The Pennsylvania State University. The Graduate School. Department of Energy and Mineral Engineering

Transcription

1 The Pennsylvania State University The Graduate School Department of Energy and Mineral Engineering REDUCING THE IMPACT OF THE POWER SECTOR ON OZONE POLLUTION: AN EVALUATION OF SPATIAL AND TEMPORAL DIFFERENTIATED PRICES FOR NITROGEN OXIDE EMISSIONS A Thesis in Energy and Mineral Engineering by Zachary O Cain Stines 2016 Zachary O. Stines Submitted in Partial Fulfillment of the Requirements for the Degree of Master of Science May 2016

2 The thesis of Zachary Stines was reviewed and approved* by the following: Mort Webster Associate Professor of Energy Engineering Thesis Advisor Seth Blumsack Associate Professor of Energy Policy and Economics Chiara Lo Prete Assistant Professor of Energy Economics Luis F. Ayala H. Professor of Petroleum and Natural Gas Engineering Associate Department Head for Graduate Education *Signatures are on file in the Graduate School

3 iii ABSTRACT Nitrogen oxide (NOx) is a common air pollutant that has impacts on human health and is a precursor to the formation of tropospheric (ground-level) ozone. The U.S. Environmental Protection Agency therefore regulates the emissions of nitrogen oxides through the Clean Air Act and the National Ambient Air Quality Standards (NAAQS). Approximately 14% of all NOx emissions are produced by the electric power sector and as a result, regulations are often applied directly to the sector. However despite the success of regulations at reducing the amount of NOx emissions over the last several decades, areas not meeting the ozone NAAQS still persist. Further reductions of emissions using current approaches result in rapidly increasing marginal costs. As a product, new research is needed to design and evaluate the effectiveness of alternative regulatory designs in reducing ground-level ozone. This thesis seeks to evaluate and compare three regulatory designs: undifferentiated pricing, time differentiated pricing, and time and space differentiated pricing. Undifferentiated pricing is used to represent the current Cross-state Air Pollution Rule with constant emissions prices. Time differentiated pricing sets a higher emission price on days designated as having increased ozone formation. Time and space differentiated pricing would operate similarly to time differentiated pricing, except that different emission prices are applied to different regions. A unit commitment model is used to simulate the different regulatory designs for a study region based on the Electric Reliability Council of Texas (ERCOT) and to evaluate the cost and emissions impacts on the system. The short-term impacts result from the redispatching of resources to lower emitting generators and to generators not located within a region marked by higher permit prices. The results show that of the scenarios analyzed, time and space differentiated pricing is cost effective at reducing NOx emission prices in the nonattainment region on high ozone days, while time differentiation was the most cost-effective method at reducing system-wide emissions on high ozone days. The study also demonstrates the importance of the relative price differential between the emission prices of regions when utilizing time and space differentiated pricing. In particular, smaller differentials result in the greatest reduction in both the nonattainment and system-wide emissions. Very large differentials induce a shift of lower emitting gas generation in nonattainment regions to higher emitting coal generation in attainment regions, which increases the net NOx emissions for the system as a whole.

4 iv TABLE OF CONTENTS LIST OF FIGURES... VIII LIST OF TABLES... X ACKNOWLEDGEMENTS... XI CHAPTER 1. INTRODUCTION Nitrogen Oxide Emissions Characteristics Production and Atmospheric Reaction Retention Time and Transport Sources Impacts Primary Tropospheric Ozone Particulate Matter Regulations National Ambient Air Quality Standards Historic Current Electricity Markets Overview of Electricity Markets Summary... 9 CHAPTER 2. LITERATURE REVIEW Variable Impacts of Ozone Methods of Emission Control Controlling Heterogeneous Pollutant Implementation Temporal Studies Spatial Studies Summary... 14

5 v CHAPTER 3. METHODOLOGY Analyzed Regulatory Designs Undifferentiated Price Time Differentiated Price Time and Space Differentiated Price Two-Phase Model for Simulation Unit Commitment Model Formulation Unit Commitment Overview Treatment of Transmission Network Non-served Energy Renewable Generation Detailed Unit Commitment Model Formulation Phase 1: Determining Control Technology Installations Phase 2: Determining Summer-wide Emissions and Costs Study System Electric Reliability Council of Texas (ERCOT) Defining Space and Time Time: Air Quality Model Output Spatial: Attainment Areas Model Implementation Data Sources CHAPTER 4. RESULTS General Regulatory Designs Overview of Regulatory Designs Results from Regulatory Design Models Generation and Non-served Energy Summer-wide Emissions High Ozone Day Emissions Cost of the System Time and Space Differentiated Pricing Effect on Generation and Non-served Energy Effect on Non-served Energy Generation for Whole Summer Generation on High Ozone Days Installation of Control Technology... 49

6 vi Effect on Nitrogen Oxide Emissions Emissions for Whole Summer Emissions on High Ozone Days Effect on Costs Effect on Total System Cost Effect on Marginal Cost of Electricity Summary Comparison to Alternative Regulatory Designs Undifferentiated Pricing: Base CSAPR Generation and Non-served Energy Emissions for Whole Summer Emissions on High Ozone Days Total System Cost Marginal Cost of Electricity Time Differentiated Pricing vs. Time and Space Differentiated Pricing Generation and Non-served Energy Emissions for Whole Summer Emissions on High Ozone Days Total System Cost Marginal Cost of Electricity Cost Effectiveness Summary CHAPTER 5. DISCUSSION AND CONCLUSIONS Conclusions Implementation of Spatial Element Legality Time and Spatial Aspects of Ozone Forecasting Models Limits of Work Ozone Concentrations and Health Impacts from Air Quality Improvement to Power System Model Uncertainty Future Work Alternative Study System Spatial Definition Differentiation of High Ozone Days... 86

7 vii REFERENCES APPENDIX A.1 Unit Commitment Generator Constraint Parameter Values A.2 Dates of High Ozone Days... 95

8 viii LIST OF FIGURES Figure 1: Ozone concentrations (lines) as a function of NOx and hydrocarbon (VOC) emissions. The thick line separates two ozone-forming regimes: above is NOx-limited and below is VOC-limited... 4 Figure 2: Areas in nonattainment for the hour NAAQS for ground-level ozone... 7 Figure 3: Profile of total installed summer generation capacity in Figure 4: Profile of total energy use from different types of generation in Figure 5: Distribution of the modeled generators heat rates based on the fuel type Figure 6: Distribution of the modeled generators NOx emission rates based on the fuel type Figure 7: Map of the ERCOT system with nonattainment regions marked in blue and generators marked with red dots Figure 8: Total emissions on high ozone days under CSAPR, time differentiated pricing, and time and space differentiated pricing the nonattainment region, attainment region, and systemwide Figure 9: Total emissions on normal summer days under CSAPR, time differentiated pricing, and time and space differentiated pricing the nonattainment region, attainment region, and systemwide Figure 10: Percentage of the total cumulative summer-wide generation that utilized coal (left axis) and natural gas (right axis) as a fuel source Figure 11: Total cumulative summer-wide generation from coal (MWh) in the attainment region (left axis) and the nonattainment region (right axis) Figure 12: Total cumulative summer-wide generation from natural gas (MWh) in the attainment region (left axis) and the nonattainment region (right axis) Figure 13: Percentage of the total cumulative generation on all normal summer days and high ozone days that utilized coal as a fuel source Figure 14: Percentage of the total cumulative generation on all normal summer days and high ozone days that utilized natural gas as a fuel source Figure 15: Total generation from coal (MWh) on all high ozone days in the attainment region (left axis) and the nonattainment region (right axis) Figure 16: Total generation from natural gas (MWh) on all high ozone days in the attainment region and the nonattainment region Figure 17: Total cumulative NOx emissions (tons) over the course of the whole summer based on the nonattainment NOx emission price Figure 18: Total cumulative summer-wide NOx emissions for the attainment region Figure 19: Total cumulative summer-wide NOx emissions from the nonattainment region Figure 20: Total cumulative NOx emissions produced on normal summer days (left axis) and on high ozone days (right axis) for the different nonattainment NOx emission prices Figure 21: The total emissions on high ozone days in the attainment region (left axis) and the nonattainment region (right axis) Figure 22: Total high ozone day NOx emissions in the attainment region from natural gas and coal generation. The percentage is of the total emissions produced in the attainment region Figure 23: Total high ozone day NOx emissions in the nonattainment region from natural gas and coal generation. The percentage is of the total emissions produced in the nonattainment region.60

