Grantee Research Project Results
2005 Progress Report: Methodology for Assessing the Effects of Technological and Economic Changes on the Location, Timing and Ambient Air Quality Impacts of Power Sector Emissions
EPA Grant Number: R831836Title: Methodology for Assessing the Effects of Technological and Economic Changes on the Location, Timing and Ambient Air Quality Impacts of Power Sector Emissions
Investigators: Ellis, Joseph H. , Burtraw, Dallas , Hobbs, Benjamin F. , Palmer, Karen
Institution: The Johns Hopkins University , Resources for the Future
EPA Project Officer: Chung, Serena
Project Period: February 1, 2005 through January 30, 2008 (Extended to January 30, 2009)
Project Period Covered by this Report: February 1, 2005 through January 30, 2006
Project Amount: $648,733
RFA: Regional Development, Population Trend, and Technology Change Impacts on Future Air Pollution Emissions (2004) RFA Text | Recipients Lists
Research Category: Climate Change , Air
Objective:
The amounts, locations, and timing of power sector emissions are sensitive to economic and technological assumptions. To more confidently estimate future temporal and spatial scenarios of emissions, a theoretically defensible, transparent, and practical methodology potentially is helpful. The objective of this research project is to develop a methodology through the use of a sequence of models representing market-driven electricity supply and facility location. The primary activities under the energy facility siting task have included improving the empirical siting analyses using an expanded dataset and performing capacity siting and NOx emissions analyses.
Progress Summary:
We summarize the major accomplishments in this progress report, including capacity siting scenarios and an analysis of the effect of climate change upon the temporal and spatial distribution of NOx emissions in the mid-Atlantic and Midwest power markets. The Haiku model, developed by Resources for the Future, is used to provide regional technology, demand, and emissions totals and disaggregates national totals to regions. We use it to explore systematically the sensitivity of both emissions and ambient air quality results to these uncertain drivers to assess what assumptions and model components matter most. Ambient air quality (tropospheric ozone and particulates), for an example set of scenarios, are simulated using Mesoscale Model 5 (MM5)/Meteorology-Chemistry Interface Processor/Sparse Matrix Operator Kernal Emissions (SMOKE)/Community Multiscale Air Quality (CMAQ). This will demonstrate the practicality of integrating the source disaggregation methodology with the SMOKE emissions processing system and subsequently the CMAQ transport and fate model itself.
Empirical Siting Analyses
The empirical siting models have been reestimated using county-level data from years 1995-2004. During 1995-2004, the generating capacity for the entire continental United States increased by 43 percent, from 686 to 982 GW. The distribution of capacity addition by technology in number of generating units defined as individual boilers or machines (and in their total generating capacity) for steam, combined-cycle units, combustion turbine (CT), and other is 29.7 percent (16.9%), 21.5 percent (46.9%), 13.4 percent (27.2%), and 43 percent (9%), respectively.
The empirical models use a logit specification that relates the probability that a generating unit of a particular type has been sited in a county to predictive variables, including air quality status, population, preexisting power plants of various types, state deregulation status, and proximity of transmission lines. Overall, the most important factor that drives siting decisions is the presence of existence generating facilities, especially coal and CT. Other variables also can be important; for instance, air quality attainment status generally has positive impact on siting probability, except for a negative impact of severe/serious status on siting CT. This seemingly counterintuitive result indicates that attainment status is a proxy for other variables (such as proximity of industrial load) that positively influences siting probabilities.
Capacity Siting
The equations are used to estimate probabilities of siting of various facility types by county within the Mid-Atlantic and Midwest study areas. These probabilities are an input to an integer programming model that allocates new capacity among counties. This new capacity has been allocated previously to multicounty subregions (power control areas) by a regional capacity expansion model. The integer program maximizes the match between the empirical probabilities and the allocation of facilities, respecting minimum and maximum capacity constraints resulting from, for instance, the need for realistic facility sizes.
Figure 1 is a graphic display of the capacity siting results for a scenario for Mid-Atlantic 2025 capacity siting patterns in which electric loads reflect no climate change (1990s climate, left) and significant climate change (2050s climate, right), respectively. As a reference, the figure also indicates the location of several large cities. Compared with the capacity allocation under 1990s climate scenario, there is more CT capacity under the 2050s climate because of the fact that higher temperatures translate into higher peak loads. This new capacity is predicted to be installed mostly in two power control areas within Pennsylvania New Jersey Maryland Interconnection, LLC (PJM)—Pennsylvania Power & Light and Baltimore Gas & Electric/Potomac Electric Power Company—with lesser amounts in Pennsylvania Electric’s and New Jersey Central’s areas. The spatial allocation of new capacity is crucial for air pollution analyses because it determines the spatial distribution of pollution emissions.
