This item has the following downloads:
1 THREE ESSAYS ON THE CHANGING U.S. ELECTRICITY INDUSTRY By THEODORE J. KURY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 2013 Theodore J. Kury
3 For my children
4 ACKNOWLEDGMENTS My career trajectory has been non traditional, to put it as politely as possible, but I never dreamed that my life would encompass all of the experiences Frost was right, I have taken the road less traveled, and it has made all the difference. But I am lucky enough to have travelled, and continu e to travel, this road with other people, and while it is impossible to I want to thank my parents for all of their love and support and instilling in me the belief t hat I could accomplish anything if I set my mind to it. Sure, I never became the centerfielder for the Mets but a Ph. D. seemed impossible once, too. I want to thank Sheree Brown, John Kersten and everyone at SVBK Consulting Group for introducing a novic e economist whose resume included hotel manager, driving instructor, and ticket. I want to th ank Jim Fort, Mark Kinevan, Joanie Teofilo and everyone at The Energy Authority for affording me the opportunity to not merely do, but to ask why and look for the answer s I want to thank Mark Jamison, Sandy Berg, and everyone at PURC for believing that there was a place for me My experience s here continue to surpass any of my expectations. Many talk about helping to change the world, while the people at PURC demonstrate how to do it every day. I want to thank all of my professo rs at the University of Florida for their time, attention, and understanding of all of my e xogenous constraints. I have learned much from all of you, despite what you might have seen on my examination papers.
5 I want to thank all of my classmates for their valuable insight throughout our studies I learned something every day from all of you, and I am excited about all that you will accomplish in your careers. I also want to extend s pecial condolences to Dave Brown, who always seemed to get stuck as the discussant for my papers, but they are much better for your insight. I want to thank the members of my committee: Sandy Berg, Jon athan Hamilton, Chuck Moss, and David Sappington for all of their guidance and support I needed to refine my communi cation skills, and you have shown me how. You have changed the way I ask questions and look for an swers and helped me to become a more effective economist. Mostly, I want to thank my wonderful wife Melissa and my children Philip, Lena, Teddy, and John. Without you, nothing else would matter. The journey of life is filled with opportunities and challen ges, and I treasure every moment because of the people I get to share it with. I Love You.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 8 LIST OF FIGURE S ................................ ................................ ................................ ......................... 9 LIST OF ABBREVIATIONS ................................ ................................ ................................ ........ 10 ABSTRACT ................................ ................................ ................................ ................................ ... 11 CHAPTERS 1 A BRIEF HISTORY OF THE UNITED STATES ELECTRICITY MARKET .................... 13 2 PRICE EFFECTS OF INDEPENDENT SYSTEM OPERATORS IN THE UNITED STATES ELECTRICITY MARKET ................................ ................................ ..................... 17 Introduction ................................ ................................ ................................ ............................. 17 Existing Literature ................................ ................................ ................................ .................. 21 Data ................................ ................................ ................................ ................................ ......... 25 The Model ................................ ................................ ................................ ............................... 26 Results ................................ ................................ ................................ ................................ ..... 32 Conclusion ................................ ................................ ................................ .............................. 40 3 THE IMPACT OF THE TRANSPARENCY OF WHOLESALE MARKETS ON MARKET PARTICIPATION: THE CASE OF THE U.S. ELECTRICITY INDUSTRY .... 41 Introduction ................................ ................................ ................................ ............................. 41 Costs and Benefi ts of RTOs ................................ ................................ ................................ ... 43 Related Literature ................................ ................................ ................................ ................... 45 Data ................................ ................................ ................................ ................................ ......... 47 Model ................................ ................................ ................................ ................................ ...... 50 Results ................................ ................................ ................................ ................................ ..... 53 Conclusions ................................ ................................ ................................ ............................. 66 4 CHALLENGES IN QUANTIFYING OPTIMAL CO2 EMISSIONS POLICY: THE CASE OF ELECTRICITY GENERATION IN FLORID A ................................ ................... 67 Introduction ................................ ................................ ................................ ............................. 67 Literature Review ................................ ................................ ................................ ................... 69 Model of Economic Dispatch ................................ ................................ ................................ 71 Data Sources ................................ ................................ ................................ ........................... 73 Model Operation ................................ ................................ ................................ ..................... 74 Model Output ................................ ................................ ................................ .......................... 75
7 Conclusions ................................ ................................ ................................ ............................. 82 5 CONCLUDING REMARKS AND OPPORTUNITIES FOR FURTHER RESEARCH ...... 84 APPENDICES TEST OF ENDOGENEITY OF SALES ................................ ................................ ....................... 86 TEST OF TH E STRENGTH OF INSTUMENTAL VARIABLES ................................ .............. 88 THE MODEL OF ECONOMIC DISPATCH ................................ ................................ ................ 91 LIST OF REFERENCES ................................ ................................ ................................ ............. 127 BIOGRAPH ICAL SKETCH ................................ ................................ ................................ ....... 132
8 LIST OF TABLES Table page 2 1 Mean and standard deviation of nominal electricity price for each state group ................ 24 2 2 Mean and standard deviation for model variables ................................ ............................. 27 2 3 2SLS estimates with entire sample ................................ ................................ .................... 32 2 4 2SLS estimates excluding states that have restructured their electric industry ................. 34 2 5 2SLS est imates with entire sample and interaction terms between state and fuel price .... 36 2 6 Selected coefficients on the interaction betwe en state and fuel prices .............................. 37 2 7 2SLS estimates with restricted sample and interaction terms between state and fuel price ................................ ................................ ................................ ................................ .... 38 2 8 2SLS estimates by customer class excluding states that have restructured their electric industry ................................ ................................ ................................ .................. 39 3 1 Mean and standard deviation of purchase sample ................................ ............................. 51 3 2 Mean and standard deviation of sales sample ................................ ................................ .... 52 3 3 Parameter estimates for initial sample ................................ ................................ ............... 54 3 4 Parameter estimates for expanded sample ................................ ................................ ......... 57 3 5 Parameter estimates for expanded sample: utilities in states that restructured their electricity industry ................................ ................................ ................................ ............. 61 3 6 Parameter estimates for expanded sample: utilities in states that did not restructure their electricity industry ................................ ................................ ................................ ..... 62 3 7 Parameter estimates for expanded sample: municipal utilities ................................ .......... 64 3 8 Parameter estimates for expanded sample: investor owned utilities ................................ 64 A 1 OLS estimates of the log return of electric sales ................................ ............................... 87 B 1 First stage estimates of log return of electricity sales with entire sample ......................... 88 B 2 Partial R 2 value s for excluded instruments ................................ ................................ ........ 89 B 3 First stage estimates of log return of electricity sales excluding restructured states ......... 89 B 4 Partial R 2 values for excluded instruments ................................ ................................ ........ 90
9 LIST OF FIGURES Figure page 2 1 Regional Transmission Organizations in North America ................................ .................. 23 2 2 Comparative state electricity prices ................................ ................................ ................... 29 3 1 Effect of the IOUxISOYrs and IOUxISOYrs2 coefficients on wholesale market purchases of the initial sample ................................ ................................ ........................... 56 3 2 Effect of the IOUxISOYrs and IOUxISOYrs2 coefficients on wholesale market purchases of the expanded sample ................................ ................................ ..................... 59 3 3 Effect of the SumPkxISOYrs and SumPkxISOYrs2 coefficients on wholesale market purchases of a 1000 MW utility in the expanded sample ................................ .................. 60 4 1 Real ($2010) incremental cost of electricity by year and emissions price ........................ 76 4 2 Emissions by year and emissions pric e ................................ ................................ .............. 77 4 3 Average cost of abatement curves ................................ ................................ ..................... 78 4 4 Average and incremental cost of abatement curves for 2015 ................................ ............ 79 4 5 Multiple abatement equilibria at a carbon tax of $700 ................................ ...................... 80 4 6 Multiple abatement equilibria at a carbon tax of $70 ................................ ........................ 81 4 7 Fuel consumption in 2013 ................................ ................................ ................................ .. 82
10 LIST OF ABBREVIATION S DOE United States Department of Energy EIA Energy Information Administration FERC Federal Energy Regulatory Commission ISO Independent System Operator RTO Regional Transmission Organization
11 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy THREE ESSAYS ON THE CHANGING U.S. ELECTRICITY INDUSTRY By Theodore J. Kury August 2013 Chair: David Sappington Major: Economics In 1996, the Federal Energy Regulatory Commission (FERC) sought to transform the wholesale electricity market with a series of market rules. A product of these rules was the establishment of regional transmission organizations (RTOs) and independent system operators (ISOs) charged with facilitating equal access to the transmission grid for electricity suppliers. The e ffect of these changes in market structure remain s an open question. This dissertation attempts to quantify the impacts of this change in market strucuture in addressing important policy issues facing the electricity sector. The first essay utilizes a pane l data set of the 48 contiguous United States and a treatment effects model in first differences to determine whether there have been changes in delivered electric prices as a result of the establishment of ISOs and RTOs. This estimation shows that electri result is statistically significant. However, this result is dependent on the presence of states that restructured their electricity markets. When these restructur ed states are removed from the data set the price effects of RTOs become indistinguishable from zero. The second essay utilizes the diversity of the United States electricity market and a panel data set of electric utilities for the period 1990 2009 to st udy the effects that RTOs have had on
12 the trade of w holesale electricity. It finds that the presence of a transparent wholesale marketplace for electricity has the effect of increasing participation, but that this participation occurs asymmetrically across types of electric utilities. The third essay utilizes a model that simulates the dispatch of electric generating units in the state of Florida under various prices for CO 2 emissions, and analyzes the challenges that may arise in the determination of opti mal emissions abatement policy. It finds that the rate of abatement varies considerably with the price of CO 2 emissions. It demonstrates how the incremental cost curve of emissions abatement may intersect with a CO 2 tax at many levels of abatement, allowin
13 CHAPTER 1 A BRIEF HISTORY OF THE UNITED STATES ELECTRICITY MARKET In 1882, the Edison Illuminating Company began to provide electricity to 59 customers in lower Manhattan from its Pearl Street Generating Station, marking it as owned electric utility. Edison Illuminating generated the electricity, transmitted it to customers 1 and distributed it to their homes and businesses. Since it performed all of the functions necessar y to supply electricity to customers, it had a vertically integrated structure. The scope and scale of electric utilities grew rapidly from those humble beginnings but the underlying vertically integrated structure of the industry remained intact for more than 100 years In 1996, the Federal Energy Regulatory Commission issued Order 888, paving the way for the restructuring of t he electricity industry in the United States. Had this order been issued in a country like Brazil, where the power of the federal government is great relative to the state governments, the sector likely would have been transformed in a uniform manner acros s the country. However, the country is called the United States of America for good reason, as the by the Constitution, nor prohibited by it to the States, are r eserved to the States respectively, or to 2 As a result of this amendment it was left to the individual states to determine the extent to which the electricity industry in their respective states was restructured. Through the actions or inact ions, of individual state legislatures, t he electricity market in the United States was fractured into three distinct structures Some states, primarily in the S outheast and the W est, maintained the verti cally integrated structure. In o ther states, primari ly in the M idwest, the 1 As it was direct current, this electricity but it was transmission nonetheless Electricity t ransmission over longer dis tances is only feasible after i nversion to alternating current. 2 United States Constitution Amendment X
14 utilities maintained ownership of all assets, but ceded control of their transmission assets to third parties known as independent system operators or regional transmission operators. A final group of states primari ly in the N orthea st but including California took advantage of the federal order and forced electric utilities to divest either their generation or distribution assets, and opened their markets for electricity supply to retail competition. Since this restructuring, the electricity industry in the United States has experienced notable events such as the California power crisis of 1999 2001, the blackout of 2003 in the N ortheast and Midwest and price spikes occurring in Texas in 2005 and New York City from 2006 to 2008. While investigations into these events have focused on the behavior of particular parties, principally electricity generators, little has been done to explore the implications of the market strucuture itself. FERC is presently attempting to assess the costs and benefits of RTOs third party administrators of the electricity transmission system that arose in response to restructuring, through the collection of performance metrics Once this data is collected, regulators will have better information with which to address the ir impact, but the metrics are limited to the performance of the wholesale market. As a result, the impact on retail consumers is not addressed by these metrics, nor is the distributional effect of a more transp arent market. New challenges facing the electricity industry center on the the externalities associated with the emission of greenhouse gases during the combustion of fossil fuels. Reaction to differences between the social and private cost of these emissi ons, even as scientists on both sides of the debate continue to argue over whether such a difference exists, has not been uniform across the globe. Even though the most widespread school of thought is that CO 2 emissions do have an effect on the global clim ate, and that there is a difference between the private and social costs, the impacts of these effects varies considerably, even among the staunchest supporters of
15 climate change. 3 Despite this uncertainty, some governments have established prices for CO 2 emissions, while others have not. The failure of governments to act uniformly has been a source of consternation for participants on either side of the debate, and global summits on climate change have not led to a resolution of the question. While the Eur opean Union has established an emissions trading system, the two other 2 emissions 4 have yet to implement nationwide programs to impose a market price on emissions. The federal legislature of the United States introduced the Waxman Markey Bill in the House of Representatives and the Kerry Boxer Bill in the Senate in 2009. Both bills proposed a reduction in CO 2 emissions to 17% of 2005 levels by 2050. The Waxman Markey bill passed in the House, but the Kerry Boxer bill languis hed. In July 2010, Senate Majority Leader Harry Reid told the New York Times, in announcing that the Senate 5 The Chinese government has shown support for volu ntary emissions reduction programs and stated national emissions reduction goals, but has not supported emissions prices. This apparent inability for nations to agree on a market treatment for CO 2 emissions could be the result of a disagreement over the ma gnitude of the externality, or the costs and benefits associated with mitigating it, but there may be complications in the behavior of the marginal costs and benefits as well. 3 statement changes in precipitation and other climate variables in addition to temperature, sea level and concentrations of atmospheric carbon dioxide. The magnitude and timing of impacts will vary with the amount and timing of climate 4 2 emissions 5 New York Times, July 22, 2010
16 The provision of electricity is critical to life in the United States, and a bet ter understanding of the effects and challenges of changes in the sector can lead to improvements in consumer and producer welfare.