9 Figure 24: Total energy cost (million $) on high ozone days as the price in nonattainment NOx emission changes Figure 25: Average marginal cost of electricity in each congestion zone Figure 26: Summary of the results for time and space differentiated pricing as the differential between the attainment and nonattainment emissions price changes Figure 27: Total summer-wide generation in the both the attainment region (left axis) and the nonattainment region (right axis) for the base CSAPR scenario, time differentiated scenario, and the time/space differentiated scenarios Figure 28: Total summer-wide generation from natural gas (left axis) and coal (right axis) in the CSAPR scenario, time differentiated pricing scenario, and the time/space differentiated pricing scenario Figure 29: The total natural gas generation produced in the attainment region (left axis) and the nonattainment region (right axis) for all scenarios Figure 30: The total coal generation produced in the attainment region (left axis) and the nonattainment region (right axis) for all scenarios Figure 31: Total summer-wide NOx emissions (left axis) and high ozone day NOx emissions (right axis) for all regulatory designs and pricing scenarios Figure 32: Total summer-wide emissions in the attainment region (left axis) and nonattainment region (right axis) for all scenarios and regulatory designs Figure 33: The total NOx emissions from the attainment region (left axis) and nonattainment attainment region (right axis) on all high ozone days Figure 34: The average marginal cost of electricity ($/MWh) of the whole system for all scenarios Figure 35: Total reduction in high ozone day NOx emissions in the nonattainment region from the CSAPR scenario and the total increase in cost of the system Figure 36: Total reduction in high ozone day NOx emissions from the CSAPR scenario and the total increase in the cost of the system ix

10 x LIST OF TABLES Table 1: Proportion of total 2013 NOx emissions from the three largest emitting sectors... 2 Table 2: National Ambient Air Quality Standards for NO2 and Ozone... 6 Table 3: Price of NOx emissions on normal summer days and in each region for the CSAPR, time differentiated, and all the time and space differentiated cases modeled Table 4: Terms for the unit commitment model for the ERCOT system with NSE Table 5: Counties defined as nonattainment Table 6: Price of NOx emissions on normal summer days and in each region for the CSAPR, time differentiated, and time and space differentiated cases Table 7: Total summer-wide generation in each region and NSE for the CSAPR, time differentiated, and time and space differentiated scenarios Table 8: Total summer-wide generation from coal and natural gas system-wide and in each respective region Table 9: Total emissions and cost for the whole summer under CSAPR, time differentiated pricing, and time and space differentiated pricing Table 10: Total cost, including the startup cost and operating cost, for the system based on the implemented regulatory design Table 11: Total non-served energy and the relative percentage of total generation in each scenario Table 12: Total cumulative generation from each region over the whole summer Table 13: Percentage of total cumulative generation from each region over the whole summer. 40 Table 14: Percentage of total cumulative summer-wide generation by fuel type Table 15: Total cumulative generation from each region on all high ozone days and normal summer days for each scenario Table 16: Average emission rate (ton/mwh) in each region over the course of the whole summer Table 17: Percentage of total system-wide summer NOx emissions based on the region and fuel type Table 18: Percentage of the total regional summer NOx emissions based on the region and fuel type Table 19: Average emissions rate (ton/mwh) from each region and fuel type over the course of the whole summer Table 20: Average emission rate (ton/mwh) of natural gas and coal generation in both the nonattainment region and the attainment region Table 21: Total system cost (million $) on all high ozone days, normal summer days, and over the course of the whole summer for the different pricing scenarios Table 22: Total cost of energy (million $) on all high ozone days, normal summer days, and over the course of the whole summer for the different pricing scenarios Table 23: Total startup cost (million $) on all high ozone days, normal summer days, and over the course of the whole summer for the different pricing scenarios Table 24: Total summer-wide energy, startup, and total cost associate with each regulatory design and pricing scenario

11 xi ACKNOWLEDGEMENTS I want to first thank my advisor, Dr. Mort Webster, for not only guiding me through this process, but also introducing myself to the complex issues surrounding power systems. While I had some understanding of power systems prior to coming to Penn State, through the guidance of Dr. Webster, I am now confident in my ability to understand and utilize the techniques needed to help solve the future issues facing the electric power sector. I am grateful to know that I will always have the resource of Dr. Webster to help direct me towards the research and individuals who are challenging these problems. Secondly, I would like to thank all my excellent professors here at Penn State who helped further my knowledge and expertise in optimization, economics, market structures and regulation. Prior to Penn State, I had not had any formal training in economics and I greatly appreciate the professors within the Agricultural, Environmental, and Regional Economics Department guiding me along the way. I am also grateful for the financial support from the U.S. Environmental Protection Agency. This research and my education has been funded through the Science to Achieve Results (STAR) Program, Grant #RD Last and far from not least, I have to thank all of those who helped prepare me and supported me while at Penn State. I am thankful for the professors at the University of Georgia who helped prepare me with a strong foundation and in particular, Dr. David Gattie, who helped spark a passion for energy systems as an undergraduate. Also my parents and family who supported me with their love and wonderful conversations on the long walks home from the office. Most importantly I have to thank my fiancée, Linda Cunningham, who has put up with a lot, yet has continued to provide so much love, support, and trust throughout these last two years. I could have not succeeded without her encouragement and support each day.

12 1 CHAPTER 1. INTRODUCTION 1.1 Nitrogen Oxide Emissions Nitrogen oxides (NOx) are a group of gasses emitted from many different sources and are identified by the U.S. Environmental Protection Agency (EPA) as one of six criteria air pollutant due their adverse impacts on human health and contributions to the formation of ground-level ozone and fine particle pollution (U.S. Environmental Protection Agency, 2016a). This study focuses on analyzing alternative regulatory designs used to reduce the overall emissions of nitrogen oxides from the electric power sector. This chapter will seek to first introduce the characteristics, sources, impacts, and current regulations regarding NOx emissions followed by a brief introduction to electric power systems. The goal is that by the end of the introduction, the reader will be familiar with the background necessary to understand the challenges facing NOx emissions regulation in power systems Characteristics Production and Atmospheric Reaction Nitrogen oxides is a group of seven reactive gases that include nitrogen dioxide (NO2) and nitric oxide (NO) (U.S. Environmental Protection Agency, 1999). Nitrogen oxides are produced from both biogenic and anthropogenic sources. With regards to anthropogenic emissions, the two most important forms of nitrogen oxides are nitrogen dioxide (NO2) and nitric oxide (NO). Generally most nitrogen oxide emissions are generated as NO. In particular, NO is most commonly produced through combustion and is a product of the three Zeldovich equations: N 2 + O NO + N N + O 2 NO + O N + OH NO + H These reactions occur at high temperatures due to the dissociation of oxygen and nitrogen in combustion air (U.S. Environmental Protection Agency, 1999). Nitrogen oxide emissions from combustion can be classified into three categories: thermal, fuel, and prompt. Thermal nitrogen oxides are controlled by the temperature of the combustion. Fuel NOx emissions are a result of the nitrogen within fuels oxidizing during the combustion process. Fuel NOx emissions are more common with nitrogen rich fuels such as coal. Prompt NOx emissions are a product of fuel combining with the nitrogen in the air. The nitrogen from the air is then oxidized in the combustion process with the fuel (U.S. Environmental Protection Agency, 1999). As previously mentioned, nitrogen oxides are highly reactive gases. As a result, the concentrations of different nitrogen oxides within the atmosphere are constantly changing as a result of different atmospheric reactions. These atmospheric reactions play a large role in the secondary impacts associated with NOx emissions. Specifically, the most important reactions are with regards to the creation of tropospheric ozone. Ozone is created through the reaction between molecular and atomic oxygen. O + O 2 + M O 3 + M