Figure 1. Allocation of New Capacity in PJM Region under Scenarios of Normal Climate (1990s, left) and Climate Change (2050s, right)
Temporal Distribution of NOx Emissions
After determining the mix and location of new power facilities for the 2025 scenarios with and without climate change, the operation of the facilities during the summer ozone season was simulated. Figure 2 shows average NOx emissions duration curves for 1990s (upper) and 2050s climates (lower), respectively, for the ozone season. The plots (from left to right) are for the combined PJM and East Central Area Reliability Council (ECAR) markets, ECAR separately, and PJM separately, respectively. The solid curve in each plot is the average over the years considered for the 1990s and 2050s series. To characterize their uncertainty, we perform bootstrapping with a sample size of 1,000; the 95 percent confidence interval then constructed using standard deviation from the results of bootstrapping is represented by dash lines. Contrasting the upper and lower sets of figures shows that the 2050s climate results in a peakier distribution of emissions; more NOx emissions are occurring during the hottest days, when their impacts on tropospheric ozone likely would be most severe.
Figure 2. NOx Emissions Duration Curves for 2025 under a 1990s Climate (Upper) and 2050s Climate (Lower)
Overall, the variability of hourly emissions is larger during high emissions periods. For instance, the coefficient of variation (ratio of the standard deviation to mean) of the highest-emission period (25 hours) is approximately equal to 7.2 percent. This is compared with 2 percent to 5 percent for other periods (50-300 hours). Thus, uncertainty associated with emissions is greatest precisely during those time periods (peak demands during hottest days) when the emissions would be of greatest concern.
Spatial Distribution of NOx Emissions
Next, we examined the spatial distribution of NOx emissions among power control areas in our Mid-Atlantic/Midwest model. Table 1 presents the 2025 NOx emission by power control areas for the 1990s and 2050s climates, respectively. These emissions are calculated from the same market simulation model used to obtain the temporal distributions shown above. Each climate represents an average of the emissions in the several years that are simulated for each climate. Total emissions under the two climates are identical because of the assumed presence of a NOx cap. The spatial distributions differ somewhat. Under both scenarios, the generating units located in ECAR are the major source of NOx emissions. The NOx emissions of a number of PJM power control areas, however, are significantly affected by climate conditions, as indicated by the percentage changes in the last column. These differences reflect the siting scenarios, which have placed more peaking (CT) capacity in certain PCAs.
Table 1. Spatial Distribution of Average NOx Emissions for 2025 Power System Under Two Climate Scenarios (Tons/Ozone Season)
Power Control Area |
1990s Climate, Average |
2050s Climate, Average |
Percentage Difference |
ECAR |
91145 |
92418 |
1.4% |
PJM-AE |
1618 |
2179 |
34.8% |
PJM-BGE2 |
2319 |
2131 |
-9.2% |
PJM-BGEPEP |
5058 |
5153 |
1.7% |
PJM-DPL |
309 |
366 |
21.8% |
PJM-JC1 |
479 |
480 |
0.1% |
PJM-JC2 |
116 |
429 |
270.4% |
PJM-ME1 |
337 |
906 |
171.8% |
PJM-ME2 |
305 |
1329 |
345.3% |
PJM-PE |
174 |
205 |
14.3% |
PJM-PN |
6055 |
6534 |
7.2% |
PJM-PPL1 |
9152 |
7517 |
-17.7% |
PJM-PPL2 |
411 |
1169 |
184.4% |
PJM-PPL3 |
10529 |
7576 |
-28.1% |
PJM-PS |
1094 |
710 |
-35.0% |
TOTAL |
129101 |
129102 |
|
Electricity Supply
The previous version of Haiku identifies detailed technology scenarios over approximately a 20-year horizon. Our challenge is to extend electricity supply in Haiku to a 50-year horizon while preserving as much technological detail as possible. This involves improvement in the solution algorithm and characterization of new and emerging supply side technologies with the goal of representing not only emerging technologies but also the economic survival of existing power plants, and uncertainties associated with these scenarios.
Improvements in the solution algorithm have involved broad scale refinements within the existing model framework. The following steps have been accomplished:
- Expansion of the number of subregions of the nation modeled as independent and linked electricity markets from 13 to 20. This change enables greater precision in the characterization of technology within the regions and enabled improvements in the pollution control algorithm.
- Full integration of the pollution control algorithm that governs choices at individual plants for compliance with the set of pollution constraints faced by the firm. This integration accounts for the interaction of pollution controls and fuel choice for various pollutants, as well as the irreversibility and opportunity cost associated with the timing of postcombustion installations.
The following activities on supply-side modeling are underway:
- Improvement in model calculations to speed convergence. This step is essential to be able to address uncertainty with either formal statistical methods or scenario analysis.
- Improved characterization of interregional transmission capability and cost, including characterization of emerging technologies that may prove to be an important part of the electricity grid in the long run.
Electricity Demand
Previously, the Haiku model embodied constant elasticity functions for electricity demand that respond to changes in price. This characterization is limited because a potentially significant feature of the electricity sector in the future is greater use of time-varying prices or alternatively greater use of demand management by supply companies and customer aggregators. The time-variation of demand has important implications for the time of supply and the formation of and exposure to air pollution. This year we have embarked a major initiative to revise the demand portion of the model. The following activities are underway:
- The demand module in Haiku is being reconfigured to allow for cross-time block elasticities. That is, the demand at one time of day will depend not only on price in that time of day but also price in other times of day. A key decision is the functional form, and numerous alternatives have been explored. The probable choice is a constant elasticity of substitution (CES) form for the demand function. Detailed documentation on this modeling choice also is being developed, and this will lead to a stand-alone publication.