17 CHAPTER 2 PRICE EFFECTS OF IND EPENDENT SYSTEM OPER ATORS IN THE UNITED STATES ELECTRICITY MARKET Introduction Before the Federal Energy Regulatory Commission (FERC) issued its landmark Order 888 in April of 1996, the electricity generation, transmission, and distribution market in the United States had functioned largely within a vertically integrated monopoly structure for over 100 years. The opening paragraph of Order 888 reads: Today the Commission issues three final, interrelated rules designed to remove impediments to competition in the wholesale bulk power marketplace and to bring more efficient, and policy cornerstone of these rules is to remedy undue discrimination in access to the monopoly owned transmission wires that control whether and to whom electricity can be transported in interstate commerce. A second critical aspect of the rules is to address recovery of the transition costs of moving from a monopoly regulated regime to one in which all sellers can compete on a fair basis and in which electricity is more competitively pr iced. 1 FERC appears to believe that the vertically integrated structure in which the generator of electricity also controls the transmission of electricity is inefficient, and that this inefficiency leads to higher prices. The issuance of this order paved the way for numerous states to introduce plans to restructure their electric markets, with varying degrees of success. This movement began most notably in California, Texas, and a number of states in the Northeast, with the separation of ration from the transmission and distribution functions. To facilitate non discriminatory access for all generators to the transmission grid, FERC conditionally approved the formation of five independent system operators (ISO) in 1997 and 1998 to oversee t he deregulated wholesale power markets. In December of 1999, FERC issued Order 2000, which stated: 1 FERC Order 888, issued April 24, 1996, Page 1 (75 FERC 61,080)
18 The Federal Energy Regulatory Commission (Commission) is amending its regulations under the Federal Power Act (FPA) to advance the formation of Regional Tran smission Organizations (RTOs). The regulations require that each public utility that owns, operates, or controls facilities for the transmission of electric energy in interstate commerce make certain filings with respect to forming and participating in an RTO. The Commission also codifies minimum characteristics and functions that a transmission entity must satisfy in order to be considered an RTO. The Commission's goal is to promote efficiency in wholesale electricity markets and to ensure that electricity consumers pay the lowest price possible for reliable service. 2 This Order suggests that FERC believed that the establishment of independent entities to control access to the electric transmission system would result in costs that are no greater than the c osts that exist at the time of the order. The focus of this paper is to identify tangible price effects as a result of the formation of RTOs and ISOs. These effects are critical to assessing the efficacy of this landmark regulatory policy. While there are structural differences 3 between the two types of organizations, their basic function of ensuring equal access for electric generators to the transmission grid and optimal dispatch of the generating system remain. Since that is the function analyzed in the paper, the terms ISO or RTO as used here are effectively indistinguishable. An RTO can impart many benefits to the market in both the short term and long term. FERC Order 2000 identified five benefits that RTOs can offer: improved efficiencies in the mana gement of the transmission grid, improved grid reliability, non discriminatory transmission practices, improved market performance, and lighter handed government regulation 4 Through the optimization of the daily and hourly decisions of system dispatch ove r a wider geographic area than the existing system, the RTO may lower the system costs required to serve electric 2 FERC Order 2000, issued December 20, 1999, Page 1 (89 FERC 61,285) 3 For example, RTOs have been tasked by the FERC to ensure the long term reliability o f the system by managing transmission investment. ISOs are nominally regulated by the Federal government, while RTOs govern themselves. 4 FERC Order 2000, issued December 20, 1999, Page 70 71 (89 FERC 61,285)
19 load. By allowing non discriminatory access to the transmission system, the RTO may also be able to incorporate lower priced resources that ma y not have enjoyed access to the market under a previous market regime, thus lowering system costs. Fabrizio, Rose, and Wolfram (2007) provide evidence that electric generators increase their operating efficiency in a market environment by reducing labor a nd nonfuel operating expenses, relative to operators in states that do not restructure their markets. An RTO may also be able to improve the reliability of the electric system by coordinating resource allocation and long term system planning. All of these benefits must be measured against the costs of operating and maintaining the RTO, and the costs incurred by market participants for compliance and regulation. However, since all costs related to the RTO are recovered through volumetric charges passed throu gh to consumers of electricity own costs by examining the rates charged to customers. A change in prices, controlling for other factors, should signal either a n et cost or net benefit associated with the RTO. FERC is presently attempting to assess the costs and benefits of RTOs. In February of 2010, FERC issued a request for comments on a series of performance metrics for ISOs and RTOs 5 This request for comment was the result of a 2008 report from the Government Accounting Office that requested that FERC work to develop metrics to track the performance of RTO operations and report this performance to the public. Once this data is collected, regulators will have b etter information with which to address the question, but the goal of this paper is to see if there is something that can be learned now, with the data available. Pricing metrics utilized by FERC include indicators of wholesale market price performance, bu t do not reflect the costs paid by retail utility customers. Any burden to the retail customer will include not only the 5 75 Fed. Reg. 7581 (2010)
20 metrics account for some of the costs to retail customers, but do not address all of them. In an effort to assess the costs of maintaining a RTO, Greenfield and Kwoka (2010) have developed an econometric model of RTO costs dependent upon the geographic scale, scope of services provided, and age o f the RTO. Such a model could be used to benchmark the relative cost effectiveness of these organizations. Kwoka, Pollitt, and Sergici (2010) have also presented evidence that forced divestiture as a result of electric restructuring has resulted in decreas es in efficiency for electric distribution systems. Because these models do not address benefits, the question of whether RTOs have provided net benefits the consumers of electricity remains open. This study employs a panel data set of the contiguous Unite d States spanning the period 1990 2008 in an attempt to determine whether the establishment of RTOs has had an effect on the prices that consumers pay for electricity. The United States electricity market is particularly attractive for studying questions r elated to market structure. For roughly 100 years, most electric utilities in the United States were vertically integrated, providing generation, transmission, and distribution of electricity. Following the issuance of FERC Order 888, industry structure ch anged. Many states restructured their electricity markets, forcing the divestiture of the generation, transmission, and distribution components of the electric utilities in their state. Utilities in other states did not restructure, but ceded control of th eir transmission assets to independent entities, the RTOs and ISOs. A third group of states retained their vertically integrated structure. This paper exploits this diversity to study the effects of changes in market structure. The analysis concludes that the price effects of RTOs, when disentangled from the effects of electric restructuring, are not statistically significant, and these general results are robust to various specifications of the model. However, when the price effects for individual
21 classes of customers are considered, there may be some slight reductions in price for residential and industrial customers. The remainder of the paper is organized as follows: Section 2 consists of a review of the existing literature, Section 3 describes the data used in the analysis, Section 4 is a description of the estimation models used, Section 5 discusses the results of the analysis, and Section 6 contains some concluding remarks. Existing Literature Coase (1937) addressed the question of why individuals orga nize into firms, observing that the degree of vertical integration varied greatly among types of industries and types of firms. Since individuals were always free to interact with the market in the absence of firms, Coase concluded that firms arise when th e costs of interacting with the market exceed the costs of interacting within an organization. So, if the regulators of a particular industry decided that the costs of interacting within an organization would exceed those of the market, they might restruct ure the firms in the industry in order to reduce transaction costs. Grossman and Hart (1986) have argued that the literature on transaction costs emphasized the conclusion that nonintegrated relationships can be inferior to relationships with complete con tracts. However, they assert that this is not due to the nature of the nonintegrated relationship itself, but because of the presence of incomplete contracts. They pointed out that this argument in the existing literature has assumed that integration leads to complete contracts, which may not be the case. They further argue that the proper comparison is that between contracts that allocate rights of ownership, residual rights, to one party and contracts that allocate them to another. They conclude that when it is too costly to specify a list of particular rights that one party desires over
22 Previous studies in the electricity area have focused on the question of whether restructuring of the ele ctricity market itself has led to changes in delivered electricity prices. Kwoka (2006) presents a review of a number of these studies. He finds that all are plagued by two underlying problems: the endogeneity issues related to the decision to restructure the electricity market, and the confounding effects of settlement agreements between the states and particular terms of these settlement agreements varied consid erably by state, but contained two common elements. The first element was some form of retail rate control, either a rate freeze that kept rates at current levels for a designated period of time, or a prescribed schedule of future rates based on current ra tes. Most often, the first year in the schedule mandated a rate decrease, and this decrease often persisted beyond the first year. The second element was a mechanism to recover the value of stranded assets, or to recover costs not recovered under the rate agreement. Restructuring in Pennsylvania, for example, was accompanied by the imposition of retail rate caps on the privately owned utilities. The expiration of the rate caps for PPL Electric Utilities in January of 2010 was accompanied by rate increases o f 30%. This dramatic increase in electric prices suggests that the realized prices in the years following the restructuring agreement did not reflect the market price for electricity in Pennsylvania. The states of Maryland and California experienced simila r price increases upon the expiration of imposed price caps, so the experiences restructuring was simply a temporal subsidy, though it is not yet clear how mu ch, as this transition cost recovery continues in many states, and the methods used to impose this subsidy were heterogeneous across states. Because temporal subsidies have been used to shift costs, the full effect of these subsidies is unknown and the eff ect of restructuring on costs is difficult to
23 determine. Therefore, any analysis utilizing electricity prices in restructured states will be tainted by those confounding effects, as well as by endogeneity issues related to the decision to restructure the e lectricity market. The present study frames the question differently to avoid those confounding effects. Rather than attempt to explain the changes in price wrought by electric restructuring, which is composed of two inter related effects 6 this paper focu ses on whether there have been changes in price as a result of the formation of ISOs or RTOs. A map of the current footprint of these organizations is shown in Figure 2 1. Figure 2 1. Regional Transmission Organizations in North America 7 6 The two effects are the effect of the change in market struct ure as well as the effect of the rate agreement used to facilitate electric restructuring. 7 From http://www.ferc.gov/industries/electric/indus act/rto.asp
24 Using a panel d ata set of the 48 contiguous United States, this paper utilizes a treatment effects model in first differences to determine whether there have been changes in delivered electric prices as a result of the establishment of RTOs. To avoid the confounding effe cts of electric restructuring, the model is initially estimated with the full panel data set as a benchmark, and then again without the 16 states that have restructured their electric markets. Of the remaining 32 states, 12 are served by one or more RTOs. Table 2 1 shows the average nominal price of electricity for each of these state groups. Table 2 1. Mean and standard deviation of nominal e lectr i city price for each state g roup Restructuring Status ISO Status N Nominal Price Restructured States Before ISO Implementation 150 8.09 1.91 After ISO Implementation 154 9.98 2.67 Non Restructured States Served by ISOs Before ISO Implementation 126 5.96 0.93 After ISO Implementation 102 6.81 1.67 Non Restructured States Not Served by ISOs 380 6.14 1.29 Table 2 1 illustrates the endogeneity issue raised by Kwoka. The states that restructured their electricity markets exhibited higher prices, on average, than the states that did not. However, among the states that did not restructure their electricity mark et, there is little difference, on average, in price level between the states that were eventually served by ISOs and those that were not. Therefore, by considering only the states that have not restructured their electric markets, this paper estimates whe ther there have been price effects due to the establishment of RTOs, in the absence of restructuring agreements.
25 Data The data used in this paper are annual data for the 48 contiguous United States, spanning the period 1990 through 2008. The data for the s tudy are primarily derived from reports and Administration (EIA). The EIA is mandated by Congress to collect survey data from electric utilities in the United States. Thes e data are collected on a variety of forms spanning electric utility operations. The EIA 860 report consists of generator specific data such as generating capacity and energy sources. The EIA 861 and EIA 826 reports contain utility specific data on sales a nd revenues by customer class. The EIA 923 report contains utility specific data on electricity generation and fuel consumption. This utility and generator specific data is ary data source for statewide generation and prices in this study. Prices used in this study are average prices across customer classes, as well as for broad customer classes, calculated by dividing revenue by the sales volume. State level data on annual h eating and cooling degree days is available from the National Climatic Data Center, which population weights the heating and cooling degree days collected from individual climate monitoring stations. Heating and cooling degree days are functions of average daily temperature often used to explain demand for electricity (Papalexopoulos and Hesterberg, 1990). They are the aggregate of the average daily temperatures either above (cooling) or below (heating) 65 degrees Fahrenheit. For example, if the average dai ly temperature is 70 degrees, then that day is said to have 5 cooling degrees 8 These degree days are then aggregated annually or monthly. Data on annual population by state 8 tion experiences 70 degree temperatures and half of the population experiences 74 degree temperatures, then the National Climatic Data Center will record 7 cooling degrees for that state, for that day.
26 is from the U.S. Census Bureau. Data on per capita income by state is from the U.S Department of Commerce, and is used as a proxy for heterogeneous economic conditions within each state. Data regarding state participation in electric restructuring activities is available from EIA 9 FERC, and the individual state regulatory agencies. Fi nally, the membership of state utilities in RTOs is available from EIA, FERC (as seen in Figure 2 1), and the individual RTOs. The Model The paper presents a model of the average electricity prices paid per kilowatthour (kWh) of consumption by the customer s in each state, and tests the treatment effect of RTOs on that price. The effects of RTOs are not limited to prices, however. The centralization of dispatch and system planning decisions may have impacts beyond electricity revenues, such as on the overal l system reliability. The effects of the RTOs on system reliability are much more difficult to assess, as most reliability data is proprietary. Further, the RTOs may be able to optimize the decisions regarding power plant investment within its region of re sponsibility, but its effects may not yet be seen. Thus, this paper studies the impact that RTOs have through the retail rates charged to customers. This is an important metric, as the portion of FERC Order 2000 cited above specifically states the Commissi on goal of lowest possible prices. The average revenue per kWh of electricity for each state i in a given year t can be expressed by the following panel equation: ( 2 1) where: 9 For example, http://www.eia.gov/cneaf/electricity/pa ge/restructuring/restructure_elect.html
27 Price Nominal state electricity revenues per kWh in cents/kWh Sales Electricity sales in MWh PCoal Nominal state price of coal in $/MMBtu PGas Nominal state price of natural gas in $/MMBtu %Hydro Percent of electric generation from hydroelectric sources %Nuc Percent of electric generation from nuclear sources RTO Whether the majority of the electric customers in the state are served by a utility that belongs to an RTO The mean and standard deviation for these variables is given for the entire sample, as well as three cohorts, in Table 2 2 Table 2 2 Mean and standard deviation for model v ariables Entire Sample Restructured States States that Did Not Restructure Electric Industry RTO States Non RTO States Price (cents/kWh) 7.16 2.25 9.05 2.51 6.34 1.38 6.14 1.29 Sales 6.75e07 6.09e07 9.28e07 8.30e07 4.16e07 2.98e07 6.28e07 4.39e07 Coal Price 1.36 0.58 1.58 0.61 1.05 0.46 1.37 0.53 Natural Gas Price 4.31 2.42 4.31 2.49 4.45 2.39 4.22 2.38 % Hydro 11.10% 20.83% 9.84% 18.89% 8.68% 15.81% 13.56% 24.44% % Nuclear 18.43% 18.53% 22.72% 18.80% 18.39% 21.36% 15.03% 15.58% N 912 304 228 380 The variable represents the fixed effects of the model, or the heterogeneous characteristics of the state that contribute to the prevailing electricity price in the state. The price
28 of electricity in a state is influenced by factors such as the types of units used to generate electricity, the price and availability of fuel, the geographic proximity to these resources, the effects of geography on the costs of electricity transmission and distribution, heterogenous ratemaking standards that might apply to that state, or the degree to which ratemaking authority is centralized 10 Because generating units are long lived assets, the composition of the generating fleet will change little over time leading to stability in the structure used to produce electricity. As a result, p rice levels might be expected to differ by state, and these differences might be expected to persist. Figure 2 2 illustrates the electricity prices in the data set for three sample ydropower resources in the region. Georgia relies primarily on coal and nuclear generation and thus experiences higher prices than Idaho. Connecticut relies on nuclear and natural gas generation, with no access to lower priced coal generation, and therefor e had the highest prices of the three states. The centralization of ratemaking authority is also a source of heterogeneity, with each state served by some combination of investor owned, municipally owned, and cooperative utilities. However, the ownership s tatus of these utilities rarely changes, so this heterogeneity will be relatively stable over the sample period. 10 State public utility commissions typically have ratemakin g authority over only investor owned utilities, while municipally owned utilities are governed by the municipalities themselves, and cooperative utilities ar e governed by the customers they serve.
29 Figure 2 2. Comparative state electricity p rices The heterogeneous effects of this variable are removed by estimating the model in firs t differences. Further, the variables Price Sales PCoal and PGas are transformed by taking logs, so that the variables in the equation, with the exception of the treatment, all represent annual percent changes. The estimation equation then becomes: (2 2 ) Lagged observations of the RTO variable are also included in the estim ated model, as the effects of the RTO may not materialize (or fully materialize) in the year of its inception. The first
30 lag will be equal to 1 if the utilities in the state became members of an RTO in the previous year, and the second lag equals 1 if two years prior. One further refinement to the model is necessary. Unless the price elasticity of electricity demand is zero, the electricity sales variable is endogenous in the price equation. While other authors have estimated the price elasticity of demand for electricity 11 that question is beyond the scope of this paper. As long as the price elasticity differs from zero, it is important for the specification of this model. Therefore, the endogeneity of the electricity sales variable is tested using the inst rumental variables heating and cooling degree days, state per capita income, and state population. Even if the price of electricity has an effect on sales, it should not have an effect on the weather, income or the population of the state, so these variabl es are exogenous. The reduced form equation for is estimated and the residuals are includ ed as explanatory variables in Equation 2 2 The coefficient on this variable is significant 12 and so Equation 2 2 is estimated using 2SLS with the instrument al variables heating and cooling degree days, state per capita income and state population for The sign of the coefficient might be positive or negative. Increased demand for electricity increases the expenditure on fuels required to pro duce electricity and may result in the utilization of higher cost generating units, which would have the effect of increasing price. However, utilities generally recover some amount of fixed costs through variable charges, so a decrease in sales could also have the effect of raising prices overall, as any fixed costs need to be recovered over a smaller volume of sales. Increasing fuel prices, the primary variable cost of electricity production, should also cause prices to increase, so the signs on and 11 See, for example, Bernstein and Griffin (2005) 12 The details of the reduced form estimation are included in Appendix A
31 coefficients should be positive, as many utilities recover fuel expenditures as they are incurred through fuel adjustment charges in their retail rates. The variable costs associated with the production of hydroelectricity are very low, but the availability of hydroelectricity varies with year to year levels of precipitation, realized as either rainfall or accumulated snow pack. However, when the electricity is available, it is available at much lower variable costs. Therefore, the sign on ro is expected to be negative, as increased volumes of hydroelectricity should displace more expensive generating resources. The sign on should also be negative, as increased availability of low priced nuclear generation should result in lower electr icity prices.