13 2 Where M in the equation represents another compound that stabilizes the ozone and helps reduce the energy required for the reaction. At lower altitudes the only significant source of atomic oxygen results from the dissociation of NO2 in the presence of sunlight. NO 2 + hv NO + O The nitric oxide produced from this reaction quickly reacts with ozone, to reproduce NO2. NO + O 3 NO 2 + O 2 These three equations effectively recycle one another and the combined product results in a steady state concentration of each compound. These concentrations are dependent upon the different initial reactants, products, and rate constants of the different equations (Seinfeld, 1989) Retention Time and Transport The two main types of NOx emissions, NO and NO2, both have relatively short lives within the atmosphere. Typically NO and NO2 are known to persist at ground-level for approximately one to two days (Dentener and Crutzen, 1993; Fowler et al., 2008) and can persist for up to 20 days at much higher altitudes in the troposphere (Stevenson et al., 2006). Even a short retention time of a day still allows for the transport of NOx emissions to other areas depending on the meteorological conditions, specifically wind speed. This allows for emissions produced in one area to have an impact of others. These impacts are typically observed within a few days (Bharvirkar et al., 2004) Sources As previously discussed, the greatest source of NOx emissions is from anthropogenic sources. The largest share of anthropogenic emissions results from the combustion of fuel in three main sources: electricity generation, industrial sector, and transportation sector. The relative proportions of these sources are presented in Table 1. Table 1: Proportion of total 2013 NOx emissions from the three largest emitting sectors (U.S. Environmental Protection Agency, 2015b) Source Category Percent Annual Emissions (%) Electricity Fuel Consumption 14 Industrial Fuel Consumption 10 Vehicles (Highway and offhighway) 58 While the power sector only contributes 14 percent of the total nitrogen oxide emissions, it is typically identified as being a sector for regulation since it is the largest single point source. Within the power sector, coal is the largest emitter of NOx emissions and also has the greatest rate of NOx emissions as a result of being a very nitrogen rich fuel. The EPA has estimated that coal accounts for more than 80% of all NOx emissions generated from the electricity generation

14 3 sector (2015b), yet only accounts for approximately 40% of all electricity generation (U.S. Energy Information Administration, 2014). For these reasons, generators utilizing coal are typically a target for regulation Impacts Nitrogen oxide emissions have been shown to have both primary and secondary impacts. Primary impacts are a result of interactions directly with NOx emissions. The secondary impacts result from compounds being produced from atmospheric reactions with NOx Primary NOx emissions are associated with harmful impacts on the human respiratory system and typically have an increased effect upon vulnerable populations (i.e. elderly and individuals with prior respiratory issues). Studies have shown a relationship between respiratory inflammation and other adverse health impact with prolonged exposure to NOx emissions. NO in particular has been shown to have similar impacts on the human body as carbon monoxide, but typically only affects infants and very sensitive populations (U.S. Environmental Protection Agency, 1999). These primary impacts generally only affect sensitive populations and do not arise at the levels emitted within the United States Tropospheric Ozone While the primary impacts of NOx emissions are important, they are typically not considered to be the greatest concern. On the contrary, the secondary impacts are of great concern and have resulted in damages throughout the U.S. The secondary impacts of NOx emissions result from the formation of ozone and the creation of fine particulate matter. Both ozone and particulate matter are known to have harmful impacts on human health and form urban smog. As a result, both are regulated by the EPA as criteria air pollutants. Tropospheric ozone, also known as ground-level ozone, have been shown to produce a number of negative impacts on human health. Studies reveal that exposure to ground-level ozone results in inflammation of the airways, loss in lung function, and respiratory symptoms (e.g. coughing, throat irritation, wheezing, etc.). Consistent exposure has been associated with increased hospital visits, asthma attacks, and mortality (Berman et al., 2012; Fann et al., 2012; Hubbell et al., 2005; U.S Environmental Protection Agency 2016b). Research has not been able to establish a threshold at which ozone does not have an adverse impact on human health (Correia et al., 2013; Hubbell et al., 2005; U.S. Environmental Protection Agency, 2006). Ozone is typically not emitted as in most pollutants, but is rather a product of atmospheric reactions. NOx emissions act as reactants in these reactions and are critical precursors. As previously discussed, ozone is formed and consumed in the reactions with NO and NO2. In the absence of other compounds, these three reactions result in a steady state concentration of ozone. When volatile organic compounds (VOC) are included in the reactions, new concentrations emerge. In particular, a new reaction occurs that converts NO to NO2 without

15 4 the consumption of O3. Thus in the present of VOC the following three equations occur and allow for the accumulation of ozone. RO 2 + NO NO 2 + RO NO 2 + hv NO + O O + O 2 + M O 3 + M From these equations, it can be observed that the production and accumulation of O3 is a product of both the concentrations of VOC and NOx emissions (Seinfeld, 1989). The complexity of these reactions within the atmosphere result in a nonlinear relationship between the concentration of the reactants and the production of ozone. The ozone concentration as a function of VOC and NOx concentration is presented in Figure 1. Figure 1: Ozone concentrations (lines) as a function of NOx and hydrocarbon (VOC) emissions. The thick line separates two ozone-forming regimes: above is NOx-limited and below is VOC-limited (Jacob, 1999). The figure reveals two regimes that can exist within the troposphere. A NOx limited regime and a VOC-limited regime. In a NOx limited atmosphere, a decrease in NOx emissions results in a decrease of O3. In a VOC limited atmosphere, a reduction in NOx emissions actually results in an increase in the O3 concentration (Jacobs, 1999). This complex relationship between ozone production and NOx emissions increases the difficulty in reducing ground-level ozone

16 5 levels. Typically rural and suburban regions are NOx limited regions and urban environments are VOC limited (U.S. Environmental Protection Agency, 2006) Particulate Matter Particulate matter (PM) is both generated through emissions, but is also a secondary impact of NOx emissions. NOx emissions react with different compounds (e.g. moisture and ammonia) to form small particles. These particles are known as particulate matter and result in harmful impacts on human health when inhaled. Due to its very small size, PM has the ability to penetrate deep into the respiratory system and exacerbate respiratory ailments (e.g. bronchitis, asthma, and emphysema) along with cardiovascular issues. As a result of these impacts on human health, PM is also regulated as a criteria air pollutant. By reducing NOx emissions, PM emissions can also be reduced Regulations National Ambient Air Quality Standards In 1970, the Clean Air Act was passed into law requiring the EPA to set National Ambient Air Quality Standards (NAAQS) for six common air pollutants, nitrogen oxides, ozone, and particulate matter being three of them. The Clean Air Act requires that a primary NAAQS is set with regards to public health, particularly children, elderly, and sensitive populations (i.e. those having preexisting health conditions). Secondary NAAQS are also set to protect public welfare. Public welfare includes damages to buildings, crops, visibility, animals, and vegetation. Areas are defined as either meeting or exceeding the NAAQS. If an area meets the NAAQS, then it is considered to be in attainment and if it exceed the NAAQS then it is considered to be in nonattainment. When an area does not meet all necessary NAAQS and is considered in nonattainment, the state is required to submit a state implementation plan (SIP) that details how the state will monitor, improve, and maintain levels that meet NAAQS. If the state does not submit an implementation plan, the Federal government is required to implement a federal implementation plan (FIP). Since NOx is a larger class of compounds, the EPA has resorted to only applying a NAAQS to NO2 and using it as a proxy for all NOx emissions. This is largely based on the fact that NO is rapidly converted to NO2 in the atmosphere and NO2 being the main precursor to ozone formation (U.S. Environmental Protection Agency, 1999). The standards for NO2 were first set in 1971 and have since been reviewed twice by the EPA. In both cases, it was determined there was no scientific evidence to support a revision of the current standard. In 2010, the EPA did add another primary standard for the concentration of NO2. This standard was based upon the average concentration over a given hour (U.S. Environmental Protection Agency 2016c). Ozone is also considered a criteria air pollutants and therefore is also regulated under the Clean Air Act, but unlike other emissions, ozone is not directly emitted. As previously mentioned, ozone is formed within the atmosphere through reactions between VOCs and NOx