- The new demand module has been tested in prototype form in the model, and it solves quickly. The closure of the model will require calculation of elasticities for each temporal relationship. Because the empirical literature does not offer any one obvious choice for functional form or parameter values, we will use conditions suggested by theory to fill in missing values such that they are consistent with the patchwork of values that exist in the literature.
Air Quality Modeling
Two major tasks have been accomplished successfully in the previous year. We have migrated to the current version (V2.2) of the SMOKE emissions processing system. This includes as well, the ability to use more recent emissions inventories (previously the National Emissions Trends 96 suite of inventories had been used). The National Emissions Inventory (NEI) 99 and NEI01 inventories now are operational (they also involve a newer, more detailed source categorization consistent with SMOKE V2.2). Our existing procedure of modifying individual plant hourly emissions via point source hour-specific emissions also was checked, debugged, and made operational with the new version of SMOKE and the newer inventories.
We also have migrated to CMAQ V4.5 and successfully configured and employed its parallel processing capabilities. This involved a considerable amount of system tuning and experimentation but appears to be well worth the effort. Using the U.S. national 36 km grid, a 1-day simulation typically took about 4 hours of total system time on a single processor machine. Using our parallel linux cluster with eight nodes devoted to the job (using Message Passing Interface) that same simulation now takes about 20 minutes of total system time. We currently have a total of 35 compute nodes in the cluster in addition to the master node, thus further speedup will be possible. The cluster is monitored using GANGLIA and can be viewed at http://128.220.52.80/index.php Exit .
Future Activities:
Siting and Dispatch
During Year 2 of the project, the following activities will be undertaken: (1) Interfacing of regional capacity allocation and siting models with Haiku, in which a given capacity mix and energy generation amounts at the regional level will be provided by Haiku and disaggregated to the power control area and county level by the capacity planning, operations, and siting models developed in Year 1; (2) improvements in the siting models to reflect emerging technologies, such as wind; (3) improvements in the capacity allocation models to reflect resource limitations that affect siting decisions, including water availability and transmission upgrades; and (4) generation of spatially and temporally disaggregate emissions scenarios for use in CMAQ.
Supply and Demand
During Year 2, we will initiate two further activities: (1) characterization of emerging generation technologies, which will involve review of cost and cost projections, learning parameters, and abstract representation of technologies that are not yet available commercially; and (2) review of technology trends for currently mature technologies.
In the future year, we will initiate a final activity to achieve closure with the demand module and make it available for the simulations that will be part of the project. The selected demand functions, probably of CES form, are reduced form representations of demand that are calibrated to observed data. Underlying these functions there should be structural parameters that characterize the driving factors that determine demand, such as persons per household, average square foot per household, miscellaneous appliance use, and so forth. This step of the project will construct and calibrate these structural relationships, which then can provide a vehicle for modeling the evolution of consumer demand in the long run as demographic features of the population evolve in a predictable manner.
Air Quality Modeling
The upcoming year initially will involve running a comprehensive suite of baseline scenarios for multiple years using the U.S. national grid. The intent here is to create time-series of concentrations for ground level ozone and particulate matter that subsequently will be used for comparison when future year scenarios are run. Given the considerable magnitude of the computing and data manipulation involved, this necessitates the development of automated scripts, for example, to identify a particular grid cell given a latitude-longitude pair (accomplished via a perl script), then extract from MM5 output, select meteorologic output (e.g., 1.5 m temperatures) and extract from CMAQ output, and select layer 1 pollutant (ozone and particulate matter) concentrations. The next step is semiautomated graphing and visualization of the extracted information using PAVE.
Journal Articles on this Report : 2 Displayed | Download in RIS Format
Other project views: | All 16 publications | 9 publications in selected types | All 7 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Chen Y, Hobbs BF. An oligopolistic power market model with tradable NOx permits. IEEE Transactions on Power Systems 2005;20(1):119-129. |
R831836 (2005) R828731 (2002) R828731 (Final) R828733 (Final) |
Exit Exit |
|
Chen Y, Hobbs BF, Leyffer S, Munson TS. Leader-follower equilibria for electric power and NOx allowances markets. Computational Management Science 2006;3(4):307-330. |
R831836 (2005) R831836 (2007) R831836 (2008) R831836 (Final) R828731 (2003) R828731 (Final) |
Exit |
Supplemental Keywords:
power sector emissions, electrical energy production and demand, ambient air quality, climate change, regional facility location, atmospheric models, climatic influence, emissions impact, environmental monitoring,, RFA, Scientific Discipline, Air, climate change, Air Pollution Effects, Environmental Monitoring, Ecological Risk Assessment, Atmosphere, air quality modeling, atmospheric carbon dioxide, ecosystem models, electrical energy, climatic influence, emissions impact, green house gas concentrations, modeling, carbon dioxide, climate models, CO2 concentrations, demographics, electric power sector emissions, ambient air pollution, atmospheric models, Global Climate ChangeRelevant Websites:
http://128.220.52.80/index.php Exit
Progress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.