32 Results The results o f the estimation of Equation 2 2 are shown in Table 2 3. Table 2 3. 2SLS estimates with entire s ample Variable Coefficient Constant 0.0180*** (0.0029) 0.0504 (0.0899) 0.1650*** (0.0279) 0.0209*** (0.0078) 0.1756*** (0.0553) 0.0143 (0.0183) RTO 0.0200** (0.0089) RTO t 1 0.0284*** (0.0092) RTO t 2 0.0043 (0.0126) R squared of 0.14 Robust standard errors clustered by state in parentheses Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level The coefficient on sales is negative, but not significantly different from zero. Since the dependent variable represents values on the average cost curve, this suggests that the utilities are operating close to the minimum point on the curve. The coefficients for the fuel prices both have the expected signs and are statistically significant at the 1% level, though the electricity pric e is eight times more sensitive to a 1% increase in coal prices than to a natural gas price increase. The broad result that electricity prices are more responsive to changes in coal price than natural gas prices is consistent with Mohammadi (2009), althoug h he finds that coal price elasticity is
33 roughly twice that of natural gas. This result offers further insight into the problem of modeling electricity prices in general. It is a common approach, in modeling electricity prices, to form a fossil fuel price index 13 and use it as a proxy for fuel costs. The result that the coefficients for natural gas prices and coal prices are significant and distinct in this specification suggests that modeling fuel prices in this manner is conveying information that would be unava ilable if the fossil fuel index approach is adopted. Increased availability of hydroelectricity causes the price to decrease, and this decrease is significant. Finally, as indicated by the sum of the coefficients on the RTO and RTO t 1 variables, electricit y prices seem to fall by about 4.8% during the first RTO t 2 variable is not statistically significant, and further lags of the variable yield similar results. This indicates that if an RT O is going to have a price impact on consumers, it occurs in the first two years of its existence. This 4.8% decrease is statistically significant and interesting, because it is at the lower range identified by Joskow (2006), who estimates the price effect s of electric restructuring, utilizing a different data set and methodology, to be 5% to 10%. However, as noted by Kwoka (2006), the effects of restructuring settlements and any imposed rate caps that accompanied those settlements can act as confounding fa ctors, by masking the market prices that might otherwise exist if not for the restructuring agreement. That is, when the equation is estimated with the full sample, the effects of RTOs are indistinguishable from the effects of these rate agreements, if mem bership in an RTO accompanies the restructuring. It would be preferable to simply account for these rate agreements with additional variables, but the form of these agreements, such as the length of time that rate controls are put in place, the restrictive ness of these controls, and the period over which these deferred costs are 13 This index is essentially a weighted average of coal and natural gas prices, as the state s in the sample do not use appreciable quantities of petroleum to generate electricity.
34 recovered, differs greatly from state to state, making the quantification of their effects difficult. Therefore, the best way to control these effects is to remove them altogether. To remove this confounding effect, the equation is estimated with only the sample of states that have not restructured their electric industry. This means that the sample is free of any of the confounding effects of rate agreements on electricity prices, a nd should truly reflect the effects of RTOs, controlling for other factors. Note that membership in an RTO does not require restructuring of the electric utility, as the RTO does not assume ownership of the transmission and distribution assets of the utili ty, so the sample includes states that are within RTOs, but have not restructured their electric industry. The results of the estimation of equation (2) with this restricted sample are shown in Table 2 4. Table 2 4. 2SLS e stimates excluding states that ha ve restructured their electric i ndustry Variable Coefficient Constant 0.0141*** (0.0033) 0.0561 (0.1007) 0.1775*** (0.0366) 0.0263*** (0.0081) 0.2053*** (0.0742) 0.0549 (0.0553) RTO 0.0127 (0.0086) RTO t 1 0.0127 (0.0106) RTO t 2 0.0043 (0.0095) R squared of 0.21 (Robust standard errors clustered by state in parentheses) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level
35 Notice that the signs and significance of most of the variables remains unchanged when the model is estimated with this subset of the data. The magnitudes of the coefficients are consistent as well. However, the variables corresponding to the establishment of an RTO and the effects of that RTO one year later have changed considerably. First, the magnitude of the variables related to the RTO has fallen by roughly half, and second, the precision of their measurement has decreased. Neither variable is signific ant at even the 10% level. Therefore, by eliminating from the sample those 16 states that have restructured their electric industry, the price effects of an RTO are reduced from approximately 4.8%, an effect significantly different from 0%, to 2.5%, but no t significantly different from 0%. This suggests that most of the realized price reductions observed in the initial estimation are not due to the change in the market structure, but the form of the restructuring agreements in the states that chose to restr ucture their markets. Therefore, if there are any cost savings that result from the establishment of RTOs in the absence of electric restructuring, they are not significantly different from zero. Alternate specifications of this model are tested, both as a check on the robustness of the results as well as a way to relax certain assumptions of the original specification of the model. First, the effect of RTO membership on real prices instead of nominal prices is considered. Using the annual consumer price i of the electricity and fuel prices are restated in real terms. Replacing nominal prices with real prices decreases the magnitude of the price effects, once the effects of inflation are rem oved, but does not change the results regarding statistical significance. 14 Second, the original model, as specified, assumes that the marginal effect of changes in fuel price does not vary by state. However, because the availability of resources necessary to 14 Estimation details are available upon request from the author.
36 generate electricity varies with individual state geography, the degree to which each state relies on different types of fuels changes. Therefore, this assumption that marginal effects are constant across states may not be valid. Therefore, another spec ification of the model is estimated with interaction terms between each state and the annual change in the price of coal and natural gas in that state. Table 2 5. 2SLS estimates with entire sample and i nte raction terms between state and fuel p rice Variable Coefficient Constant 0.0175*** (0.0030) 0.1015 (0.0955) 0.1819*** (0.0527) 0.0026 (0.0527) RTO 0.0288*** (0.0086) RTO t 1 0.0325*** (0.0097) RTO t 2 0.0017 (0.0126) R squared of 0.38 (Robust standard errors clustered by state in parentheses) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level The 96 coefficients for the state and fuel price interaction have been omitted from Table 2 5 for the sake of parsimony, but a Wald test rejects the hypothesis that the coefficient of each state with respect to coal prices are equal at the 1% level, and a test of the coefficients on gas prices yields similar results. For illustrative purposes, selected coefficients are listed in Table 2 6 15 15 The coefficients for all 96 interaction terms are available from the author upon request.
37 Table 2 6. Selected coefficients on the interaction between state and fuel p rices State Coefficient Change in log coal p rices Alabama 0.3910*** (0.0081) Florida 0.4741*** (0.0136) Georgia 0.5288*** (0.0226) Minnesota 0.2633*** (0.0093) Change in log natural gas p rices Colorado 0.0678*** (0.0007) Louisiana 0.2260*** (0.0015) Oklahoma 0.1384*** (0.0101) Texas 0.1717*** (0.0062) (Robust standard errors clustered by state in parentheses) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level fuel prices, and the coefficients are consistent with the degree to which these states rely on these y was coal fired, as was 18% of gas for 44% of its generating capacity, while Louisiana, Oklahoma, and Texas are much more reliant on gas for 76%, 65%, and 69% of their capacity, respectively. It is not surprising, then, that the electricity prices in these states would be sensitive to the prices of these fuels. The addition of these variables does not change the results of the analysis, however, as shown in the restricted sample regression results in Table 2 7.
38 Table 2 7. 2SLS estimates with restricted sample and interaction terms between state and fuel p rice Variable Coefficient Constant 0.0191*** (0.0026) 0.0848 (0.0957) 0.1853** (0.0752) 0.0190 (0.0546) RTO 0.0152* (0.0089) RTO t 1 0.0009 (0.0084) RTO t 2 0.0077 (0.0057) R squared of 0.44 (Robust standard errors clustered by state in parentheses) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level The individual state interaction terms change slightly, but remain largely consistent between the two samples. Once again, the effect of the RTO is reduced dramatically, as is the precision with which it is measured. However, with this specification, a reduction of approximately 1.5% in realized electricity prices is observed, and this result is significant at the 10% level. Finally, the state data set also includes prices and sales reported by broad customer class (i.e. residential, commercial, and industrial customers). To see if benefits from RTOs have accrued to particular customer classes, Equation 2 2 is estimated using the prices and sales for each class of customer and the results are reported in Table 2 8. The coefficients are similar in sign and magnitude to the ones in Table 2 4, but now there are two statistically significant results for the RTO variables. The first is a 1.44% decrease in prices for residential customers in the firs t
39 customers, the change in market structure may be producing tangible cost benefits. Residential customers are typically voters, so this group exerts political influence, and industrial customers are important consumers of electricity, so the price benefits for these groups may not be surprising. However, given that roughly 35 different organizations representing large industrial users of electricity contributed to the final version of FERC Order 888, these customers have not seen a sizable reduction in price. Table 2 8. 2SLS estimates by customer class excluding s tate s th at have restructured their electric i ndustry Variable Residential Commercial Industrial Constant 0.0172*** (0.0028) 0.0160*** (0.0043) 0.0064 (0.0054) 0.1175 (0.0830) 0.2058* (0.1099) 0.0483 (0.1417) 0.1355*** (0.0325) 0.1523*** (0.0414) 0.2830*** (0.0897) 0.0054 (0.0075) 0.0075 (0.0104) 0.0616*** (0.0195) 0.0302 (0.0539) 0.0318 (0.0908) 0.6212** (0.3020) 0.0610 (0.0416) 0.0443 (0.0685) 0.0051 (0.0932) RTO 0.0144** (0.0070) 0.0153 (0.0150) 0.0031 (0.0135) RTO t 1 0.0065 (0.0097) 0.0186 (0.0139) 0.0249* (0.0143) RTO t 2 0.0103 (0.0073) 0.0125 (0.0081) 0.0051 (0.0149) (Robust standard errors clustered by state in parentheses) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level
40 Conclusion When FERC established rules to change the structure of the electricity market, it did so under the belief that the existing system was inefficient, and that the change in structure would provide benefits to consumers. Ten years after these original orders, the question regarding benefits of the changes in market structure was raised by the Government Accounting Office, leading to a FERC Request for Comment on the establi shment of performance metrics for ISOs and RTOs. Once these data have been collected, greater insight into the net benefits of the establishment of ISOs and RTOs may be possible. However, the present study provides some immediate insight into this important issue. Utilizing a panel data set of the United States over the past 18 years, this paper estimates equations for annual percentage changes in electricity price, and attempted to identify the degree to which membership in an RTO affects costs. Th ere is a significant effect, a decrease of 4.8% over two years, when estimating these price changes with the entire data sample. However, the entire sample includes the effects of rate agreements that accompanied restructuring agreements in states that cho se to restructure their market. When the equation is estimated excluding the states that restructured their electric industry, the significance of the price change disappears. Therefore, if ISOs and RTOs have led to changes in the price of electricity, the n these changes are indistinguishable from zero or may only apply to certain classes of customer. However, there may be other benefits of RTOs relating to reliability of electricity service or the optimization of long term resource planning that are not es timated here. The question of whether RTOs have influenced system reliability or the long term planning process would be interesting avenues for further research. However, given the time and effort required to comply with the changes in market structure ne cessitated by FERC rules, it is worth asking the question whether all of this effort has provided tangible benefits to electricity consumers at least in terms of lower prices.
41 CHAPTER 3 THE IMPACT OF THE TRANSPARENCY OF WHOLESALE MARKETS ON MARKET PARTICIP ATION: THE CASE OF THE U.S. ELECTRICITY INDUSTRY Introduction 2 000, a docket opened to explore the role of Regional Transmission Organizations (RTO) in the restructured electricity marketplace. The role of an RTO is to administer the electric transmission system, ensuring open access to the grid for all electricity generators. The FERC noted that since FERC Order 888 wa s issued in 1996, trade in the bulk electricity markets had increased significantly. FERC also noted that during the Notice of Proposed Rulemaking process for the instant docket, the Commission smission grid by vertically integrated electric utilities was inadequate to support the efficient and reliable operation that is needed for the continued development of competitive electricity markets, and that continued discrimination in the provision of transmission services by vertically integrated utilities may also 1 FERC further enjoined utilities, state officials, and affected interest groups to voluntarily develop RTOs. Despite the urging of FERC, t here remain substantial portions of the United States electricity grid that are not administered by RTOs or Independent System Operators (ISO). Coase (1960) observed that there are costs involved in carrying out transactions in the over who it is that one wishes to deal with, to inform people that one wishes to deal and on what terms, to conduct negotiations leading up to the bargain, to draw up 1 FERC Order 200 0, issued December 20, 1999, Page 2 (89 FERC 61,285)
42 2 Milgrom and Roberts (1992) categorize these costs as either coordinatio n or motivation costs. They define coordination costs as the need to determine the price and other parameters of the transaction, make the existence of buyers and sellers known to one another, and bring buyers and sellers together. Motivation costs arise f rom incomplete and asymmetric information and imperfect commitment. The wholesale market for electricity, where the relevant product is one kilowatthour (kWh) of electricity delivered to a particular location at a particular point in time, is prone to coor dination costs, as the product has a very short useful life. RTOs and ISOs can have an explicit influence on the coordination costs in the wholesale electricity market, but the direction of that influence is not always clear. One way in which RTOs can infl uence coordination costs is by publishing wholesale electricity prices in a manner in which any interested party can access them 3 This paper employs a panel data set of United States electric utilities spanning the period 1990 2009 to investigate whether the existence of a transparent wholesale market increases the degree to which an electric utility participates in the wholesale market. I find that privately owned utilities and larger utilities increase their participation in a transparent wholesale marke t, while participation of municipally owned utilities is only slightly affected. This indicates that the distribution of the benefits afforded to participants in market administered by RTOs is not uniform across all market participants. The remainder of th e paper is organized as follows: Section II provides a discussion of the costs and benefits of RTOs, Section III provides a review of related literature, Section IV describes the data utilized, Section V describes the empirical model and estimation methodo logy, Section VI reports the results of the estimation, and Section VII offers concluding remarks. 2 Coase (1960) p. 15 3 Further discussions of these costs and benefits follow in Section II.
43 Costs and Benefits of RTOs One way that ISOs and RTOs can influence the development of electricity markets is by providing a transparent wholesale market, wh ich can be defined as a market in which the prices for a unit of electricity delivered to a given location at a given point in time are posted in a manner that is easily accessible by any interested party, such as on a public web site. 4 Consider the case o f a naive electric utility, Alpha, operating as an island, isolated from the electricity grid around it. The utility dispatches its generating units to supply electricity to its customers, and attempts to do so in a manner that optimizes performance, typic ally measured in terms of least cost or some standard of reliability. If electricity demand and the criteria under which the utility determine which of its generating units will be dispatched at any given time. Alpha assesses the availability or operating characteristics, determines how much electricity it must sup ply, and dispatches units sufficient to meet the prevailing demand at the least possible cost. Now consider the existence of a second electric utility, Beta, physically interconnected to Alpha in a neighboring area. Operating as an island, Beta faces the s ame decision as Alpha. However, if both utilities seek to minimize costs, and in a particular hour there is a difference from Beta to Alpha, then the opportun ity for Pareto improvement exists. If Alpha has a higher marginal cost of generation than Beta in a given hour 5 then Beta can generate that marginal kWh 4 Per Bakos (1998). For an example from the Midwest ISO, see https://www.midwestiso.org/MARKETSOPERATIONS/REALTIMEMARKETDATA/Pages/LMPContourMap.asp x 5 This might be due to a difference in the fuel used to generate the electricity or the efficiency with which the fuel is use d by the marginal generating unit of each utility.
44 and sell to Alpha at a price somewhere between their respective marginal costs, and both utilities hav e lowered their effective average costs of generation; Alpha by buying the marginal kWh at less than it would cost to generate it with its own units and Beta by realizing sales revenue greater than the cost to generate the marginal kWh. But the costs that must be incurred in order to achieve this benefit are not limited to the cost of any transmission and the transaction itself. As Milgrom and Roberts observe, coordination costs also arise. Each utility must expend resources to gather information about the electricity system around it. First, each must gather information regarding the number of potential trading partners. Second, each needs information regarding the costs and availability of electricity in any given hour, for every one of those potential tra ding partners, to identify profitable trading opportunities. Third, each needs to know how to make the arrangements necessary to have that electricity delivered to the purchasing utility system if a transaction is agreed upon. Before the advent of RTOs and ISOs, the first and third tasks were often performed by roughly 140 regional balancing authorities (Joskow 2005), organizations registered by the North American Electric Reliability Corporation (NERC) to integrate future resource plans, maintain the balan ce between load, interchange, and generation, and support real time interconnection frequency for a given area. The second function was accomplished primarily through bi lateral contact between utilities, though confederations of utilities also existed. Fo r example, before ISOs and RTOs existed, the Orlando Utilities Commission, the City of Lakeland, and the Florida Municipal Power Agency formed the Florida Municipal Power Pool in ources to meet the
45 By establishing a transparent wholesale market place, however, the RTO can fulfill the second task for the utility, either by maintaining a centralized databank of hourly prices, or by collecting hourly bids and offers from utilities interested in participating in the market. While the RTO can lower the costs required to gather this information, other costs to participate in this market still exist. Utilities must incur costs in orde r to conform to the rules and procedures of this wholesale market, and the ability to trade with utilities that are members of other RTOs may be constrained. In a survey of RTO cost/benefit studies, Eto, Lesieutre, and Hale (2005) report that while utiliti es will incur costs to participate in these markets, these costs had not been explicitly studied. Additionally, Newell and Spees (2011) find that gaps in realized sales of electricity capacity 6 across the PJM/MISO border are caused by institutional barrier s. These barriers include difficulty in obtaining long term firm transmission service to support capacity sales, and energy market must offer requirements that impose risks on capacity importers. Participation in these markets also imposes educational burd ens on utilities. In the PJM Interconnection, a prominent ISO, the manuals describing the administrative, planning, operating, and accounting procedures number over 3000 pages in 34 separate volumes. Therefore, there are countervailing factors that may inf Related Literature The majority of the existing literature on electricity market restructuring has focused on the impacts of the restructuring itself. Kwoka (2006) reviews a number of studies on the price effects of electricity restructuring, and finds that they are plagued by the endogeneity of the treatment variable, electric restructuring, as the states with higher prices tended to restructure their electric industry. He also finds that it is difficult to disentangle the two effects of 6 The capacity product in the electricity industry is the ability to generate electricity on demand, but not the electricity itself.