17 6 emissions in the presence of sunlight. The formation of ozone is greatest on hot, sunny days and can vary quite rapidly over a short number of days as a result of different meteorological conditions (Fowler et al., 2008; Mauzerall et al., 2005). Ozone concentrations even fluctuate over the course of the day as the concentrations increase in the presence of sunlight and are depleted over the course of the night (Bloomer, Vinnikov, & Dickerson 2010; Fowler et al 2008; U.S. Environmental Protection Agency, 1999; World Health Organization 2000). The Clean Air Act requires that the NAAQS are reviewed every five years to evaluate if the standard is supported by current scientific research. The primary standard for ozone was recently reduced in 2015 to reflect new scientific research that suggested the previous standard was not adequately protecting human health. The current standards for both NO2 and ozone are presented in Table 2. Table 2: National Ambient Air Quality Standards for NO2 and Ozone (U.S. Environmental Protection Agency, 2016a). Criteria Pollutant Nitrogen Dioxide Ozone Standard Type Averaging Time Level (parts per billion) Primary 1 hour 100 Primary and 1 year 53 Secondary Primary and 8 hour 70 Secondary Currently, there are no areas within the United States that are in exceedance of the primary or secondary standards for NO2 (U.S. Environmental Protection Agency, 2016a). The concentrations of NO2 have decreased by 40% since 1980 and current annual concentrations are in the range of parts per billion (ppb). As a result of no areas being in nonattainment for NO2 NAAQS, no SIPs or FIPs are required to be implemented. On the contrary to NO2, 224 counties within the United States observe concentrations greater than the NAAQS for ozone and have remained in nonattainment despite increased regulatory efforts (U.S. Environmental Protection Agency, 2015a). A map of areas in nonattainment for ground-level ozone is presented in Figure 2.

18 7 Figure 2: Areas in nonattainment for the hour NAAQS for ground-level ozone (U.S. Environmental Protection Agency, 2015a). To understand if an area meets the ozone NAAQS requires for the maximum daily 8-hour average of ozone concentrations to be calculated. These values are calculated by taking the average of the ozone concentration from an air quality monitor for 8 consecutive hours beginning at midnight. These averages are performed on a rolling horizon and thus result in 24 8-hour average concentrations for each day (i.e. 0:00 to 8:00, 1:00 to 9:00, etc.). From these 24 8-hour averages, the maximum is recorded and used as the day s maximum 8-hour average concentration. The fourth highest daily maximum in a year is then averaged with the previous 2 years fourth-highest daily maximum 8-hour average concentration to result in the final value that is compared to the NAAQS. Depending on this final 3 year average of the fourth-highest daily maximum 8-hour average results in an area being defined as in attainment or nonattainment (U.S. Environmental Protection Agency, 1998). As a result of large areas being in nonattainment for ozone NAAQS, the EPA has implemented regulations that have further sought to reduce ozone levels and help states reach attainment. As previously discussed, ozone is not emitted and therefore the regulatory actions have been aimed at the primary precursor to ground level ozone, NOx emissions Historic Regulation seeking to reduce ground-level ozone through the reduction of NOx emissions began in the 1970 s with the EPA creating the Clean Air Act. The original regulations focused

19 8 on command and control policy. These regulations set levels at which generators were allowed to emit and mandated different control methods be adopted. In 1977, New Source Performance Standards (NSPS) were established for generators and required the lowest achievable emissions reduction technology to be installed on all new generation (Burtraw and Evans 2003; Swift 2001). While these regulations reduced NOx emissions, they were not necessarily the most costeffective. Following the amendments to the Clean Air Act in 1990, policy began moving towards market-based strategies with the implementation of the Acid Rain Program in 1995 (U.S. Environmental Protection Agency, 2016d). The first market-based regulation implemented with the goal of reducing ground-level ozone was in 1999 with the Ozone Transport Commission (OTC) NOx Budget Program. This program created the foundation for market-based NOx regulation. In particular, this program was founded on the principal that ozone was transported between states and thus required regulations that could assist multiple states in achieving attainment. The program was applied to states within the Northeast to assist in implementing an emissions budget program for summer NOx emissions. It set a budget to less than half of the emissions level in 1990, but allowed the states to be responsible in creating the specific regulations, allocation of allowances, and implementation (U.S. Environmental Protection Agency, 2016c). The OTC NOx Budget Program was eventually replaced in 2003, following the NOx SIP Call in The NOx SIP Call requested that 20 states and the District of Columbia submit SIPs that addressed the transportation of ozone between states. The NOx Budget Trading Program (NBP) was implemented from 2003 to 2008 and was a cap and trade program aimed at reducing summer-wide NOx in the areas covered by the NOx SIP Call. As a result of the program being so effective in reducing summer NOx emissions, it was eventually expanded and replaced in 2009 by the Clean Air Interstate Rule (CAIR) (U.S. Environmental Protection Agency, 2016e). CAIR built upon the successes of NBP by expanding to include 27 states and the District of Columbia in the goal of addressing interstate transport of NOx and SO2. The CAIR program also created three separate cap and trade programs: annual SO2, annual NOx, and summer NOx. The program was once again successful in reducing annual NOx emissions and resulting in an emissions reduction of 21 percent below the 2013 CAIR regional NOx budget. Even though CAIR was implemented, a 2008 court decision required the EPA to create a new program to replace the CAIR rule. In response to the 2008 court order, the EPA designed the current policy of the Cross-state Air Pollution Rule (CSAPR) Current The Cross-State Air Pollution Rule (CSAPR) was first submitted in 2011 as a replacement to the CAIR program with the goal of meeting the 1997 NAAQS for PM and ozone. CSAPR was initially stalled in court appeals as a result of many states having concerns about the NOx summer budgets and was only recently approved after alterations supplied by the EPA. The alterations reevaluated the NOx summer budgets for individual states and also updated the rule to reflect the new 2008 NAAQS for ozone. Following the court s approval, the timeline of the original ruling was adjusted by 3 years. Phase 1 of CSAPR was officially implemented in

20 9 January, 2015 and Phase 2 is scheduled in begin in 2017 (U.S. Environmental Protection Agency, 2016f). As in previous programs that have sought to reduce regional transport of emissions, CSAPR is based upon the good neighbor provision of the Clean Air Act. This provision requires states to not emit emissions that will contribute significantly to nonattainment in, or interfere with maintenance by, any other State with respect to any such national primary or secondary ambient air quality standard (42 U.S.C. 7410(a)(2)(D)(I)). The rule requires the states to submit SIPs that address the good neighbor provision otherwise have the EPA implement a FIP. CSAPR is particularly aimed at reducing emissions from the electric power sector since the EPA has identified the sector as a having a substantial amount of NOx emissions that can be cost-effectively reduced through current technology (U.S. Environmental Protection Agency, 2015b). 1.2 Electricity Markets Since this study seeks to evaluate alternative regulatory designs for NOx emissions within the power sector, it is important to understand the basis for electricity markets and the complexity of the electric power system Overview of Electricity Markets Traditionally electrical systems were operated using a regulated monopoly system, where one single company acted as the electrical provider. This company generated, transmitted, and distributed all the electricity within the system. The rates at which the companies provided electricity to consumers was regulated by a Public Utility Commission. In recent years, certain regions within the United States have made alterations to this traditional model of operation and have created wholesale electricity markets. This is known as deregulation or liberalization, and has created competitive markets for electricity generation. The system and market is then operated by an Independent System Operator (ISO) or a Regional Transmission Operator (RTO). These entities ensure that the market functions efficiently and that the system is operated reliably. The United States currently has seven different regions that have deregulated and created wholesale electricity markets. The systems are the following: PJM, New York ISO, ISO New England, Southwest Power Pool (SPP), Midcontinent Independent System Operator (MISO), California ISO, and Electric Reliability Council of Texas (ERCOT). This research was performed using the Electric Reliability Council of Texas as the study space Summary NOx emissions are defined as a criteria air pollutant by the EPA due to its associated primary and secondary impacts, in particular its role as the main precursor for the formation of ground-level ozone. Therefore NOx emissions are regulated through direct NAAQS and also programs aimed at reducing NOx emissions in an effort to meet the NAAQS for ground level ozone. The regulatory designs used for reducing NOx emissions have altered throughout the

