (g/s)18
600
)EMVAP 50th Percentile Results for Lovett Generating Station
450
500
550
ue*) (u
g/m3)
EMVAP results are all much better h i k i i
300
350
400
(Design Va
lu than using peak emissions.
200
250
300
ncen
tration (
50
100
150
1‐hr SO
2Co
n
0
Observed (Monitor)
AERMOD with Actual
Emissions
AERMOD with Maximum Emissions
EMVAP 5‐Case Equal Freq. Distribution
EMVAP 10‐Case Equal
Freq. Di t ib ti
EMVAP 20‐Case Equal
Freq. Di t ib ti
EMVAP 5‐Case Emission Rate Distribution
EMVAP 10‐Case Emission
Rate Di t ib ti
EMVAP 20‐Case Emission
Rate Di t ib ti
1
19
Distribution Distribution Distribution Distribution
EMVAP Cases * Design Value is 99th Percentile of the daily maximum 1‐hour average
CliftyCreek
20
Frequency Distribution of SO2 Emissions at Clifty Creek Stack 1 1975
3500
4000
Clifty Creek Stack 1, 1975
3000
3500
ons (g/s)
2500
SO2Em
issio
1500
2000
Hou
rly
1000 0 10 20 30 40 50 60 70 80 90 100
f i i
21
Percent of Emissions
Frequency Distribution of SO2 Emissions at Clifty Creek Stack 2 1975
3500
4000
Clifty Creek Stack 2, 1975
3000
3500
ns (g
/s)
2500
SO2Em
ission
1500
2000
Hou
rly S
1000
1500
0 10 20 30 40 50 60 70 80 90 100
22
Percent of Emissions
Frequency Distribution of SO2 Emissions at Clift C k St k 3 1975
4000
Clifty Creek Stack 3, 1975
3000
3500
ns (g
/s)
2500
SO2Em
ission
1500
2000
Hou
rly S
1000 0 10 20 30 40 50 60 70 80 90 100
f i i
23
Percent of Emissions
15001550160016501700
m3)
EMVAP 50th Percentile Results for Clifty Creek Generating StationEMVAP results are improved over using peak emissions.
1150120012501300135014001450
lue*) (ug
/m
g p
750800850900950100010501100
n (Design Va
400450500550600650700750
oncentration
50100150200250300350400
1‐hr SO
2Co
0
Observed (Monitor)
AERMOD with Actual
Emissions
AERMOD with Maximum Emissions
EMVAP 5‐Case Equal Freq. Distribution
EMVAP 10‐Case Equal
Freq. Distribution
EMVAP 20‐Case Equal
Freq. Distribution
EMVAP 5‐Case Emission Rate Distribution
EMVAP 10‐Case Emission
Rate Distribution
EMVAP 20‐Case Emission
Rate Distribution
24
EMVAP Cases * Design Value is 99th Percentile of the daily maximum 1‐hour average
Kincaid
25
Frequency Distribution of SO2 Emissions at
13000
q y 2
Kincaid Power Station
9000
11000
ns (g
/s)
7000
O2Em
ission
3000
5000
Hou
rly S
1000
3000
0 10 20 30 40 50 60 70 80 90 100
f i i
26
Percent of Emissions
125013001350
3)EMVAP 50th Percentile Results for Kincaid Generating Station
100010501100115012001250
ue*) (u
g/m3
700750800850900950
(Design Va
lu
400450500550600650
ncen
tration
100150200250300350
1‐hr SO
2Co
n
050
Observed (Monitor)
AERMOD with Actual
Emissions
AERMOD with Maximum Emissions
EMVAP 5‐Case Equal Freq. Distribution
EMVAP 10‐Case Equal
Freq. Di t ib ti
EMVAP 20‐Case Equal
Freq. Di t ib ti
EMVAP 5‐Case Emission Rate Distribution
EMVAP 10‐Case Emission
Rate Di t ib ti
EMVAP 20‐Case Emission
Rate Di t ib ti
27
Distribution Distribution Distribution Distribution
EMVAP Cases* Design Value is 99th Percentile of the daily maximum 1‐hour average
Sensitivity Analysis of Design Conc. By Number 3of Iterations Performed – Lovett (g/m3)
Iterations 50th 75th 90th Max.
50 348.99 374.81 413.85 458.40
500 347.32 374.81 412.31 460.45500 347.32 374.81 412.31 460.45
1000 347.32 374.81 412.31 534.39
2500 347.32 374.81 412.31 541.62
5000 347.32 374.81 412.31 614.7828
Current Limits in EMVAPCurrent Limits in EMVAP
• Receptors: no effective limit (tested so far with p (10000 receptors)
• Source groups to be combined: 10 (can include h b k d)groups with constant emissions, or background)
• Load cases per source group: 20It ti 5000 i l t d• Iterations: 5000 simulated years
• Years of modeled data per iteration: 5
Typical run time is a few minutes to an hour on a standard computing platform.standard computing platform.
29
Conclusions and StatusConclusions and Status
• EMVAP is currently operational for EPRI beta testingEMVAP is currently operational for EPRI beta testing and consideration of implementation approaches
• Evaluation against field data shows expected results: g pcritical predictions are somewhat higher than those from actual emissions and lower than those from peak emissions
• Further development and testing is currently d bunderway by EPRI
30