46 restructuring, the change in market s tructure and the effects of the rate agreements that accompanied restructuring. Fabrizio, Rose, and Wolfram (2007) examine the effects on restructured markets on electric generators and find increases in operating efficiency through reductions in labor and nonfuel operating expenses. Kwoka, Pollitt, and Sergici (2010) study electric distribution systems and find that forced divestiture as a result of electric restructuring has resulted in decreases in efficiency in distribution. Hogan (1995) argues that the ISO must be actively involved with the operation of the wholesale market and system dispatch. However, little empirical work has been done to assess the benefits of the RTOs and ISOs themselves. Blumsack (2007) outlined the problem of evaluating the effic acy of RTOs and found that the metrics used to evaluate them were incomplete and not objective. He proceeded to enumerate nine areas on which evaluation metrics should focus. Davis and Wolfram (2011) studied the changes in operating efficiency of nuclear p ower plants in the United States and found that those operating in competitive wholesale markets increased their efficiency by 10%. Fabrizio (2012) studied the make or buy decisions of 240 investor owned electric utilities from 1990 to 2007 and found that utilities in ISOs tended to meet increases in demand with more purchased power than non ISO members. Outside of the electricity industry, Garicano and Kaplan (2000) have studied the changes in transaction costs as a result of business to business e commerc e and find that the Internet reduces coordination costs. The question of the effects of market transparency has been well studied in the finance long believed that t ransparency the real time, public dissemination of trade and quote information 7 7 Securities and Exchange Commission (1994), Page IV 1
47 Pagano and Roell (1996) studied stylized trading systems with differing degrees of transparen cy and found that while greater transparency lowered trading costs for uninformed traders on average, it did necessarily lower them for every size trade. Madhavan (1996) demonstrated that market transparency can increase price volatility and lower market l iquidity, but that these effects disappear in markets that are sufficiently large. Chandley and part of the purpose of RTO de sign was to facilitate trading 8 and show that the day ahead net exports from the Midwest to the PJM region tripled when American Electric Power became a member of PJM in October of 2004. 9 FERC Order 2000 identified five benefits that RTOs can offer, relative to the existing market structure: improved efficiencies in the management of the transmission gr id, improved grid reliability, non discriminatory transmission practices, improved market performance, and lighter handed government regulation. 10 This study examines whether utilities within RTOs participate more in the wholesale electricity markets, relat ive to utilities outside of RTOs, either due to improved efficiencies in grid management or non discriminatory transmission practices. Data The primary data source for this study is the Form 861 database compiled by the U.S. Information Administration. The reporting of information collected on the Form 861 is an annual requirement for all privately and publicly owned electric utilities in the United States and its territories. Data collected includes the quantity of wholesale and retail purchases and sales revenues, number of customers, annual system peak load, as well as information on demand side management programs, green pricing and net metering programs, 8 Chandley and Ho gan (2009), Page 33 9 Chandley and Hogan (2009), Page 34 10 FERC Order 2000, issued December 20, 1999, Page 70 71 (89 FERC 61,285)
48 and distributed generation capacity. The utilities also report their control area operator on the form, which allows the identification of the time periods during which the utility is a part of an RTO that has established a transparent wholesale market. The transparency mechanism employed by the various RTOs, the posting o f wholesale prices on a public website, are nearly identical, so the effect of this mechanism is treated as homogenous across RTOs. Total sources and disposition of energy on the form is disaggregated into several categories that are important for this stu (reported as net generation), and purchases from the wholesale market (reported as purchases). Together, these accounts are aggregated as total electricity sources for the utility. The total sources of electricity in a given year must always equal the total disposition of electricity, which is disaggregated into sales to ultimate consumers (retail sales), sales for resale (wholesale sales), and electricity losses (losses du e to the transmission or distribution of electricity). The data set consists of over 64,000 data points, each representing the response of one electric utility to the EIA 861 survey for one year from 1990 through 2009. This data set is an unbalanced panel, with roughly 3000 to 4000 utilities responding in any given year. However, these utilities enter and exit the sample in a non random fashion, and the inclusion of all utilities in the sample can lead to selection bias (Heckman 1979). Therefore, this analy sis employs a balanced panel, only those utilities which have submitted data over the entire 20 year data collection period. The questions of whether utilities purchase or sell more electricity in the wholesale markets, in the presence of an RTO, will be a ddressed separately. Initially, only utilities with positive sales to ultimate consumers, that is, utilities which serve retail electric load are considered. This is designated the initial purchase sample. Utilities that do not themselves
49 generate electric ity in any year of the sample are excluded from this sample. These utilities are 11 of another utility, and therefore lack the means to serve their electric load, except by purchasing electricity on the wholesale market. T he wholesale market interactions of these utilities would therefore be unaffected by the presence of a transparent market because they are restricted to purchasing 100% of their electricity regardless of whether the wholesale market is transparent. The dep endent variable for this sample is the fraction of the total sources of energy that is purchased from the wholesale market. The nave utility Alpha in the initial example would purchase none of its energy requirements in the wholesale market, and its parti cipation in the market may be limited by the coordination costs. As these coordination costs change, the utility may find it beneficial to participate in the market. Initially, the utility may only participate in the market when necessary (i.e. when it has insufficient generation to meet its needs, perhaps due to unit outages), and the percentage of its energy that it purchases in the wholesale market may be very low. However, as coordination costs evolve, the utility may also look for economic opportunitie s to displace its own generation with market purchases, thus increasing the percentage of its requirements that it purchases. In this manner, the dependent variable might change for each utility over time with changes in coordination costs. Participation c ould also be measured by the volume of wholesale purchases, but this would be expected to increase as the electricity requirements of the utility grow. Normalizing these purchases by the total sources of electricity removes this potential bias. Similarly, the initial sales sample includes all utilities with positive net electricity generation in a given year, with the exception of any utility that sold all of that generation in the wholesale market over the entire time period in the study. These utilities a re likely wholesale 11 These are utilities that serve retail electricity customers but purchase all of the electricity required to serve them on the wholesale market.
50 generators, and the presence of a transparent wholesale market will have no effect on whether they participate in the wholesale market. The dependent variable in this case is the fraction of total disposition of energy that is sold on t he wholesale market. Broader criteria may be used to derive the samples, however. Recall that the initial purchase sample excluded any utility that did not generate electricity in any year during the sample period. However, a transparent wholesale marketp lace might afford utilities that do not generate electricity the opportunity to purchase electricity not needed to serve retail load, and then resell that electricity to another retail provider. Utilities that exploit this opportunity in the wholesale mark et are excluded from the initial sample, but the presence of a transparent wholesale market may still influence their behavior. Therefore, the second purchase sample includes all utilities in the initial purchase sample, and all utilities that reported sal es for resale during the sample period. This sample is much larger, and affords the opportunity to use the majority of the data points. Similarly, the second sales sample encompasses generating utilities that serve ultimate consumers during some period dur ing the sample. Unlike the broader purchase criteria, this does not lead to a sizable increase in the portion of the sample used. Model The model to be estimated is the dependent variable ( DV ), which is either the fraction of the total sources of energy th at comes from the wholesale market (for the Purchase regressions), or the fraction of total disposition of energy that is sold on the wholesale market (for the Sale regressions). (3 1)
51 Changes in the dependent variable are explained by a utility specific fixed effect, the number of utilities that exist in the 48 contiguous United States in the given year ( Mk tUtils ), a linear time trend ( Time ), an indicator variable equal to 1 if the utility is a member of an RTO that operates a transparent wholesale market in that year ( ISO_Whl ), the number of years that the utility has been in a transparent wholesale market ( ISOYrs ), the size of the utility measured by its summer peak demand ( SumPk ), and indicator variables equal to 1 depending on the ownership of the utility ( Federal if it is a federal power project, Muni if a municipally owned utility, and IOU if a privatel y owned utility). Our variables of interest include the ISO_Whl and ISOYrs variables, as well as the interaction between these variables and the size and ownership variables. Finally, the error terms for each utility were found to exhibit first order seria l correlation, and are thus modeled as AR(1) processes. Descriptive statistics for the purchase samples are given in Table 3 1, and the sales samples in Table 3 2. Table 3 1. Mean and s tandard deviation of purchase s ample All Purch1 Purch2 Purchase% 0.9297 0.2136 0.8335 0.2942 0.9419 0.1911 MktUtils 3230.8 213.4 3217.8 190.7 3217.8 190.7 ISO_Whl 0.1458 0.3560 0.1623 0.3688 0.1436 0.3507 ISOYrs 0.6127 1.7849 0.7317 2.0041 0.6061 1.7818 SumPk 281.16 2765.44 634.86 2491.64 266.04 2821.21 Federal 0.0022 0.0471 0.0048 0.0694 0.0020 0.0453 Muni 0.5873 0.4923 0.7299 0.4440 0.6367 0.4809 IOU 0.0612 0.2397 0.1339 0.3406 0.0528 0.2237 N 61370 19405 55484
52 Table 3 2. Mean and s tandard deviation of sales s ample All Sales1 Sales2 Sales% 0.1567 0.3239 0.2284 0.3452 0.2583 0.3697 MktUtils 3277.0 265.9 3217.8 190.7 3217.8 190.7 ISO_Whl 0.2231 0.4164 0.1480 0.3551 0.1510 0.3581 ISOYrs 0.9561 2.1612 0.6359 1.8277 0.6551 1.8607 SumPk 463.74 2016.93 1081.18 3027.79 1056.29 2982.65 Federal 0.0045 0.0671 0.0134 0.1148 0.0129 0.1127 Muni 0.5701 0.4951 0.6231 0.4846 0.6061 0.4886 IOU 0.0997 0.2997 0.2066 0.4049 0.2137 0.4099 N 35784 9819 10165 utilities in its area. This variable is especially notable because it is the catalyst for the interaction in the hypothetical example of utilities Alpha and Beta. However, hourly wholesale price data is not available for utilities that do not participate in transpar ent wholesale markets, the control group for this study. In lieu of this data, the effect of cost differentials could be modeled with a variety of annual aggregated regional price differentials, such as mean and maximum differentials. Doing so failed to ge nerate coefficients on these variables that were significant at any reasonable level, and did not affect the magnitude or statistical significance of other variables in the model. Moreover, the relatively high R 2 values in the regressions reported below su ggest that the explanatory power of any omitted variables is relatively small. The treatment effect in the model, whether the utility is a member of an organization that operates a transparent wholesale market, might be seen as endogenous, but it is import ant to note that membership in an RTO or ISO is mandatory for any utility located in a state that
53 restructured its electricity market, and that the decision to restructure the market was made by the state legislatures, and not the utility itself. Further, utilities that operate within the control area of a larger utility may find themselves compelled to join an RTO if their control area operator does so. Finally, as argued in Kwoka (2006), price is often cited as the decision to initiate changes in the elec tricity market, not purely participation in the market itself. However, additional analyses are performed in this paper with a sample free from endogeneity concerns, and the basic results still stand. The utility specific fixed effect accounts for the fac t that utilities serve their load obligations with different combinations of owned generation and purchased power. Due to the long lived nature of generating assets, this fixed effect will simply reflect the average purchases and sales of the utility over time, and will be relatively stable. The Market Utilities variable is expected to be positive, as the liquidity of the market should increase as more utilities are participating in it. The remaining variables are the variables of interest, although the nul l hypothesis suggests that the effects of the constraints imposed by the transparent wholesale markets would be less than the effects of the cost reduction of the information regarding electricity availability and price, and that the coefficients on these variables will be positive. A variable to track how long the utility has been involved with a transparent wholesale market is also included, to discern whether the length of time that utilities have been exposed to this market changes the degree to which t hey participate. Results The results of the estimation with the initial sample are given in Table 3 3
54 Table 3 3. Parameter estimates for i nitial s ample Variable % Purchased % Sold Constant 0.5532*** (0.0088) 0.1572*** (0.0034) MktUtils 1.19e 05*** (2.60e 06) 1.62e 05*** (2.70e 06) Time 0.0040*** (0.0004) 0.0011** (0.0005) ISO_Whl 0.0140 (0.0159) 0.0180* (0.0095) ISOYrs 3.55e 03 (4.73e 03) 2.93e 03 (4.14e 03) ISOYrs 2 3.07e 04 (2.78e 04) 5.91e 04 (6.48e 04) SumPk 2.72e 06*** (5.86e 07) 1.04e 06*** (3.66e 07) Federal 3.36e 05 (8.24e 02) Muni 0.2360*** (0.0341) 0.0035 (0.0115) IOU 0.2181*** (0.0723) 0.0431 (0.0469) SumPk x ISO_Whl 5.15e 06** (2.13e 06) 2.38e 06 (1.51e 06) Muni x ISO_Whl 0.0150 (0.0163) 0.0273*** (0.0102) IOU x ISO_Whl 0.0016 (0.0211) 0.0132 (0.0134) SumPk x ISOYrs 2.47e 07 (1.06e 06) 4.72e 07 (8.79e 07) SumPk x ISOYrs 2 1.19e 07 (1.13e 07) 4.71e 08 (9.29e 08) Muni x ISOYrs 0.0049 (0.0039) 0.0084*** (0.0032) IOU x ISOYrs 0.0461*** (0.0086) 0.0043 (0.0069) IOU x ISOYrs 2 0.0025*** (0.0007) 0.0006 (0.0006) N 18425 9295 Number of clusters (utilities) 980 524 R squared 0.8736 0.9532 Rho 0.6820 0.7784 (Standard errors in parentheses) (Blanks indicate coefficients omitted due to collinearity) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level
55 The number of utilities in the market has a positive and significant effect on the fraction of wholesale purchases and sales for the utilities, but the magnitude of the effect is not large. The coefficient implies that an additional 1000 utilities, increasing the market size by approximately 25%, would result in an extra 1.2% in purchases or 1.6% in sales. Since the relev ant product in the wholesale electricity market is a kWh of electricity delivered to a particular location, the presence of an additional utility in the state of Ohio, say, would likely have little effect on the degree of market participation of a utility near Los Angeles, and this effect is reflected in the magnitude of this coefficient. It appears, from the time trend, that utilities have been purchasing about 0.4% more and selling about 0.1% more of their electricity in the wholesale market every year. T he number of years exposed to the wholesale market does not have a statistically significant effect, but does when interaction terms are considered. The coefficient on the size of the utility alone indicates that larger utilities have a tendency to purchas e and sell less electricity. However, the magnitude of this effect is very small. For a utility with a peak demand of 1000 MW, for example, slightly smaller than the utility in Knoxville, Tennessee, the effect on purchases would be 0.3% and on sales would be 0.1%. The interaction terms are far more interesting. They indicate that a larger utility sells more in a transparent wholesale market. The same Knoxville sized utility purchases an additional 5% of its electricity in an ISO. They also indicate that m unicipally owned utilities decrease their sales into a transparent wholesale market by approximately 2.7%, but that experience in the market increases sales by 0.8% per year. This may occur if the transactions costs of the market are not fully understood, but more information regarding their magnitude is gained over time 12 Meanwhile, privately owned utilities participate in the markets to a much greater degree, increasing their purchases by 4.3% but this participation 12 For example, on April 25, 2006, FERC ordered MISO to recalculate revenue sufficiency guarantee charges retroactive to May 1, 2005, as a result of the misapplication of their tariff. (115 FERC 61,108)
56 increases at a decreasing rate. The coe fficients for the purchase sample imply the relationship shown in Figure 3 1. Figure 3 1 Effect of the IOUx ISOYrs and IOUxISOYrs2 coefficients on wholesale market purchases of the initial s ample While the magnitude of the coefficients imply that the effect on participation will eventually become negative, it is important to realize that this point, sometime in the 19 th year, is beyond the time horizon of the sample. Transparent wholesale markets have only existed in the sam ple for 12 years, so these coefficients may not reflect the nature of this relationship over a longer period of time. It is clear that, in the time frame of this analysis, experience in the markets increases participation at a decreasing rate. Similarly, t he coefficients for privately owned utilities in the sales sample, and larger utilities in the purchase and sales samples, imply that market participation increases at an increasing rate within the time period of study sample, but this behavior cannot be e xpected to continue indefinitely.