21 years. In particularly moving from traditional command and control regulations to more marketbased policies. In recent years the regulations have continued to use cap and trade policies to reduce annual and summer NOx emissions in order to reduce ground-level ozone. While these programs have been effective in their reduction of NOx emissions, ozone levels are still in exceedance of the NAAQS in many urban environments throughout the nation. This has led to research that seeks to evaluate how emissions regulations could be redesigned in order to achieve greater reductions in ozone concentrations. 10

22 11 CHAPTER 2. LITERATURE REVIEW 2.1 Variable Impacts of Ozone As discussed in the previous chapter, the formation of ozone is the result of highly complex and nonlinear reactions that occur simultaneously in the atmosphere. The main reactants that contribute to the formation of ozone are volatile organic compounds (VOC) and nitrogen oxides (NOx). When VOCs and NOx emissions are emitted, they non-uniformly mix with the surrounding atmosphere and result in the non-uniform formation of ozone. This results in ozone formation and concentrations that vary across space. The reactions that form ozone not only require VOC and NOx concentrations, but are also very dependent on the presence of sunlight. This results in ozone formation being generally favored on hot, sunny days (Fowler et al., 2008; Mauzerall et al., 2005). Hence why regulations designate a high ozone season over the course of the summer and set specific budgets for the summer season. Temperature and sunlight are both products of meteorological conditions that also vary across not only space, but also time. As a result of the meteorological condition s temporal and spatial variability of VOC and NOx concentrations, the formation and thus impacts of ozone vary across both time and space. Studies throughout the years have supported this claim. 2.2 Methods of Emission Control Regulations seeking to reduce emissions have evolved over the years. Traditionally regulations utilized a method of command and control. These regulations were characterized by requiring generators to implement specific control technologies or meet specific emissions rates. An example of this type of regulation is the New Source Performance Standards (NSPS). These type of standards generally disproportionately impact new generators since they are subject to the regulations while older, installed generators are not (Stavins, 1998; Keohane, Revesz and Stavins, 1998; Swift, 2001). Most research has shown that these forms of regulation are not as effective as marketbased regulatory instruments (Carlson et al., 2000; Ellerman, 2003; Schmalensee & Stavins, 2012). Specifically, markets have been shown to provide greater flexibility in meeting emissions standards by allowing the cheapest means of abatement to be prioritized (Hahn & Stavins, 1991). These market based regulations are also supported by traditional theory of controlling negative externalities (Viscusi, Vernon, & Harrington, 2005). Market-based regulations were first implemented with regards to reducing ozone concentrations in 1999 with the implementation of the Ozone Transport Commission (OTC) NOx Budget Program. Since its advent, new market-based policies have replaced it and continued to reduce NOx emissions in a more cost-effective manner.

23 Controlling Heterogeneous Pollutant While NOx emissions continue to decrease, ozone concentrations continue to perpetuate at nonattainment levels in spite of increased regulations (Martin, 2008; Mauzerall et al., 2005). As a result, research has sought to investigate how new regulatory designs could increase the effectiveness of the current market-based policies with regards to reducing ozone concentrations. Since ozone s formation, concentration, and impacts vary across both time and space, ozone can be considered a heterogeneous pollutant. Heterogeneous pollutants are characterized by having marginal damages and costs that alter across space and time (Levy et al, 2009; Martin, 2008; Mauzerall et al., 2005; Muller, 2011; Tong & Muller, 2006). The traditional economic theory supports the idea that the most efficient means of controlling a heterogeneous pollutant is to set the marginal cost of abatement equal to that of the marginal damages (Hahn & Stavins, 1991; Tietenberg, 1995). Since these marginal damages vary with regards to time and space, it requires the marginal cost of abatement to be differentiated to account for the variation in time and space (Tietenberg, 2010). When the price of the pollutant is differentiated and set equal to the marginal cost and damages at that point in time and space, the price results in a welfare dominating solution, but is only welfare dominating under perfect information and certainty (Fowlie & Muller, 2013). This theory of differentiating prices has resulted in research seeking to evaluate how differentiation could be applied to ozone regulation. 2.4 Implementation Controlling NOx emissions can be implemented through several means. It is most commonly controlled through the use of cap and trade policies or taxes. Under perfect certainty, both a quantity based (i.e. cap and trade) and price based (i.e. taxes) regulation produce the same outcome. When utilizing taxes, the tax rate should be set equal to the marginal damages and thus would adjust over time and space when differentiated. When using permits from a quantity based method to spatially differentiate, the permit ratio between locations should be set inversely proportional to that of the marginal damages (Muller & Mendelsohn, 2009). If the same quantity based system is being used for time differentiation, the number of permits required for each emissions would have to alter with time. 2.5 Temporal Studies The impacts and formation of ozone alter on a daily basis as the result of meteorological conditions. If these impacts are translated into the ozone precursor, NOx, the damages associated with an emission of NOx on a given day would vary depending on the current and near future weather conditions. This idea has lead researchers to evaluate how NOx emissions could be differentiated with regards to time in order to reduce emissions on days when emissions incur greater damages through the creation of ozone. Currently, NOx emissions are not differentiated with regards to time and analysis by Martin has shown that generators make temporal decisions

24 13 about when to use summer season permits (2008). Specifically, generators have been observed saving seasonal permits early in the summer in order to be utilized later in the summer when the weather is more extreme and results in higher electricity prices. These days also happen to correlate with days that promote the formation of ozone and therefore could be resulting in negative impacts on the overall ozone concentrations (Martin, 2008). Initial studies have performed empirical analysis and have supported the idea of differentiating NOx emissions with regards to time. Bharvirkar et al. and Sun et al. have also both proposed using time differentiated methods of regulating NOx emissions (2004; 2012). Bharvirkar et al. used a model to simulate the electric power system in the state of Maryland. The analysis revealed a significant reduction in ozone concentrations (Bharvirkar et al., 2004). Sun et al. performed a similar analysis using the regional electricity market of PJM to see if differentiating NOx prices would be supported by the system. Their analysis revealed that the system was flexible enough to support time differentiated prices. While these studied helped further the understanding of time differentiated pricing, both were limited due to a lack of realworld constraints applied to their models. These models failed to include many import aspects of power systems that could result in different final outcomes. Currently the amount of research that has incorporated these real-world constraints is extremely limited. The most complete study on differentiated emissions with regards to time was performed by Craig (2013). The study utilized a model that evaluated both the long-term and short-term effects of time differentiated pricing. Moreover, the study used a unit commitment model to calculate the short-term changes in the dispatch while also allowing for the installation of control technology. The analysis was performed using both the ERCOT and PJM Classic systems (i.e. Pennsylvania, New Jersey, Maryland, D.C., and Delaware). This research found that differentiating prices with regards to time resulted in a more cost-effective means at reducing NOx emissions on high ozone days. It also found that time differentiation was only cost-effective at reducing emissions on high ozone days and not summer-wide. Undifferentiated prices still were the most cost-effective at reducing summer-wide emissions. Therefore concluding that time differentiation should act as a supplement to current, undifferentiated regulations. 2.6 Spatial Studies The research surrounding differentiating NOx emissions with regards to space has been much less abundant than those with regards to time. Most research has not evaluated actual spatial differentiation of NOx emission prices, but rather just evaluated the spatial variability of the impacts from ozone and NOx emissions (Levy et al., 2009; Mauzerall et al., 2005; Muller, 2011; Tong & Muller, 2006). In recent years, there has been a large increase in research utilizing adjoint methods to model and estimate the marginal damages of NOx emissions from different sources. These models calculate the partial differentials of the ozone emissions function with respect to emissions from each source and are able to approximate each source s marginal damage. Again while these studies have continued to support the wide variation in the marginal

25 14 impacts of NOx emissions, they have lacked in truly evaluating the impacts of differentiating NOx prices. Some research performed has even supported a different conclusion than that of the economic theory. In particular three studies resulted in findings that spatial differentiation might not necessarily result in regulatory gains and that the cost of implementing spatially differentiated prices might outweigh the gains from alternative designs (Fowlie & Muller, 2013; Krupnick et al., 2000; Muller, 2011). 2.7 Summary From this review of previous research, it was observed that while there was a strong theoretical support for the differentiation of NOx emissions with regards to time and space, there was a lack of research that supported the theory through the use of models that adequately reflected the real-world. In particular only one study, performed by Craig (2013), utilized a realistic power system model to simulate the differentiation of emissions prices. There was also a general lack of research that evaluated either spatial or both spatial and time differentiated emissions. This thesis seeks to fill this gap in research by providing an extensive analysis of time and spatial differentiated emission prices using a complex power system model that adequately reflects reality. This research will help further inform the understanding of how differentiated regulatory designs of NOx emissions could result in more effective policy.