57 So, while the participation of municipal utilities in transparent wholesale markets increases gradually in time, larger utilities and privately owned utilities seem to participate more in a transparent whole market. These broad results are similar in concept to the results of Rose and Joskow (1990) who concluded that larger utilities and privately owned utilities adopted new gas fired generating technologies sooner than smaller and municipally owned utilities. In this insta nce, the creation of a transparent wholesale electricity market can be seen as the technological innovation being adopted by the utilities. Further, the results for privately owned utilities are consistent with the results of Fabrizio (2012). Estimating th e regression for the expanded sample changes the coefficients, but does not change the basic results, as shown in Table 3 4. Recall that this expended sample includes utilities that may not own generation themselves, but purchase electricity in excess of t he needs of their customers to resell on the wholesale market. The effect of the number of utilities is positive and significant for both samples. The time trend is still positive and significant, but smaller in magnitude. The presence of the market itself increases sales by 2.0%. Again, larger utilities increase participation in the ISO markets, with the sales for a 1000 MW utility increasing almost 5.0%. Municipal utilities in the sample exhibit a similar pattern to the initial sample, with an initial dec rease in sales, and a subsequent increase over time. Larger utilities again exhibit a quadratic increase in purchases, to the temporal limit of our sample. Table 3 4. Parameter estimates for expanded s ample Variable % Purchased % Sold Constant 0.5405*** (0.0030) 0.1576*** (0.0032) MktUtils 6.55e 06*** (9.21e 07) 1.72e 05*** (2.60e 06) Time 1.83e 03*** (1.48e 04) 1.33e 03*** (4.69e 04) ISO_Whl 2.42e 03 (3.29e 03) 0.0203** (0.0090)
58 Table 3 4. Continued Variable % Purchased % Sold ISOYrs 2.92e 03** (1.27e 03) 1.16e 03 (3.92e 03) ISOYrs 2 1.43e 04 (1.04e 04) 8.01e 04** (3.26e 04) SumPk 1.08e 07 (6.98e 08) 1.03e 06*** (3.62e 07) Federal 0.0349 (0.0497) Muni 0.5546*** (0.0146) 5.19e 04 (1.01e 03) IOU 0.1587*** (0.0222) 0.1158*** (0.0444) SumPk x ISO_Whl 4.91e 06*** (1.23e 06) 2.40e 06 (1.48e 06) Muni x ISO_Whl 3.15e 03 (3.67e 03) 0.0299*** (0.0097) IOU x ISO_Whl 0.0127 (0.0084) 0.0100 (0.0127) SumPk x ISOYrs 4.32e 07 (6.11e 07) 4.67e 07 (8.58e 07) SumPk x ISOYrs 2 1.62e 07** (6.46e 08) 4.56e 08 (9.02e 08) Muni x ISOYrs 0.0011 (0.0010) 7.35e 03** (3.01e 03) IOU x ISOYrs 0.0427*** (0.0041) 1.60e 03 (6.60e 03) IOU x ISOYrs 2 0.0023*** (0.0004) 7.21e 04 (6.13e 04) N 52682 9621 Number of clusters (utilities) 2802 544 Rho 0.6907 0.7787 R squared 0.8839 0.9578 (Standard errors in parentheses) (Blanks indicate coefficients omitted due to collinearity) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level The pattern for privately owned utilities is similar as well. The coefficients for market experience imply the relationship in Figure 3 2, but it is still important to consider the temporal
59 limits of the sample. Once again, the effect on market participation for municipal utilities is small relative to the effect for larger and privately owned utilities. Figure 3 2 Effect of the IOUxISOYrs and IOUxISOYrs2 coefficients on wholesale market purchase s of the expanded s ample Relative to the ownership status of the utility, the size of the utility has a smaller effect on the degree of market participation. Figure 3 3 shows the effect of the interaction between the size of the utility and its experience in an ISO. These coefficients are from Table 3 4 and illustrate the change in market purchases for a 1000 MW utility. Note that for the first four years, the effect is s peak load would have to be 10,000 MW, or larger than the City of Los Angeles, for the magnitude of the size and experience effect to be equivalent to the privately owned utility experience effect.
60 Figure 3 3 Effect of the S umPkxISOYrs and SumPkxISOYrs2 coefficients on wholesale market purchases of a 1000 MW utility in the expanded s ample As discussed earlier, the dependent variable for market participation could be thought of as endogenous. In order to evaluate whether this endogeneity might have an effect on the results, the estimation is repeated using only states that restructured their electricity industry. This restructuring, enabled by FERC and initiated by state legislatures, required utilities to either assets, this third party was the ISO or RTO. Thus, utilities in restructured states that joined RTOs did so not of their own accord, but because they were compelled by the state legislature. As discussed in Kwoka (2006), the motivation for states to restructure was high electrici ty prices, and not the dependent variable in this analysis, so a sample consisting only of restructured states should be free of these endogeneity concerns. The results of this estimation are shown in Table 3 5.
61 Table 3 5. Parameter estimates for e xpande d sample: utilities in states that restructured their electricity i ndustry Variable % Purchased % Sold Constant 0.6523*** (0.0062) 0.1682*** (0.0055) MktUtils 5.79e 06** (2.38e 06) 3.23e 05** (5.88e 06) Time 4.14e 03*** (6.29e 04) 0.0035** (0.0017) ISO_Whl 6.18e 04 (6.57e 03) 0.0325** (0.0153) ISOYrs 5.95e 03** (2.53e 03) 6.50e 03 (7.06e 03) ISOYrs 2 2.15e 04 (1.79e 04) 8.29e 04 (5.60e 04) SumPk 4.49e 07 (4.84e 07) 9.51e 07* (4.91e 07) Muni 0.2813*** (0.0335) 0.0034 (0.0275) IOU 0.1389*** (0.0427) SumPk x ISO_Whl 8.84e 06*** (2.08e 06) 5.48e 06* (3.03e 06) Muni x ISO_Whl 2.13e 04 (7.32e 03) 0.0437** (0.0172) IOU x ISO_Whl 0.0316** (0.0142) 0.0181 (0.0231) SumPk x ISOYrs 1.99e 06** (9.73e 07) 3.00e 07 (1.62e 06) SumPk x ISOYrs 2 2.23e 07** (9.80e 08) 2.61e 08 (1.54e 07) Muni x ISOYrs 8.24e 04 (1.77e 03) 1.08e 02** (5.12e 03) IOU x ISOYrs 0.0687*** (0.0069) 0.0078 (0.0120) IOU x ISOYrs 2 0.0041*** (0.0006) 7.13e 04 (1.05e 03) N 12675 2898 Number of clusters (utilities) 686 173 Rho 0.7467 0.8136 R squared 0.8525 0.9298 (Standard errors in parentheses) (Blanks indicate coefficients omitted due to collinearity) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level The coefficients in Table 3 5 differ from those in Table 3 4, but the basic results of the analysis remain. Market participation in the purchase market tends to increase for larger utilities
62 and privately owned utilities. Municipal utilities experience an initial drop in level of sales participation, but increase sales with experience. Larger utilities overall tend to sell less, but the amount is small in magnitude. Thus, the potential endogeneity of the dependent variable is not driving the results of the analysis. Since the results for restructured states are consistent with those for the entire sample, this might beg for the question of whether restructuring is solely responsible for the results. To test whether this is t rue, the estimation is repeated using the complement of the data set in Table 3 5, just those states that did not restructure their electricity industry. The results of this estimation are shown in Table 3 6. Table 3 6. Parameter estimates for expanded s am ple: utilities in states that did not restructure their electricity i ndustry Variable % Purchased % Sold Constant 0.4463*** (0.0035) 0.1822*** (0.0031) MktUtils 5.92e 06** (9.50e 07) 4.79e 06* (2.52e 06) Time 1.23e 03*** (1.10e 04) 9.70e 04*** (3.30e 04) ISO_Whl 3.22e 03 (3.93e 03) 1.75e 03 (1.13e 02) ISOYrs 9.88e 04 (1.75e 03) 4.98e 03 (4.92e 03) ISOYrs 2 1.99e 04 (2.02e 04) 5.10e 04 (4.33e 04) SumPk 9.10e 08 (6.40e 08) 1.06e 06 (1.02e 06) Federal 0.0157 (0.0415) Muni 0.7452*** (1.41e 02) 7.81e 04 (8.96e 03) IOU 0.0540** (0.0257) 0.1060*** (0.0334) SumPk x ISO_Whl 1.04e 06 (1.76e 06) 7.36e 09 (1.90e 06) Muni x ISO_Whl 2.66e 03 (4.34e 03) 3.36e 03 (1.20e 02)
63 Table 3 6. Continued Variable % Purchased % Sold IOU x ISO_Whl 0.0030 (0.0127) 9.40e 03 (1.62e 02) SumPk x ISOYrs 1.84e 06* (1.05e 06) 7.99e 07 (1.52e 06) SumPk x ISOYrs 2 5.35e 07*** (1.41e 07) 9.24e 08 (2.50e 07) Muni x ISOYrs 6.79e 05 (1.32e 03) 0.0014 (0.0040) IOU x ISOYrs 1.77e 02*** (1.06e 02) 7.79e 03 (8.24e 03) IOU x ISOYrs 2 1.60e 03** (6.90e 04) 1.30e 03 (8.25e 04) N 39986 6704 Number of clusters (utilities) 2137 390 Rho 0.6092 0.7193 R squared 0.9201 0.9782 (Standard errors in parentheses) (Blanks indicate coefficients omitted due to collinearity) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level The observed pattern in the purchase market continues to hold, with larger utilities and privately owned utilities tending to purchase more. No variables of interest remain statistically significant in the sales sample, suggesting that restructuring may be driving the results in the sales sample. However, the overall results for the purchase s ample seem robust to different subsamples. Finally, to test whether the results are driven by the relationship between municipally owned utilities and privately owned utilities, the participation equations can be separately estimated with each subsample. T he results of these estimations are shown in Table 3 7 and Table 3 8.
64 Table 3 7. Parameter estimates for expanded sample: municipal u tilities Variable % Purchased % Sold Constant 0.8858*** (0.0016) 0.0260*** (0.0025) MktUtils 1.35e 05*** (1.24e 06) 1.31e 05*** (3.14e 06) Time 2.84e 03*** (1.65e 04) 1.35e 03*** (4.72e 04) ISO_Whl 1.25e 03 (2.49e 03) 4.59e 03 (4.62e 03) ISOYrs 4.74e 03*** (1.29e 03) 2.61e 03 (2.90e 03) ISOYrs 2 8.47e 05 (1.22e 04) 4.26e 04 (3.03e 04) SumPk 2.24e 06 (4.39e 06) 1.15e 06 (3.99e 06) SumPk x ISO_Whl 2.46e 05** (1.15e 05) 6.77e 06 (1.07e 05) SumPk x ISOYrs 1.60e 05** (6.56e 06) 4.95e 06 (6.82e 06) SumPk x ISOYrs 2 2.29e 06*** (7.39e 07) 4.78e 07 (7.26e 07) N 33471 5709 Number of clusters (utilities) 1785 338 Rho 0.6304 0.7803 R squared 0.8567 0.9293 (Standard errors in parentheses) (Blanks indicate coefficients omitted due to collinearity) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level Table 3 8. Parameter estimates for expanded sample: investor owned u tilities Variable % Purchased % Sold Constant 0.3397*** (0.0075) 0.1879*** (0.0067) MktUtils 4.05e 05*** (7.31e 06) 2.86e 05*** (6.71e 06) Time 7.60e 03*** (2.52e 03) 5.50e 04 (1.68e 03) ISO_Whl 0.0219 (0.0156) 0.0346** (0.0167)
65 Table 3 8. Continued Variable % Purchased % Sold ISOYrs 0.0235** (0.0092) 0.0045 (0.0095) ISOYrs 2 1.48e 03* (7.91e 04) 3.48e 04 (8.65e 04) SumPk 5.14e 06*** (1.83e 06) 6.75e 07 (1.75e 06) SumPk x ISO_Whl 6.20e 06** (2.80e 06) 3.71e 06 (3.02e 06) SumPk x ISOYrs 4.60e 07 (1.43e 06) 1.18e 06 (1.64e 06) SumPk x ISOYrs 2 5.59e 09 (1.49e 07) 1.07e 07 (1.60e 07) N 2757 2131 Number of clusters (utilities) 151 117 Rho 0.8263 0.7881 R squared 0.8153 0.8454 (Standard errors in parentheses) (Blanks indicate coefficients omitted due to collinearity) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level The basic results still hold when the individual ownership samples are considered. Municipal utilities show a decrease in purchases in ISOs, but this effect is mitigated for larger utilities. For the privately owned utilities, experience in ISOs increases purchase market participation at a decreasing rate, consistent with the earlier results. And once again, participation in the purchase market increases with the size of the utility. In the sales sample, privately owned utiliti es in ISOs tend to sell about 3% more. So while the coefficients change in these subsamples, the basic results of the analysis remain the same: municipal utilities tend to participate less in the ISOs, while privately owned and larger utilities tend to par ticipate more, and these results are robust to different samples.
66 Conclusions It is clear that RTOs and ISOs can provide opportunities in the electricity market that might not otherwise exist. One such opportunity is the facilitation of the transparent who lesale electricity market. Transparent wholesale markets can reduce coordination costs that limit the participation of utilities in the marketplace, and thus limit opportunities that might arise with that participation. However, these formal markets also i mpose costs that may discourage participation in the wholesale market. This paper estimates the determinants of market participation, and shows that the presence of a transparent wholesale marketplace for electricity has the effect of increasing participat ion in the wholesale market, but that this participation does not occur symmetrically across all types of electric utilities. Greater participation is induced in privately owned and larger utilities, reflecting both the results of Rose and Joskow, who foun d that privately owned and larger electric utilities are more willing to adopt technological innovations in the electricity industry, and Fabrizio, who found that privately owned utilities in ISOs tend to meet more of their growing electricity needs by pur chasing electricity. These results have important implications for public policy aimed at increasing transparency in wholesale electricity markets, and the organizations that facilitate it, as the opportunities afforded by this policy may not be uniformly distributed across all market participants.
67 CHAPTER 4 CHALLENGES IN QUANTI FYING OPTIMAL CO2 EM ISSIONS POLICY: THE CASE OF ELECTRICITY GENERATI ON IN FLORIDA Introduction Questions regarding the economic impact of carbon dioxide (CO 2 ) emissions co ntinue to accompany any discussion regarding the imposition of emission limits. However, most of these discussions focus on only one side of the relationship between CO 2 prices and emissions levels. That is, they attempt to quantify the resulting CO 2 price implied by an exogenous level of emissions, or the emissions level that would result from a given emissions price. Studies that take either price or emissions level as exogenous may not offer insight into the question of the optimal level of abatement 1 b y ignoring the interaction between them. This paper, considers the effects of a range of CO 2 prices, thereby informing an analysis of the average cost curves for emissions abatement, which provides insight into the unusual behavior of the marginal costs of abatement. Such insight is necessary in any discussion of optimal levels of emissions abatement. Florida Summit on Global Climate Change ght business, government, science, and stakeholder leaders together to discuss the effects of climate change on Florida and the nation. On the second day of the summit, July 13, the Governor signed three Order 07 126 mandated a 10% reduction of greenhouse gas emissions from state government by 2012, 25% by 2017, and 40% by 2025. Order 07 127 mandated a reduction of greenhouse gas emissions from the state of Florida to 2000 levels by 2017, 1990 levels by 20 25, and 20% of 1990 levels by 2050. Finally, Order 07 128 established the Florida Governor s Action Team on Energy and Climate Change and 1 Target ed levels of emission r eduction are frequently alliterative, such as 50% by 2050.
68 charged the t eam with the development of a comprehensive Energy and Climate Change Action Plan. On June 25, 2008, Flor ida House Bill 7135 was signed into law by Governor Crist, creating Florida Statute 403.44 which states: The Legislature finds it is in the best interest of the state to document, to the greatest extent practicable, greenhouse gas emissions and to pursue a market based emissions abatement program, such as cap and trade, to address greenhouse gas s tate government is to place a cap on the amount of carbon dioxide emitted by the electric power generation sector. Florida legislature commissioned a study of the economic impacts on the state of such a program. This paper utilizes a version of the model 2 constructed for that study (Kury and Harrington 2010) to simulate the dispatch of electric generating units in the state of Florida over a range of CO 2 prices. The analysis concludes that the marginal cost of abatement curve may not be well behaved, implying several points where the marginal costs of abatement are equal to the marginal benefits. This behavior can complicate the question of an optimal level of CO 2 abatement. The remainder of this paper is organized as follows: Section II provides a review of the literature on the eco nomic effects of CO 2 emissions, Section III describes the model of economic dispatch, Section IV describes the data sources utilized, Section V discusses the mechanics of the simulation, Section VI discusses the model results, and Section VII provides some concluding remarks. 2 The inputs to the model have been updated.
69 Literature Review Nordhaus (1980) is credited as being the first to derive optimal levels of CO 2 emissions, in a model of the CO 2 cycle and CO 2 abatement. He further discussed a model of the effects of CO 2 buildup in the environment and the intertemporal choice of consumption paths, and ended with suggestions regarding how to compare control strategies. He also identified three empirical issues with policy implementation: the problem that CO 2 emission is an externality across space and time, whether to control CO 2 emissions with quantities or prices, and the effects of uncertainty regarding the costs and benefits of CO 2 abatement. Further theoretical research has explored aspects of the Nordha us model, such as Goulder and Mathai (2000), who characterized optimal carbon taxes and CO 2 abatement under different channels for knowledge accumulation, under cost effectiveness and benefit cost criterion. The bulk of the literature consists of ex ante studies of proposed levels of emissions abatement. In the United States, the Congressional Budget Office (CBO), Environmental (EIA) have all studied the effects of le gislation proposed in the House of Representatives and the Senate. These analyses typically treat the levels of emissions proposed in the bills as exogenous, and attempt to determine their economic impact. For example the EIA analysis of the American Power Act of 2010 concluded that CO 2 emissions prices in the Base Case would reach $32 per ton in 2020 and $66 per ton in 2035. This analysis is limited in its ability to offer insight into policy alternatives, however. Studies on the regional economic impact o f CO 2 pricing on the market for electric generation have been performed for the ERCOT region in Texas 3 as well as the PJM region in 3 http://www.ercot.com/content/news/presentations/2009/Carbon_Study_Report.pdf
70 the Northeastern United States 4 Examining the conclusion for those two studies shows how the relative carbon intensity of the electric generation fleet can have a marked impact on the economic effects of CO 2 pricing. Pennsylvania relies on more coal fired generation, and therefore the impact of a $1 increase in carbon prices results in a $0.70/MWh increase in wholesale electr icity prices. Texas, which relies more on natural gas sees its wholesale prices increased approximately $0.50/MWh with a $1 increase in CO 2 prices. Similar analyses have been conducted for the European market. Honkatukia et. al (2006) studied the degree to which allowance prices in the European Union Emissions Trading System for CO 2 get passed through to the wholesale prices in Finland, and concluded that 75% to 95% of the price change is passed through to the spot price. A comparative analysis was conducte d by Newcomer et al. (2008) who modeled the short run effects of a range of CO 2 prices on the price of electricity and level of carbon dioxide emissions in three regions of the United States, but the determination of an optimal level of abatement was beyon d the scope of their work. The literature on the social costs of CO 2 emissions presents a diverse range. The Contribution of Working Group II to the Fourth Assessment Report to the Intergovernmental Panel on Climate Change (2007) cited the results of a sur vey of 100 estimates of the social cost of CO 2 that reported a range from $3 per ton to $95 per ton. This survey was taken from Tol (2005), which reported a mean of $43 per ton (in 2005$) of carbon with a standard deviation of $83 per ton of carbon 5 in th e peer reviewed studies. In its modeling, the Interagency Working Group on the Social Cost of Carbon, United States Government (2010), uses values from $5.70 4 http://www.pjm.com/documents/~/media/documents/reports/20090127 carbon emissions whitepaper.ashx 5 Th ese figures convert to $11.62 and $22.43 per ton of CO 2 respectively.
71 to $72.80 (in 2007$) for the social cost of CO 2 in 2015, and $15.70 to $136.20 for 2050. Anthoff a nd Tol (2013) analyze the factors that affect the uncertainty in the social cost of carbon and find that the influence of parameters changes depending on the time scale of the analysis or the region considered. They also find that some parameters are more certain than others. Ackerman and Stanton (2012) demonstrate that with high climate sensitivity, high climate damage, and a low discount rate, the social cost of CO 2 could be almost $900 per ton in 2010. Model of Economic Dispatch The problem of least co st economic dispatch of a group of n electric generating units is to minimize the aggregate costs required to provide the amount of electricity demanded by end users in each hour. The costs to produce this electricity will be driven by the type of generating unit, its thermal efficiency 6 the variable costs required to operate and maintain the unit, and the price of its fuel. For each hour, the problem can be stated: (4 1) subject to the constraints: where: G i MWh generated by the i th generating unit C i Maximum hourly generating capacity in MWh of the i th generating unit. L Electricity demand by consumers in MWh CO2 i Tons of CO 2 emitted per MMBtu of fuel consumed by the i th generating unit 6 The thermal efficiency of a power plant is the rate at which it converts units of fuel to a given unit of electricity. This is typically called the heat rate of a power plant, and all else equal, a lower heat rate is preferred to a higher one.
72 Emit$ Emissions cost per ton of CO 2 Fuel$ i Cost of fuel per MMBtu consumed by the i th generating unit HeatRate i Heat rate of the i th generating unit in MMBtu of fuel required to produce one MWh of electricity O&M$ i Hourly operation and maintenance expenses of the i th generating unit in $/MWh Without a price for emitting CO 2 the value of Emit$ is zero and the amount of CO 2 emitted by that generating unit does not enter the dispatch equation. With a positive value for Emit$ the total cost of emissions is driven by the operating efficiency of the generating unit and by the type of fuel utilized as some generating fuels emit relatively more carbon dioxide when burned. Such fuels which include coal and petroleum coke, are often referred to as dirty fuels F uels that emit relatively less carbon dioxide when burned, such as natural gas, are referred to as clean fuels. There fore, the price of emissions may necessitate the switch from a dirtier generating fuel to a cleaner one by an individual generator capable of burning more than one type of fuel, or may lead to a generator that burns a dirtier fuel being replaced by a gener ator that burns a cleaner fuel. The strategies to reduce emissions from the electric generation sector are limited in the short run. Generators can adjust the types of fuels that they use, known as fuel switching, or reduce the amount of electricity that t expand to strategies such as improving the heat rate of existing power plants (thus reducing fuel consumption), construction of new power plants that produce electricity while emitting less (or no) carb on dioxide, or developing and exploiting technologies that captures a portion of the carbon dioxide emitted. The model utilized in this paper allows for both short run and long run strategies
73 The determination of the optimum hourly unit dispatch is conduc ted in two stages. First, the hourly operating cost is minimized for each available generating unit. For units with the capability to burn different fuels, the cost and emissions rate of each fuel are considered and the least cost alternative is selected. Second, all of the generating units are ordered from lowest cost to highest, and the units with the lowest hourly costs are dispatched until the hourly electric loads are met. Data Sources Data on the hourly marginal costs for individual generating units i s considered proprietary, so these costs must be estimated. Data for individual generating units, such as summer and winter generating capacity the type of generating unit and fuel sources, are available from the EIA Form 860 ( Annual Electric Generator R eport ) and Form 861 ( Annual Electric Power Industry Database ) databases. Data on generating unit operating efficiency, such as heat rate, are available from EIA Form 423 ( Monthly Cost and Quality of Fuels for Electric Plants Data ) filings. The heat rate da ta utilized in this simulation represents the annual average heat rate for each generating unit. Some unit level operating data, such as variable operating and maintenance expenses, are available from utility resp onses to the Form 1 (Annual Report of Major Electric Utility) required by the Federal Energy Regulatory Commission (FERC). Other operating data is derived from industry averages published by the EIA for use in its Annual Energy Outlook. Unit specific operating and contract data 7 as well as long term load forecasts, are available from the Regional Load and Resource Plan published by the Florida Reliability Coordinating Council. Actual hourly loads are available from utility responses on Form 714 (Annual Electric Control and Planni ng Area Report) to the FERC Data for planned generating 7 Contract data includes power purchased from other states under long term contracts As a result, the costs associated with these contracts are sunk, and their marg inal cost of dispatch is zero.
74 units are available from the FRCC Regional Load and Resource Plan. Projected fue l prices and levelized costs of new generation are taken from the 2013 Annual Energy Outlook published by the EIA. Mode l Operation Within each month of a given run, the model first determines the order in which the generating units will be dispatched to meet electric load, often called the generation stack, and then dispatches the generation stack against the monthly load shape on an hourly basis using Equation 4 1 When dispatching each unit, the model discounts each uni t s production capacity by the stinct operating characteristics of different types of gene rating units. Electrical generation for different types of units may or may not be controlled by the operator of the unit For a unit that burns fossil fuels, if the power plant is running and has fuel available, it will generate electricity. These types o f units are also called dispatchable units. For a unit that relies on the sun or the wind to generate electricity, however, that power plant will not produce electricity if the sun is not shining or the wind is not blowing. These types of units are called nondispatchable units. For nondispatchable units, the availability factor reflects the amount of time that the sun is shining or the wind is blowing. For dispatchable units, this availability factor reflects the times when the unit is available to generate accounts for t he unit being unavailable due to either a planned or unplanned outage. Ideally, two factors would be used to reflect unit availability. The first would reflect planned unit outages, mos t commonly for routine maintenance. The second factor would reflect unplanned, or forced, outages ; the instances where a unit breaks down unexpectedly. However, individual unit outage
75 schedules are proprietary and dynamic To ameliorate these modeling limi tations, this availability factor is employed. The long run strategies employed by the model consider the decisions to build new power plants. The model can either be allowed to build new generating units only when they are necessary to serve electric load or might be allowed to build new units opportunistically, that is, when the wholesale price of electricity is sufficient to allow the new units to earn a profit. The former approach may not induce generation sufficient to reach more aggressive emission r eduction targets, as the composition of the generation fleet is more static, while the latter approach may lead to the problem of stranded investment. Because the construction of new generating units in Florida is regulated through a determination of need proceeding at the state Public Service Commission 8 the former approach has been modeled in this analysis. The opportunistic approach was also modeled, but yielded similar results. Changes in the outlook for natural gas prices limited the emissions reducti ons that could be achieved even with the opportunistic approach, however. In Kury and Harrington (2010), a carbon price of $90 per ton was sufficient to induce a change in construction to zero emitting technologies (nuclear and biomass), while the latest p rices for natural gas and generating technologies now require a carbon price of $135 per ton to induce the same behavior. Model Output During its execution, the model tracks the e lectricity production for each unit, as well as the units of fuel burned, the total dispatch costs, and the carbon emissions. These output variables are be aggregated by utility, type of plant, fuel type, and plant vintage 8 Florida Statutes 403.519
76 The aggregate model output consists of mat ched sets of emissions prices, emissions levels, and the volume of each generating fuel burned for each model year. Therefore, each level of emissions in a particular year implies a price of emissions and a fuel mix, and vice versa. In that manner, the mod el determines the price of emissions and mixture of generating fuels that correspond to each level of carbon dioxide emissions, for each compliance year in the analysis. Further, it also compute s the effects of different levels of emissions (and the result ing emissions prices) to allow the characterization of the marginal effects of the emissions policy. The model was run for the years 2012 2025, varying the CO 2 price from $0 to $100 per ton, and the change in several output variables is presented. The first variable is the change associated with the real incremental cost component of electricity production, shown in Figure 4 1. Figure 4 1. Real ($2010) incr emental c ost of electricity by year and emissions p rice While the relationship between emissions prices and incremental cost does change slightly as we look further into the future, t he relationship between emissions prices and
77 incremental cost is fairly stable, as a $1 increase in emissions prices tends to raise the price of electricity in Fl orida by approximately 50 ¢ per MWh, or about $6 per year for a family that uses 1000 kWh per month T his effect drops to about 40 ¢ per MWh as emissions prices increas e to $100 per ton. The magnitude of the effect of CO 2 prices on incremental cost reflects the relative carbon intensity of the generating units utilized to produce electricity, so a decrease in the effects of an emissions price as the emissions per MWh of electricity decreases is expected. Figure 4 2 Emissions by year and emissions p rice Figure 4 2 illustrates the effects of carbon dioxide emissions prices on the emissions of the electric generating sector. Emissions levels are initially reduced 2 3% under relatively low emissions prices. This is primarily due to the displacement of petroleum coke and inefficient coal generators as a source of electricity in Florida. However, emissions levels then reach a plateau whose magnitude varies with the year, during which incre asing the price of emissions has relatively little effect on overall emissions levels. Once emissions prices exceed a critical value
78 however, a rapid decline in emissions levels occurs. This decline in emissions occurs at $15 per ton in the short run, as coal in the surface, however, are cause for concern for policymakers. These areas are regions in which costs are increasing for consumers 9 in the form of higherrealized costs, but with little corresponding decrease in emissions. Consumers are thus paying higher costs without any concurrent benefit. The results shown in Figures 4 1 and 4 2 can be consolidated to construct the average cost curves for emission s abatement in a given year. Figure 4 3 shows these average cost curves for selected years in the simulation. Figure 4 3 Average cost of a batement c urves While the marginal cost of abatement cannot be observed from a discrete model, the shape of the marginal cost curve can be inferred from the behavior of these average cost curves. 9 As seen in Figure 4 1.
79 The marginal cost curves for the years 2015, 2020, and 2025 clearly cross the average cost curve multiple times. Therefore, the marginal benefits curve for emission abateme nt, even if it is itself well behaved, may intersect the marginal cost curve at more than one level of emissions abatement. To illustrate this phenomenon, the average and incremental costs of abatement for 2015 are shown in Figure 4 4. Figure 4 4 Ave rage and incremental cost of abatement c urves for 201 5 The incremental cost curve for 2015 is clearly not well behaved, sloping upward over several levels of abatement. T here is considerable uncertainty surrounding the social cost of CO 2 abatement. If a CO 2 tax of $700 per ton is established, a value at the upper range of the social cost of CO 2 established in the literature, the tax would be equal to the incremental cost of abatement at approximately 1 million, 42 million, and 45 million tons of avoided CO 2 This phenomenon is illustrated in Figure 4 5.
80 Figure 4 5. Multiple abatement e quilibria at a carbon t ax of $700 If the tax were equal to $70 per ton, a level within the range of the social cost of CO 2 established by the Interagency Working Group of the United States Government (2010), it would be equal to the incremental cost of abatement at approximately 18 million and 34 million tons of avoided CO 2 This is illustrated in Figure 4 6.
81 Figure 4 6 Multiple abatement e quilibria at a carbon tax of $70 The challenge for policymakers is that when the optimal level of CO 2 abatement is considered, using the criteria of equating marginal costs with marginal benefits, there may not be a single optimum leve l. Therefore, even if global leaders were to agree on the marginal costs and marginal benefits of CO 2 abatement, an accomplishment that is likely difficult to achieve, there is still the potential to disagree on the optimum level. This would make it diffic ult to agree on the level of an emissions cap, if that method of regulation is implemented. Further, if emissions control through a carbon tax is considered, it may not result in the desired level of emissions abatement. Therefore, if a specific level of C O 2 abatement is desired by policy makers, the implementation of an emissions cap is the only reliable way to achieve it.
82 Figure 4 7 Fuel c onsumption in 2013 Figure 4 7 illustrates the amount of coal (BIT) natural gas (NG) and petroleum coke (PC) burned during the simulation of 2013. These results provide insight into the shape of the emissions surface. Initial reductions in emissions levels occur as petroleum coke and inefficient coal plants, relatively dirty sources of electricity, are displaced. Once the petroleum coke has been fully displaced, further increases in emissions prices eliminate half of the remaining coal consumption and emissions levels decrease rapidly. Once an CO 2 price of $30 per ton is reached, the displacement of the remaining coal fired capacity continues, but at a much lower rate. At CO 2 prices of $70 per ton, for example, the initial consumption of coal has been reduced by 60%. Conclusions It is easy to find discussions of government imposed carbon dioxide abatement targets a nd the emissions prices that result from these targets, but the literature on discussions of policy alternatives or the establishment of optimal emissions abatement is not well developed. Since emissions abatement carries a cost to the consumer, however, i t is important to ensure that those
83 costs are commensuarate with the benefits that consumers are receiving from this abatement policy. This paper present s the results of an analysis of the units used to generate electricity in Florida and the marginal eff ects of carbon prices on their dispatch. Using the operating cost economic dispatch model, this paper analyze s the effects that various emissions prices (and their concurrent emissions levels) have on generation. We find that at relatively low emissions prices emissions levels decrease as fuel sources such as petroleum coke and coal burned in less efficient pla nts are displaced. Once this initial reduction has been achieved, further increases in carbon prices may do little to decrease emissions until a critical point has been achieved and most coal generation can be displaced by natural gas. These results sug gest that the marginal effects of emissions prices may vary greatly with the level of emissions abatement and the fundamenta l characteristics of the market. The question of what constitutes optimal emissions abatement policies is complicated by the potenti al for the marginal cost of abatement curves to oscillate. This paper demonstrates how the incremental cost of abatement curves may intersect with a CO 2 tax at many levels of ts over the optimum level of abatement, then, can occur even if parties agree completely on the marginal costs and marginal benefits of abatement, complicating the formation of public policy.