26 15 CHAPTER 3. METHODOLOGY 3.1 Analyzed Regulatory Designs This study evaluates the effectiveness of alternative regulatory designs for nitrogen oxide emissions, building on previous research. In particular, a prior study examined the effectiveness of differentiating the price of nitrogen oxide emissions with regards to time (Craig 2013; McDonald-Buller et al. 2015). In this study, I explore differentiating emission prices with regards to their spatial location as well as time. This section describes the range of scenarios simulated Undifferentiated Price The regulatory design used as the base case is an undifferentiated permit price. I simulate this regulatory approach by setting an equal price on all summer-wide emissions without regard for where or when the emissions occur during the summer. In particular, the base case is intended to approximate the current Cross-State Air Pollution Rule (CSAPR). CSAPR sets three emission budgets: an annual SO2 budget, an annual NOx budget, and a summer seasonal NOx budget. Since this study was only conducted over the course of a summer, the undifferentiated price is meant to be reflective of the summer seasonal NOx budget. Under CSAPR, a summer seasonal NOx budget and annual SO2 budget were both implemented with the expectation that the marginal cost of abatement for both pollutants would result in $500 per ton of emissions reduced (U.S. Environmental Protection Agency, 2011). I therefore model this regulatory design with a constant price of $500 per ton of NOx emissions and $500 per ton of SO2 emissions over the whole summer. All other regulatory designs modeled in this study are layered on top of the base CSAPR case; i.e., the price of NOx emissions never goes below $500 per ton on any day or any region Time Differentiated Price The proposed regulatory design would differentiate the current NOx permit price by the time at which the emission occurs. This approach was previously studied in Craig (2013) and McDonald-Buller et al. (2015). Although theoretically the permit prices could be differentiated continuously, this study and the previous work both differentiate permit prices daily. Specifically, I model a permit price that is increased on days designated as high ozone days, thereby creating two prices for each generator: a high ozone day price and a normal summer day price. By temporally differentiating the permit price, the intent is to target and reduce those emissions that are associated with the greatest marginal damages (i.e., emissions produced on days with higher ozone formation). In the current study, I do not focus on evaluating a wide range of time differentiated prices, and instead use a single case for comparison. The case used assumes a time differentiated price of $5,000 per ton of NOx on high ozone days and was chosen based on the results from previous research by Craig (2013). As previously mentioned, the case was layered on a CSAPR

27 16 case and so the normal summer day price for NOx and SO2 were set to $500 per ton of emissions. The model specifically implemented these prices by calculating the operating cost of each generator on every day. When the day was a high ozone day, an emissions price of $5,000 per ton of NOx was applied along with a $500 per ton of SO2. When the day was not a high ozone day, a price for both pollutants was set to $500 per ton Time and Space Differentiated Price In addition to the time differentiated price, I analyze a hypothetical regulatory design that also differentiates permit prices based on the location of the source. By differentiating the permit price spatially, generators with higher marginal damages can be targeted with higher prices. In theory, this would improve the efficiency of the regulation. All the cases modeled were layered on top of a base CSAPR case and a time differentiated price of $5,000 per ton of NOx. In these scenarios as well, the emissions prices were only differentiated daily resulting in two emissions prices: high ozone day price and a normal summer day price. As in the case of time differentiation, the prices could also be infinitely differentiated spatially using geographic coordinates. This study only differentiated the prices spatially by creating two zones: a higher priced zone and lower priced zone. As a result of differentiating prices in this manner for time and space, 3 separate prices were created: a high ozone price for the higher priced zone, a high ozone day price for the lower priced zone, and a normal summer day price that is applied to all zones. The scenarios modeled span a range of price pairs for high-ozone days: one price for the lower priced region and another price for the higher priced region. For example, one of the time and spatially differentiated cases placed a high ozone day price of $5,000 per ton of NOx emissions on the lower priced area and a high ozone day price of $55,000 per ton of NOx emissions on the higher priced area. Both areas had the same non-high ozone day price of $500 per ton of NOx and SO2. These prices were then incorporated into the calculation of operating cost for each generator on every day based on the area the generator was located in (i.e. high priced or low priced) and on which day (i.e. high ozone day or normal summer day). Table 3 details the prices in each region on both normal and high ozone days for all the scenarios analyzed.

28 17 Table 3: Price of NOx emissions on normal summer days and in each region for the CSAPR, time differentiated, and all the time and space differentiated cases modeled. Scenario Normal Summer Day Price of NOx ($ per ton) High Ozone Day Attainment Nonattainment Attainment Nonattainment CSAPR Time: $5, ,000 5,000 Time/Space: $5,000;$10, ,000 10,000 Time/Space: $5,000;$15, ,000 15,000 Time/Space: $5,000;$25, ,000 25,000 Time/Space: $5,000;$35, ,000 35,000 Time/Space: $5,000;$55, ,000 55,000 Time/Space: $5,000;$75, ,000 75,000 Time/Space: $5,000;$105, , ,000 Time/Space: $5,000;$155, , , Two-Phase Model for Simulation A two-phase model is utilized in this study to determine and analyze the total cost and emissions from a spatially and temporally differentiated pricing regime. The two-phase model was originally developed by Craig (2013). This study extends that model to implement spatial price differentiation. The two-phase model first determines which plants would install emission control technology based upon the current emission pricing regime in the first phase. The installation is based solely upon profit maximization and only evaluates the installation of Selective Catalytic Reduction (SCR) technology. In the second phase, the model calculates the generator commitments, power output, total emissions, and total cost for the entire summer. A unit commitment model is utilized in both phases and is therefore presented below in greater detail. Also the formulation of the original two-phase model and the additions made within this research will also be presented Unit Commitment Model Formulation Unit Commitment Overview A unit commitment model is a means of optimizing the behavior of generators within a given time frame of a few days to a week. In particular, it seeks to optimize the daily commitments (i.e. which units are on and which are off) and the dispatch of generators, subject to operational constraints. The model is implemented as a mixed-integer linear program that minimizes system-wide cost. The constraints that the model incorporates are the following: limitations on generators ramping ability (i.e., the rate at which a generator can increase or decrease their output in a given amount of time), minimum up and downtime (i.e., minimum

29 18 required time between shutdown and startup), minimum loads, and startup costs. Majority of these values were assumed for the different types of generators based on the prime mover. The Appendix provides a table of all the values used within the model. The model generates a commitment state (i.e. 0 for off and 1 for on) and a power output level for each generator in every hour that results in the lowest cost possible. The model also results in the price of electricity in each hour based on the marginal cost of the supply of electricity meeting the demand Treatment of Transmission Network Unit commitment models often do not incorporate detailed transmission constraints and make the assumptions that the system is sufficiently connected to avoid congestion. While this might not necessarily be reflective of some realistic conditions, it is frequently modeled this way because of computational limitations. I adopt the same approach to modeling the transmission network as Craig (2013) to ensure consistency. Therefore I assume power does not leave or enter the system boundary since ERCOT has a very weak interconnection with surrounding power systems and is considered to be an electrical island. Also transmission constraints were not accounted in their full extent within the study. Based on research from Baldick and Baughman (2003), while most of the system was well connected, significant transmission constraints did exist between four congestion zones: North, Houston, South, and West. As a result, zonal transmission constraints are incorporated into the model. This limits the export and import of electricity between the congestion zones. The definition of the congestion zones and the hourly levels of allowable imports and exports come from historical data (ERCOT, 2010) Non-served Energy Another attribute of unit commitment models is the incorporation of non-served energy (NSE). This variable is the unmet demand within any given hour (Morales-Espana et. al., 2013). While in actuality NSE is uncommon, it is frequently included in unit commitment models to avoid infeasible solutions to the mixed-integer problem. As in the transmission constraints, the treatment of non-served energy was the same as in previous research to maintain consistency. The value of NSE is set to the offer cap allowed within the system. In particular, I use the previous cap of $4,500 per megawatt-hour and was not updated to the newest offer cap of $9,000 per megawatt-hour for consistency with the previous research (Texas Administrative Code ). When the resulting NSE is non-zero, demand is not fully supplied and the marginal cost of electricity for that hour is equal to the value of NSE or the offer cap (i.e., $4,500/MWh). As previously mentioned though, this rarely occurs in the United States, but can be an artifact of unit commitment models. Previous research by Craig found that this assumption has a strong effect on the cost of electricity and can drastically alter the results of a model. As a result of this, there are two variants of the model, one with NSE and one without (i.e. supply must equal demand at every hour). In particular, the unit commitment model formulation used in the first phase does not include NSE since the decisions upon installing control technology are based upon profit

30 19 maximization. The incorporation of NSE could have a very large impact on the price of electricity and therefore alter the installation decisions based on an artifact of the model rather than real-world conditions. Therefore Craig expanded the optimality gap in the first phase unit commitment model to assist the solver in finding an optimal solution. The second phase unit commitment model does incorporate NSE and requires a smaller optimality gap in order to capture the total cost and emissions of the system (Craig, 2013) Renewable Generation Renewable generation was accounted for and considered dispatchable in the model. Historical hourly wind and solar generation and the installed capacities of wind and solar were used to calculate hourly capacity factors for the different technologies. These capacity factors were then used to adjust the maximum capacity of the different generators. Therefore the model allowed for renewable generation to be dispatched up to the level equivalent to the historical available capacity. When a generator is not dispatched to a level equivalent to this maximum capacity, it effectively acts as a means of curtailment and therefore is a better representation of real-world systems Detailed Unit Commitment Model Formulation The detailed, mathematical formulation of the unit commitment model is from Craig (2013) and is based on Morales-Espana et al. (2013). The formulation included here incorporates non-served energy and is identical to the formulation used in the second phase of the model.

31 20 Table 4: Terms for the unit commitment model for the ERCOT system with NSE (Craig, 2013). Sets t i z l hour of week, t T generator, i N ERCOT demand zone (North, Houston, South, or West), z Z transmission lines between zones (to and from North and each other zone, l L Parameters Pz,t D demand in hour t in zone z [GW] Pt R system spinning reserves in hour t, equal to 1% of system demand in hour t [GW] Fl,t MAX maximum flow over transmission line l in hour t [GW] RLi ramping limit (up and down) for unit i [GW] Pi MIN minimum power output of unit i [GW] Pi MAX maximum power output of unit i [GW] SUi FIXED start-up fixed cost for unit i [thousands $] SUi FUEL start-up fuel cost for unit i [thousands $] MDTi minimum down time for unit i [hours] MUTi minimum uptime for unit i [hours] Oi,t operating cost for unit i in hour t [thousands $/GWh] LZl IN zones that line l transmits power into {1,2,3,4} LZl OUT zones that line l transmits power out of {1,2,3,4} Scalars CNSE cost of non-served energy [thousand $/GWh] Variables nsez,t pi,t gi,t non-served energy in zone z at hour t [GW] total power output of unit i in hour t [GW] power output above minimum load of unit i in hour t [GW] ui,t binary variable indicating unit i is operating above its minimum load in hour t {0,1} wi,t binary variable indicating unit i shut down in hour t {0,1} vi,t binary variable indicating unit i starts up in hour t {0,1} line flow over transmission line l in hour t [GW] fl,t The objective function of the unit commitment model is to minimize the total cost of the system. The total cost (TC) is defined by the sum of all operating costs (pi,t*oi,t) of each generator in each hour, all startup costs (SU i FIXED + SU i FUEL ), and the cost of non-served energy (nse z,t CNSE).

32 21 TC = [p i,t O i,t + v i,t (SU i FIXED + SU i FUEL )] + nse z,t CNSE i,t The operating costs is defined as the total fuel cost (i.e. product of fuel cost and heat rate), total cost of emissions (i.e. product of emissions price and emissions rate), and the variable operations and maintenance cost. O i,t = HR i C FUEL i + C Emission ER i + VarOM i The startup costs include the fixed cost of startup plus the cost of fuel and any associated emissions costs. Startup cost were based upon the capacity of the generator and the prime mover. All fuel and emissions costs were calculated using either generic or generator-specific heat rates. Generators associated costs, emission rates, and heat rates are all the same as in Craig (2013). The model is constrained by demand in each zone (P D z,t ) equaling the generation produced (pi,t) within each zone, the power flow imports into the zone (f l,t ), the power flow exports from the zone, and any non-served energy within the zone (nse z,t ). The power flows into and out of each zone are limited by historical, hourly constraints (F l,t MAX ) (ERCOT, 2010). z,t P D z,t = p i,t + nse z,t + f l,t f l,t i l (LZ l IN =z) f l,t F l,t MAX l (LZ l OUT =z) Spinning reserves are also included in the model and assumed to be one percent of hourly demand. This is enforced by ensuring that the total capacity of all operating plants that is not being utilized is at least greater than the reserves requirement (P t R ). P t R [u i t P i MAX p i t ] i As previously mentioned, unit commitment models are implemented as mixed integer linear programs. Three binary variables are used to define the status of each generator. The generator s current status in each hour (ui,t) is indicated by either a 1, meaning on, or a 0, meaning off. Two other variables indicate when a generator is starting up (vi,t,) and shutting down (wi,t,) in each hour. u i,t = u i,t 1 + v i,t w i,t for all t 2 Each generator also has a minimum up (MUT) and down time (MDT) that is limited by using the generator s current operating status in each hour (ui,t) and its shutdown (wi,t) or startup

33 22 (vi,t) status. This is a critical constraint that is often not enforced in other more simplistic models (i.e., economic dispatch models). t 1 u i,t w i,t for all t [MDT, T] i=t MDT t u i,t v i,t for all t [MUT, T] i=t MUT A generator s power output (pi,t) is constrained by its maximum capacity (P i MAX ). This maximum capacity was not set equal to the nameplate capacity, but rather a 95% of the nameplate capacity to better represent realistic operations. p i,t P i MAX The generator s power output is represented using a variable that represented the level of output above the generator s minimum output level (g i,t ). This value is constrained to be between the minimum and maximum power output and is utilized in constraining the ramping rate of each generator. g i,t = p i,t u i,t P i MAX g i,t u i,t (P i MAX P i MIN ) RL g i,t g i,t 1 RL for all t Phase 1: Determining Control Technology Installations This section briefly outlines the first phase of the model used in this research. See Craig (2013) for a more detailed explanation of the model. The first phase of the model determines whether any Selective Catalytic Reduction (SCR) technologies are installed at any coal generators. The decisions are made on the individual generator level and is only applied to generators that have not already installed the post-combustion technology. The decisions are made solely on profit maximization. For a given scenario of emissions prices, the model calculates the decisions by comparing the profits of the generator with and without installation of the technology. If the profit is increased by SCR, then the generator installs the technology. The profits take into account the cost of the installation including variable operational cost, fixed cost, and a heat rate penalty. SCR installations are assumed to produce a 90% reduction in the generator s NOx emissions. A minimum emissions rate is set to 0.06 pounds per MMBtu (U.S. Environmental Protection Agency Office of Air and Radiation, 2010). SCR technology is also considered dispatchable (Patino-Echeverri et al., 2007).