84 CHAPTER 5 CONCLUDING REMARKS AND OPPORTUN ITIES FOR FURTHER RE SEARCH The provision of electricity is critical to life in the United States, and a better understanding of the effects and challenges of changes in the sector can lead to improvements in consumer and producer welfar e. The electric ity industry has experienced significant changes over the last twenty years, new structure and new priorities, and analysis of the effects of these changes can advance this understanding. Electric restructuring led to a change in the organiz ation of the electricity industry in the United States. The problems that have accompanied this restructuring have received a considerable amount of attention, but it is also important to consider the benefits that have accrued. New organizations to facili atate access to the transmission grid have resulted in opportunities, but as discussed here, no significant price effects for consumers. And while access to the grid has resulted in more participation in the wholesale market, this benefit of access has not been universally exploited. These are not the only possible benefits of RTOs, however. RTOs may also improve system reliability and improve the long term planning process for generation and transmission resources. These questions are not addressed here an d remain avenues for further research. Another challenge facing the electricity industry center s on the the externalities associated with the emission of greenhouse gases during the combustion of fossil fuels. Most of the work inthis area has focues on the quantification of the costs and benefits of emissions abatement, and the theoretical perspective on what constitutes an optimal level of emissions abatement. However, these models rely on cost and benefit curv es that are well behaved in the economic sense. But this assumption of well behaved cost curves may not be a valid one. I have derived the curves for electricity generation in Florida here, but the model could be generalized
85 and the scope could be exp anded in further research. Indeed, an understanding of the marginal cost curves for the electricity generation and transportation sectors is crucial to the question of what constitutes an optimal level of abatement for the United States. The challenges we face are formidable, but we have tools with which to address them, and I hope that we can advance our understanding of these issues, to the benefit of society.
86 APPENDIX A TEST OF ENDOGENEITY OF SALES To test whether is endogenous in Equation 2 2 Equation A 1 is estimated ( A 1 ) Where: Pop State population PCI State per capita income CDD State population weighted cooling degree days HDD State population weighted heating degree days Sales Electricity sales PCoal Nominal price of coal PGas Nominal price of natural gas %Hydro Percent of electric generation from hydroelectric sources %Nuc Percent of electric generation from nuclear sources RTO Whether the majority of the electric customers in the state are served by a utility that belongs to an RTO The results of this estimation are shown in Ta ble A 1. The residuals from this reduced form estimation are included as independent variables in the estimation of Equation 2 2 The coefficient on the residuals i s significant at the 1% level 1 indicating that the variable Sales is 1 Coefficient was 0.6788 with a standard error of 0.1175
87 endogenous in Equation 2 2. Therefore, Equation 2 2 is estimated with the two stage least squares technique (2SLS) utilizing the variables , and as instrumental variables for Table A 1. OLS estimates of the log r e turn of electric s ales Variable Coefficient Constant 0.0053 (0.0033) 0.5659*** (0.0802) 0.2200*** (0.0512) 0.0436*** (0.0052) 0.0982*** (0.0152) 0.0393*** (0.0141) 0.0131*** (0.0044) 0.1431*** (0.0498) 0.0042 (0.0276) RTO 0.0020 (0.0027) RTO t 1 0.0017 (0.0034) RTO t 2 0.0070** (0.0035) R squared of 0.22 F test statistic is 15.70 and significant at the 1% level (Robust standard errors clustered by state in parentheses) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level
88 APPENDIX B TEST OF THE STRENGTH OF INSTUMENTAL VARIA BLES The results of the first stage r egressions of the estimates of Equation 2 2 are provided below. Table B 1 and Table B 2 apply to the entire sample, while Table B 3 and Table B 4 apply to the restricted sample. Table B 1. First stage estimates of log return of electricity sales with entire s ample Variable Coefficient Constant 0.0053 (0.0032) 0.0393*** (0.0141) 0.0131*** (0.0044) 0.1431*** (0.0498) 0.0042 (0.0276) RTO 0.0020 (0.0027) RTO t 1 0.0017 (0.0034) RTO t 2 0.0070** (0.0035) 0.0436*** (0.0052) 0.0982*** (0.0152) 0.5659*** (0.0802) 0.2200*** (0.0512) R squared of 0.22 (Robust standard errors clustered by state in parentheses) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level
89 Table B 2. Partial R 2 values for excluded i nstruments Variable Partial R 2 0.1101 0.0626 0.0394 0.0171 All of the coefficients on the IV for sales are significant at the 1% level. Additionally, the Cragg Donald Wald F statistic for this regression is 47.83, and exceeds the 5% critical value from Stock and Yogo (2005) at the 5% level, so the null hypothesis that the instrumental variable s are weak in this estimation is rejected. Table B 3. First stage estimates of log return of electricity sales excluding restructured s tates Variable Coefficient Constant 0.0092** (0.0041) 0.0291 (0.0193) 0.0185*** (0.0053) 0.1820*** (0.0678) 0.0033 (0.0596) RTO 0.0049 (0.0032) RTO t 1 0.0002 (0.0039) RTO t 2 0.0114*** (0.0040) 0.0465*** (0.0072) 0.0856*** (0.0167) 0.4930*** (0.1028) 0.2116*** (0.0584)
90 R squared of 0.21 (Robust standard errors clustered by state in parentheses) Statistically significant at the 10% level ** Statistically significant at the 5% level *** Statistically significant at the 1% level Table B 4. Partial R 2 values for excluded i nstruments Variable Partial R 2 0.1154 0.0475 0.0327 0.0163 Again, all of the coefficients on the IVs for kWh sales are significant at the 1% level, and the Cragg Donald Wald F statistic for this regression is 30.04, and exceeds the 5% critical value from Stock and Yogo (2005) at the 5% level, so the null hypothesis that the instrumental variables are weak in this estimation is rejected.
91 APPENDIX C THE MODEL OF ECONOMI C DISPATCH This appendix provides details on the operation and use of the model that solves the problem of least cost dispatch of the electricity generation system. Main Routine The Main routine acts a program shell that calls all of the other subroutines and manages the main program loop, the hourly system dispatch. The routine begins by calling the routines WriteInputLoad WriteInputFuelPrice and WriteInputGenUnits These routines archive the input data used to create the run: the hourly MWh load, the fuel prices, and the beginning generating units for the system. It t hen reads in the parameters for the beginning and ending dates, and the high, low, and interval CO 2 prices used in the simulation. The module loads the generating units used and loads each data element associated with the unit. Any missing elements are ass igned default ratings for that particular generator type. The data elements include: Owner Plant Name Unit Number Type of generator Summer Capacity Rating in MW Winter Capacity Rating in MW Unit Availability Percentage Unit Heat Rate Variable Operating and Maintenance Expenses in $/MWh The number of fuels utilized by the plant Avail a bility date of the unit Retirement date of the unit The program then begins the main operating loop over the CO 2 prices for the run. There are also nested loops over the months over the simulation time period and the hours in each month. Within each month, the fuel and emissions prices are used to rank the different generators available from lowest variable cost of operation to greatest. Then, within each hour of the month,
92 unit s are dispatched to serve the hourly electric load until the demand is satisfied. The hourly marginal cost of generation is added to a flat file. The model then repeats this dispatch process for each hour of the month. Once the month has been simulated, th e model repeats the rank ordering process for the next month, updating the fuel prices used in the simulation. Once all of the time periods have been processed, the model repeats the process with the new CO 2 price. The model writes detailed output on costs and generator statistics to a flat file in comma delimited format, and aggregate output to tabs of the Excel workbook. A description of each intermediate subroutine and function follows.
93 No No No No Yes Yes Yes START Read Run Date and CO2 Price Parameters Read Electric Load Forecasts Read Fuel Price Forecasts Read Available and Planned Generation Loop Over CO2 Prices Loop Over Months Sort Generation Stack by Variable Cost of Operation Utilize Lowest Cost Unit to Serve Demand Excess Demand? Loop Over Hours Hours Remaining? Months Remaining? Prices Remaining? Write Output Data for Month END Set Hourly Demand Write Output Data for Time Period Yes Yes Yes Yes No No No No
94 All code written by Ted Kury Director of Energy Studies Public Utility Research Center Option Explicit Type GenFuel Name As String Index As String Type As String Adder As Double Type Index As Long PriceIndex As Long End Type Type GenFuelBurn Type As String BurnMMBtu As Double End Type Type GenPlant Owner As String PlantName As String UnitName As String PlantType As String DEPTypeCode As String DEPCapUnit As Integer SumCap As Double WinCap As Double HeatRate As Double VOM As Double AvailPct As Double NumFuels As Integer Fuel() As GenFuel CCS As Double 'Percentage of emissions captured, default is 0 AvailDate As Date RetireDate As Date DispatchFlag As Long LevelCost As Double FixedCost As Double End Type Type PlantOutput MWh As Double FuelCost As Double FuelBurn() As GenFuelBurn VOMCost As Double End Type Dim dL oadShape() As Double
95 Dim aGenUnits() As GenPlant Dim aOutput() As PlantOutput Dim dGenStack() As Double Dim dMaxMWhByFuel() As Double Dim dCarbonPrice As Double Dim dtBeginDate As Date Dim dtEndDate As Date Dim lGenCnt As Long Dim bConstrainGenPort As Boolean
96 Sub Main() Dim dDispatch() As Double Dim dDispatchSum() As Double Dim dAnnualSum() As Double Dim dAnnualSumType() As Double Dim dAnnualSumVintage() As Double Dim dAggregateSum() As Double Dim dAggregateFuelCosts() As Double Dim dAggregateFuelMM Btu() As Double Dim dREMIFuelData() As Double Dim i As Long Dim j As Long Dim k As Long Dim m As Long Dim dTemp As Double Dim dTempMWh As Double Dim dMaxLoad As Double Dim dGenCap As Double Dim dReserveMargin As Double Dim tempPlant As GenPlant Dim iFuels As Integer Dim iPlantIdx As Integer Dim dtCurrDate As Date Dim iHours As Integer Dim dCarbonHigh As Double 'Carbon price upper bound Dim dCarbonLow As Double Dim dCarbonInt As Double Dim dHourlyCap As Double Dim dSeasonalCap As Double Dim iDateIdx As Integer Dim iDateCnt As Integer Dim iTemp As Integer Dim iLoadShapeIdx As Integer Dim iDEPTypeIdx As Integer Dim iNumDEPTypes As Integer Dim iNumFuelTypes As Integer 'Number of fuel types Dim lLoadDateIdx As Long Dim lOutRow As Long 'Row for summary output Dim lAggOutRow As Long Dim lAggSectionOffset As Long Dim lInitialGenCnt As Long 'Number of stock plants before additions Dim dtLoadDate As Date Dim sDir As String Dim sOutfile As String Dim sUnitFile As String Dim sPriceFile As String Dim sOutstring As String Dim sREMIFile As String
97 Dim oldStatusBar As String Dim dElapsedRunTime As Double Dim dTotalRunTime As Double Dim dFixedCostAdjust As Double Dim dFixedGenCost As Double Dim dNewGenCost As Double Dim NewPlan t() As GenPlant iNumDEPTypes = 12 dReserveMargin = 0.15 dFixedCostAdjust = 58 oldStatusBar = Application.DisplayStatusBar Application.DisplayStatusBar = True Application.ScreenUpdating = False sDir = ActiveWorkbook.Path & \ & Format(Now(), "yymmddhhm m") & \ 'get path of workbook MkDir sDir Call WriteInputLoad(sDir) Call WriteInputFuelPrice(sDir) Call WriteInputGenUnits(sDir) sOutfile = "PlantOutput.csv" sUnitFile = "NewUnits.csv" sPriceFile = "MarginalCosts.csv" sREMIFile = "REMIInputs.csv" Open sDir & sOutfile For Output As #1 Open sDir & sUnitFile For Output As #2 Open sDir & sPriceFile For Output As #3 Open sDir & sREMIFile For Output As #4 'Read in run parameters dtBeginDate = Range("BegDate") dtEndDate = Range("EndDate") iDateIdx = DateDiff( "m", dtBeginDate, dtEndDate) 'Read in carbon price parameters dCarbonHigh = Range("CarbonHigh") dCarbonLow = Range("CarbonLow") dCarbonInt = Range("CarbonInt") bConstrainGenPort = Range("GenPortCheckStatus") dTotalRunTime = (iDateIdx + 1) (((dCarbonHi gh dCarbonLow) / dCarbonInt) + 1) 'Find number of fuel types