34 23 While the best means of determining installation decisions would be to solve the unit commitment problem for each generator for a time horizon of the whole summer, this is computationally expensive. Instead, the SCR installation analysis is performed for two noncontinuous weeks in the summer. These weeks are selected to be representative of the demand and fraction of high ozone days for the whole summer. The two weeks were chosen using the method from de Sisternes and Webster (2013). The weeks were June 29 to July 5 and September 7 to 13. These are the same weeks used in Craig (2013). The model first determines the profits of every generator assuming no SCR installations at any generator, using the unit commitment model for the 2 weeks described above. Following this initial run, the model then re-solves the unit commitment problem with SCR assumed to be installed at each coal-fired generator, on one coal generator at a time. Then a decision is made for each generator depending on if the profit before or after the installation was greater. The process is then repeated allowing for generators to alter their decision by uninstalling or installing SCR depending on other generators decisions. This process continues until the decision of each generator does not alter as a result of the other generators decisions and a Nash equilibrium is reached Phase 2: Determining Summer-wide Emissions and Costs Once an equilibrium is reached, the model sets all installation decisions and begins running a unit commitment model that includes non-served energy for every week over the whole summer. After the model has completed and is finished dispatching the system for the whole summer, the total emissions and costs for the whole summer are calculated. The summer season was defined as April 21 through October 31. The total summer-wide emissions is the sum of the emissions produced from all power generation and startup. An individual generator s total emissions is calculated by using the generator s power output, heat rate, and emissions rate in the following equation: T E i, total = p i,t HR i ER i t The startup emissions are based on the generator s nameplate capacity, fuel consumed in startup, and emission rate. The following equation is used to define the startup emissions: E startup = Fuel startup [ ton MWh ] P nameplate[mw] ER fuel [ ton NO x MMBtu ] The total summer-wide cost of the system is calculated as the sum of the cost to generate the power, startup costs, and any capital cost from SCR installations. Unlike the total cost used to perform the unit commitment problems, the summer-wide total cost does not include any of the costs from emissions and is presented in the following equation:

35 24 TC summer = [p i,t O i,t + v i,t (SU FIXED i + SU FUEL i )] + nse z,t CNSE + SCR capcost i,t The total cost of the system also includes the incurred operating cost from any SCR installations and the annualized capital cost of SCR. In this research, the price of emissions was not used as an actual cost, but rather a means of representing scarcity. If time and space differentiated pricing was implemented using cap and trade policies, the permits would be allocated at no cost to the system. As previously discussed in the previous chapter, the permit trading ratios would adjust according to the spatial aspect and the permit per emission ratio would alter with regards to time. This cap and trade system would only increase cost as a result of redispatching of the system. Similarly, if a tax or emissions price was used to differentiated prices, extra cost would be passed on to the consumer. The revenues then from the tax or emission fees could be used to compensate consumers for the increase in electricity prices. Therefore in both cases the result is the same under perfect certainty and all costs calculated by the model do not include any additional cost from the price of emissions. Also by utilizing emissions prices, it allows the incorporation of emissions into the objective function rather than the problem becoming an optimization of stock allocation. 3.3 Study System Electric Reliability Council of Texas (ERCOT) This research was conducted using the Electric Reliability Council of Texas (ERCOT) as the study system. ERCOT is an independent system operator (ISO) that oversees an area that encompasses 75% of the land area within Texas, approximately 90% of the state s electric load, and 1,400 market participants that provide electricity to over 34 million customers. ERCOT was founded in 1999 when the Texas Legislature restructured the electric utility market and was assigned four primary responsibilities: ensure system reliability, open access to transmission, retail switching process for customer choice, and settling the wholesale market for electricity (ERCOT, 2016). The current peak demand is around 69.9 GW, observed in August The dominant resource for electricity production is natural gas with a total installed capacity share of 53% (ERCOT, 2016). The profile of the 2015 generation capacity is presented in Figure 3. The percentage of energy use from the different types of generation is also presented in Figure 4. The distribution of the generators modeled in this study is representative of this generation profile, with some simplifications; e.g., some individual generators were omitted from the model due to a lack of information. The generator characteristics vary in terms of heat rates and emission rates. The distributions of these values for each fuel type are presented in Figure 5 and Figure 6. ERCOT currently contains more than 46,500 circuit miles of high-voltage transmission, and does not have a large connection with surrounding transmission systems. This characteristic z,t i

36 25 of being an isolated power system, or electrical island, helps support key assumptions made with regards to the study system. Other (Hydro, Biomas s, & Solar) 1% Nuclear 6% Wind 18% Natural Gas 53% Coal 22% Figure 3: Profile of total installed summer generation capacity in 2015 (ERCOT, 2016)

37 26 Other (Hydro, Biomass, & Solar), 1% Nuclear, 11% Wind, 12% Coal, 28% Natural Gas, 48% Figure 4: Profile of total energy use from different types of generation in 2015 (ERCOT, 2016)

38 Figure 5: Distribution of the modeled generators heat rates based on the fuel type 27

39 28 Figure 6: Distribution of the modeled generators NOx emission rates based on the fuel type Time: Air Quality Model Output 3.4 Defining Space and Time In many scenarios, prices are differentiated with respect to time on a daily basis, and result in a day either being a high ozone day or a normal summer day. Currently in the model, high ozone days are being modelled based upon air quality simulation data for the year This data resulted in 29 high ozone days out of the 194 summer days. The specific dates of the high ozone days are listed in the Appendix. While historical data for other years are available, this study is intentionally consistent with Craig (2013). In both this study and the previous research, the high ozone days are defined as days that exceeded the EPA s maximum allowable concentration for ozone in air quality simulations produced by researchers at the University of Texas (McDonald-Buller et al, 2015) Spatial: Attainment Areas The EPA has identified counties for which the 8-hour ozone concentration has exceeded the NAAQS standard and therefore are considered to be in non-attainment as of October, The primary goal of the EPA with its NOx emissions regulations is to transition these counties

40 29 into attainment. By defining the spatial elements with regards to these counties, the study explores the alignment of its regulatory design with the goals of the EPA. As a result, this study only spatially differentiates prices with respect to only two zones: attainment or nonattainment. A higher, price for a high ozone day was placed on the nonattainment zone and these zones were defined at the county level. The counties in nonattainment correspond to the EPA s definition of 8-hour ozone nonattainment counties and they are listed in Table 5. All the nonattainment counties were seen as a single region with respect to emissions prices. Each generator was then defined with a binary value to represent whether the generator was located in a county in attainment (i.e., value of 0) or nonattainment (i.e., value of 1). Figure 7 shows a map of the system as it was modeled with the red points marking the location of each generator and the blue counties being the nonattainment regions. All generators located within the nonattainment counties are therefore given a greater high ozone day price. Individual cases will be referred in later sections with the following notation: $(Attainment high ozone day price); $(Non-attainment high ozone day price). Table 5: Counties defined as nonattainment Nonattainment Counties Metropolitan Area Brazoria Houston Chambers Houston Fort Bend Houston Galveston Houston Harris Houston Liberty Houston Montgomery Houston Waller Houston Collin Dallas-Fort Worth Dallas Dallas-Fort Worth Denton Dallas-Fort Worth Ellis Dallas-Fort Worth Johnson Dallas-Fort Worth Kaufman Dallas-Fort Worth Parker Dallas-Fort Worth Rockwall Dallas-Fort Worth Tarrant Dallas-Fort Worth Wise Dallas-Fort Worth

41 30 Figure 7: Map of the ERCOT system with nonattainment regions marked in blue and generators marked with red dots. 3.5 Model Implementation As in the previous research by Craig, the model was implemented using MATLAB R2014a ( ) and General Algebraic Modeling System (GAMS) Version In particular, the main model was operated in MATLAB while the unit commitment sub-problems were performed in GAMS using CPLEX Version 12 (Craig 2013; GAMS, 2013). The initial parameters of the specific policy being analyzed were input using MATLAB and all initial data is imported utilizing comma-separated files (CSV) in MATLAB. All relevant data was then imported into GAMS to perform any necessary unit commitment problems. The results were then returned to MATLAB. Assuming no control technology installation occur, GAMS is called upon 68 times as the model runs. In particular it is calls upon GAMS 40 times in Phase 1 and 28 times for Phase 2. It is initially called twice to develop a baseline for the 2-weeks used in Phase 1, followed by 2 times for each generator that is deciding upon control technology (total of 19 generators). This results in a total of 40 times for Phase 1. As previously mentioned, the unit commitment model for Phase 1 did not include non-served energy (NSE) and required the optimality gap to be