98 Do Until IsEmpty(Sheet5.Cells(iNumFuelTypes + 2, 7)) iNumFuelTypes = iNumFuelTypes + 1 Loop iNumFuelTypes = iNumFuelTypes 1 ReDim dMaxMWhByFuel(iNumFuelTypes, 1) 'Element 0 is Row Number 'Element 1 is MWh limit Call GetPortConstraint 'Clear output sheets Sheet4.Range("A2:AE65536").Clear Sheet9.Range("A1:AF65536").Clear Sheet10.Range("A2:P65536").Clear Sheet13.Range("A3:BE65536").Clear 'Read in generating units lGenCnt = 0 Do Unt il IsEmpty(Sheet1.Cells(lGenCnt + 2, 1)) iPlantIdx = 0 ReDim Preserve aGenUnits(lGenCnt) aGenUnits(lGenCnt).Owner = Sheet1.Cells(lGenCnt + 2, 1) aGenUnits(lGenCnt).PlantName = Sheet1.Cells(lGenCnt + 2, 2) aGenUnits(lGenCnt).UnitName = S heet1.Cells(lGenCnt + 2, 3) aGenUnits(lGenCnt).PlantType = Sheet1.Cells(lGenCnt + 2, 4) aGenUnits(lGenCnt).SumCap = Sheet1.Cells(lGenCnt + 2, 6) aGenUnits(lGenCnt).WinCap = Sheet1.Cells(lGenCnt + 2, 7) aGenUnits(lGenCnt).AvailPct = Sheet1.C ells(lGenCnt + 2, 8) aGenUnits(lGenCnt).HeatRate = Sheet1.Cells(lGenCnt + 2, 9) aGenUnits(lGenCnt).VOM = Sheet1.Cells(lGenCnt + 2, 10) aGenUnits(lGenCnt).NumFuels = Sheet1.Cells(lGenCnt + 2, 11) aGenUnits(lGenCnt).AvailDate = Sheet1.Cells(l GenCnt + 2, 18) aGenUnits(lGenCnt).RetireDate = Sheet1.Cells(lGenCnt + 2, 20) aGenUnits(lGenCnt).DEPCapUnit = Sheet1.Cells(lGenCnt + 2, 21) aGenUnits(lGenCnt).DEPTypeCode = Sheet1.Cells(lGenCnt + 2, 22) aGenUnits(lGenCnt).DispatchFlag = Sheet1.Cells(lGenCnt + 2, 19) iFuels = Sheet1.Cells(lGenCnt + 2, 11) 1 ReDim aGenUnits(lGenCnt).Fuel(iFuels) For i = 0 To iFuels aGenUnits(lGenCnt).Fuel(i).Type = Sheet1.Cells(lGenCnt + 2, i + 12) aGenUnits(lGenCnt).Fuel(i).Ad der = Sheet1.Cells(lGenCnt + 2, i + 15) aGenUnits(lGenCnt).Fuel(i).TypeIndex = GetFuelIndex(aGenUnits(lGenCnt).Fuel(i).Type) aGenUnits(lGenCnt).Fuel(i).PriceIndex = GetFuelPriceIndex(aGenUnits(lGenCnt).Fuel(i).Type) Next i 'Check fo r missing plant data and fill in the blanks If IsEmpty(Sheet1.Cells(lGenCnt + 2, 8)) Then
99 iPlantIdx = PlantTypeLookup(aGenUnits(lGenCnt).PlantType, aGenUnits(lGenCnt).Fuel(0).Type) aGenUnits(lGenCnt).AvailPct = Sheet5.Cells(iPlantIdx + 1, 5) End If If IsEmpty(Sheet1.Cells(lGenCnt + 2, 9)) Then If iPlantIdx = 0 Then iPlantIdx = PlantTypeLookup(aGenUnits(lGenCnt).PlantType, aGenUnits(lGenCnt).Fuel(0).Type) aGenUnits(lGenCnt).HeatRate = Sheet5.Cells(iPlantIdx + 1, 3) End If If IsEmpty(Sheet1.Cells(lGenCnt + 2, 10)) Then If iPlantIdx = 0 Then iPlantIdx = PlantTypeLookup(aGenUnits(lGenCnt).PlantType, aGenUnits(lGenCnt).Fuel(0).Type) aGenUnits(lGenCnt).VOM = Sheet5.Cells(iPlantIdx + 1, 4) End If If IsEmpty(Sheet1.Cells(lGenCnt + 2, 18)) Then aGenUnits(lGenCnt).AvailDate = #1/1/1980# If IsEmpty(Sheet1.Cells(lGenCnt + 2, 20)) Then aGenUnits(lGenCnt).RetireDate = #1/1/2080# lGenCnt = lGenCnt + 1 Loop lGenCnt = lGenCnt 1 lInitialGenCnt = lGenCnt 'Start loop by carbon prices and by date lOutRow = 2 lAggOutRow = 2 For dCarbonPrice = dCarbonLow To dCarbonHigh Step dCarbonInt dFixedGenCost = 0 lGenCnt = lInitialGenCnt ReDim Preserve aGenUnits(lGenCnt) lLoadDa teIdx = dtBeginDate Sheet3.Cells(1, 1) + 1 'Determine least cost form or input mix of new generation Call DefineNewPlant(NewPlant, dCarbonPrice) dNewGenCost = PriceNewPlant(NewPlant) ReDim dDispatchSum(iNumFuelTypes, 3, lGenCnt) ReDim dAnnualSum(iNumFuelTypes, 3) ReDim dAnnualSumVintage(1, iNumFuelTypes, 3) ReDim dAggregateFuelCosts(iNumFuelTypes, 3) ReDim dAggregateFuelMMBtu(iNumFuelTypes) ReDim dAnnualSumType(iNumDEPTypes, 3) ReDim dAggregateSum(3) For iDateCn t = 0 To iDateIdx Application.StatusBar = "Program & Round(dElapsedRunTime / dTotalRunTime 100, 0) & "% Completed" dtCurrDate = DateAdd("m", iDateCnt, dtBeginDate) iHours = (DateAdd("m", 1, dtCurrDate) dtCurrDate) 24 1 ReDim dLoadShape(iHours) 'Read in load file
100 iLoadShapeIdx = 0 dMaxLoad = 0 Do For i = 2 To 25 dLoadShape(iLoadShapeIdx) = Sheet3.Cells(lLoadDateIdx, i) If Sheet3.Cells(lLoadDateIdx, i) > dMaxLoad Then dMaxLoad = Sheet3.Cells(lLoadDateIdx, i) iLoadShapeIdx = iLoadShapeIdx + 1 Next i lLoadDateIdx = lLoadDateIdx + 1 dtLoadDate = Sheet3.Cells(lLoadDateIdx, 1) Loop While Month(dtL oadDate) = Month(dtCurrDate) And Year(dtLoadDate) = Year(dtCurrDate) 'Check to add new generating units dGenCap = 0 For i = 0 To lGenCnt If dtCurrDate >= aGenUnits(i).AvailDate And dtCurrDate < aGenUnits(i).RetireDate T hen Select Case Month(dtCurrDate) Case 5 To 9 dGenCap = dGenCap + aGenUnits(i).SumCap Case Else dGenCap = dGenCap + aGenUnits(i).WinCap End Select End If Next i dMaxLoad = dMaxLoad (1 + dReserveMargin) If dGenCap < dMaxLoad Then dTemp = dMaxLoad dGenCap Do Until dTemp < 0 If Not Range("CheckBoxStatus") Then j = 0 Else j = UBound(NewPlant) End If For i = 0 To j lGenCnt = lGenCnt + 1 ReDim Preserve aGenUnits(lGenCnt) ReDim Preserve dDispatchSum(iNumFuelTypes, 3, lGenCnt) aGenUnits(lGenCnt) = NewPlant(i) aGenUnits(lGenCnt).AvailDate = dtCurrDate aGenUnits(lGenCnt).RetireDate = DateAdd("yyyy", 50, dtCurrDate) dFixedGenCost = dFixedGenCost + aGenUnits(lGenCnt).FixedCost sOutstring = dCarbonPrice & "," & dtCurrDate & "," & aGenUnits(lGenCnt).SumCap & "," & aGenUnits(lGenCnt).PlantName
101 Print #2, sOutstrin g Next i dTemp = dTemp 1000 Loop End If Call SetGenStack(dtCurrDate, dCarbonPrice) Call BubbleSortGenStack(1) ReDim dDispatch(lGenCnt + 30, iHours, iNumFuel Types, 3) 'dDispatch elements are: 0 Dispatch MW 1 Variable costs (fuel, emissions, VOM) 2 Emissions 3 Fuel Burn For i = 0 To iHours dTemp = dLoadShape(i) If bConstrainGenPort Then For j = 0 To UBound(dMaxMWhByFuel) If dMaxMWhByFuel(j, 0) <> 0 Then dMaxMWhByFuel(j, 1) = dTemp Sheet14.Cells(dMaxMWhByFuel(j, 0), Year(dtCurrDate) 2007) End If Next j End If j = 0 Do Until dTemp <= 0 If j > UBound(dGenStack, 2) Then 'Add a new plant If Not Range("CheckBoxStatus") Then m = 0 Do k = NewPlant(m).Fuel(0).TypeIndex If dMaxMWhByFuel(k 2, 1) <> 1 Then lGenCnt = lGenCnt + 1 ReDim Preserve aGenUnits(lGenCnt) ReDim Preserve dDispatchSum(iNumFuelTypes, 3, lGenCnt) aGenUnits(lGenCnt) = NewPlant(m) aGenUnits(lGenCnt).Ava ilDate = dtCurrDate aGenUnits(lGenCnt).RetireDate = DateAdd("yyyy", 50, dtCurrDate) dFixedGenCost = dFixedGenCost + aGenUnits(lGenCnt).FixedCost Call AddGenStac k(dGenStack, NewPlant(m), lGenCnt) sOutstring = dCarbonPrice & "," & dtCurrDate & "," & aGenUnits(lGenCnt).SumCap & "," & aGenUnits(lGenCnt).PlantName
102 Print #2, sOutstring Exit Do Else m = m + 1 End If Loop Else m = UBound(NewPlant) For k = 0 To m lGenCnt = lGenCnt + 1 ReDim Preserve aGenUnits(lGenCnt) ReDim Preserve dDispatchSum(iNumFuelTypes, 3, lGenCnt) a GenUnits(lGenCnt) = NewPlant(k) aGenUnits(lGenCnt).AvailDate = dtCurrDate aGenUnits(lGenCnt).RetireDate = DateAdd("yyyy", 50, dtCurrDate) dFixedGenCost = dFixedGenCost + aG enUnits(lGenCnt).FixedCost Call AddGenStack(dGenStack, NewPlant(k), lGenCnt) sOutstring = dCarbonPrice & "," & dtCurrDate & "," & aGenUnits(lGenCnt).SumCap & "," & aGenUnits(lGenCnt).PlantName Print #2, sOutstring Next k End If End If tempPlant = aGenUnits(dGenStack(0, j)) Select Case Month(dtCurrDate) Case 1, 2 dHourlyCap = tempPlant.WinCap dSeasonalCap = 0.05 Case 3, 4, 10, 11, 12 dHourlyCap = tempPlant.WinCap dSeasonalCap = 0.05 Case 5 dHourlyCap = tempPlant.SumCap dSeasonalCap = 0.05 Case 6 To 9 dHourlyCap = tempPlant.SumCap dSeasonalCap = 0.05 End Select If tempPlant.AvailPct + dSeasonalCap > 1 Then dSeasonalCap = 1 tempPlant.AvailPct If tempPlant.AvailPct + dSeasonalCap < 0 Then dSeasonalCap = tempPlant.AvailPct dHourlyCap = dHourlyCap (tempPlant.AvailPct + dSeasonalCap) Selec t Case dMaxMWhByFuel(dGenStack(2, j) 2, 1) Case 1 dHourlyCap = 0
103 Case Is > 0 dMaxMWhByFuel(dGenStack(2, j) 2, 1) = dMaxMWhByFuel(dGenStack(2, j) 2, 1) dHourlyCap If dMaxMWhByFuel(dGenStack(2, j) 2, 1) <= 0 Then dMaxMWhByFuel(dGenStack(2, j) 2, 1) = 1 End Select If dTemp > dHourlyCap Then dDispatch(dGenStack(0, j), i, dGenStack(2, j) 2, 0) = dHourlyCa p dDispatch(dGenStack(0, j), i, dGenStack(2, j) 2, 1) = dHourlyCap dGenStack(1, j) dDispatch(dGenStack(0, j), i, dGenStack(2, j) 2, 2) = dHourlyCap dGenStack(4, j) dDispatch(dGenStack(0, j ), i, dGenStack(2, j) 2, 3) = dHourlyCap dGenStack(3, j) dTemp = dTemp dHourlyCap Else dDispatch(dGenStack(0, j), i, dGenStack(2, j) 2, 0) = dTemp dDispatch(dGenStack(0, j) i, dGenStack(2, j) 2, 1) = dTemp dGenStack(1, j) dDispatch(dGenStack(0, j), i, dGenStack(2, j) 2, 2) = dTemp dGenStack(4, j) dDispatch(dGenStack(0, j), i, dGenStack(2, j) 2, 3) = dTemp dGenStack(3, j) dTemp = 0 sOutstring = dCarbonPrice & "," & dtCurrDate & "," & i & "," & dGenStack(1, j) Print #3, sOutstring End If j = j + 1 Loop For j = 0 To lGenCnt Sheet4.Cells(i + 1, j + 1) = dDispatch(j, i) Next j Next i 'Aggregate monthly dispatch results 'dDispatchSum elements are: Fuel types Element (MW,VC,CO2,MMBtu) GenID For i = 0 To lGenCnt iDEPTypeIdx = DEPTypeLookup(aGenUnits(i).DEPTypeCode) For j = 0 To iHours For k = 0 To iNumFuelTypes For m = 0 To 3 dDispatchSum(k, m, i) = dDispatc hSum(k, m, i) + dDispatch(i, j, k, m) Select Case aGenUnits(i).AvailDate Case Is < #1/1/2009#
104 dAnnualSumVintage(0, k, m) = dAnnualSumVintage(0, k, m) + dDispatch(i, j, k, m) Case Else dAnnualSumVintage(1, k, m) = dAnnualSumVintage(1, k, m) + dDispatch(i, j, k, m) End Select dAnnualSumType(iDEPTypeIdx, m) = dAnnualSumType(iDEPTypeIdx, m) + dDispatch(i, j, k, m) Next m dAggregateFuelMMBtu(k) = dAggregateFuelMMBtu(k) + dDispatch(i, j, k, 3) Next k Next j Next i 'Calculate Aggregate Fuel Costs For i = 0 To iNumFuelTypes dTemp = Sheet2.Cells(DateDiff("m", Sheet2.Cells(2, 1), dtCurrDate) + 2, GetFuelPriceIndex(Sheet5.Cells(i + 2, 7))) 'Fuel Price dAggregateFuelCosts(i, 0) = dAggregateFuelCosts(i, 0) + dAggregateFuelMMBtu(i) dT emp Next i ReDim dAggregateFuelMMBtu(iNumFuelTypes) 'Write annual results If Month(dtCurrDate) = 12 Or iDateCnt = iDateIdx Then 'Write individual generator records For i = 0 To lGenCnt dTemp = (5 aGenUnits(i).SumCap + 7 aGenUnits(i).WinCap) / 12 dTempMWh = 0 sOutstring = Year(dtCurrDate) sOutstring = sOutstring & "," & dCarbonPrice sOutstring = sOutstring & "," & i For j = 0 To 3 For k = 0 To iNumFuelTypes sOutstring = sOutstring & "," & dDispatchSum(k, j, i) If j = 0 Then dTempMWh = dTempMWh + dDispatchSum(k, j, i) dAnnualSum(k, j) = dAnnualSum(k, j) + dDispatchSum(k, j, i) Next k Next j 'Write output line for individual generator sOutstring = sOutstring & "," & dTempMWh / (dTemp 8760) Print #1, sOutstring Next i 'Write annual summaries and reset summary arrays 'Write Output detail by fuel type and DEP type
105 lAggSectionOffset = CLng((dCarbonHigh dCarbonLow) / dCarbonInt) + 1 For i = 0 To 3 Sheet4.Cells(lOutRow, 1) = dCarbonPrice Sheet4.Cells(lOutRow, 2) = Year(dtCurrDate) Sheet10.Cells(lOutRow, 1) = dCarbonPrice Sheet10.Cells(lOutRow, 2) = Year(dtCur rDate) Sheet13.Cells(lOutRow + 1, 1) = dCarbonPrice Sheet13.Cells(lOutRow + 1, 2) = Year(dtCurrDate) Select Case i Case 0 Sheet4.Cells(lOutRow, 3) = "MWh" Sheet10.Cells(lOutRow, 3) = "MWh" Sheet13.Cells(lOutRow + 1, 3) = "MWh" Case 1 Sheet4.Cells(lOutRow, 3) = "Variable Costs" Sheet10.Cells(lOutRow, 3) = "Variable Costs" Sheet13.Cells(lOutRow + 1, 3) = "Variable Costs" Case 2 Sheet4.Cells(lOutRow, 3) = "Emissions" Sheet10.Cells(lOutRow, 3) = "Emissions" Sheet13.Cells(lOutRow + 1, 3) = "E missions" Case 3 Sheet4.Cells(lOutRow, 3) = "Fuel Burn" Sheet10.Cells(lOutRow, 3) = "Fuel Burn" Sheet13.Cells(lOutRow + 1, 3) = "Fuel Burn" End Select For j = 0 To iNumFuelTypes Sheet4.Cells(lOutRow, 4 + j) = dAnnualSum(j, i) Sheet13.Cells(lOutRow + 1, 4 + j) = dAnnualSumVintage(0, j, i) Sheet13.Cells(lOutRow + 1, 5 + iNumFuelTyp es + j) = dAnnualSumVintage(1, j, i) dAggregateSum(i) = dAggregateSum(i) + dAnnualSum(j, i) Next j For j = 0 To UBound(dAnnualSumType, 1) Sheet10.Cells(lOutRow, 4 + j) = dAnnualSumType (j, i) Next j lOutRow = lOutRow + 1 Next i 'Write Aggregated Annual Output If lAggOutRow = 2 Then 'Set section headers Sheet9.Cells(1, 1) = "MWh" Sheet9.Cells(4, 1) = "Variable Costs" Sheet9.Cells(lAggSectionOffset + 6, 1) = "Emissions" Sheet9.Cells(lAggSectionOffset 2 + 8, 1) = "Average Variable Costs" End If
106 Sheet9.Cells(1, Year(dtCurrDat e) Year(dtBeginDate) + 2) = Year(dtCurrDate) Sheet9.Cells(2, Year(dtCurrDate) Year(dtBeginDate) + 2) = dAggregateSum(0) Sheet9.Cells(4, Year(dtCurrDate) Year(dtBeginDate) + 2) = Year(dtCurrDate) Sheet9 .Cells(lAggOutRow + 3, 1) = dCarbonPrice Sheet9.Cells(lAggOutRow + 3, Year(dtCurrDate) Year(dtBeginDate) + 2) = dAggregateSum(1) Sheet9.Cells(lAggSectionOffset + 6, Year(dtCurrDate) Year(dtBeginDate) + 2) = Year(dtCurrDate) Sheet9.Cells(lAggSectionOffset + lAggOutRow + 5, 1) = dCarbonPrice Sheet9.Cells(lAggSectionOffset + lAggOutRow + 5, Year(dtCurrDate) Year(dtBeginDate) + 2) = dAggregateSum(2) Sheet9.Cells( lAggSectionOffset 2 + 8, Year(dtCurrDate) Year(dtBeginDate) + 2) = Year(dtCurrDate) Sheet9.Cells(lAggSectionOffset 2 + lAggOutRow + 7, 1) = dCarbonPrice Sheet9.Cells(lAggSectionOffset 2 + lAggOutRow + 7, Year(dtCurrDate) Y ear(dtBeginDate) + 2) = dAggregateSum(1) / dAggregateSum(0) 'Construct and Write REMI file For i = 0 To lGenCnt For j = 1 To aGenUnits(i).NumFuels iTemp = aGenUnits(i).Fuel(j 1).TypeIndex 2 dAggregateFuelCosts(iTemp, 1) = dAggregateFuelCosts(iTemp, 1) + aGenUnits(i).VOM dDispatchSum(iTemp, 0, i) dAggregateFuelCosts(iTemp, 2) = dAggregateFuelCosts(iTemp, 2) + aGenUnits(i).Fuel(j 1).Adder dDispatchSum (iTemp, 3, i) dAggregateFuelCosts(iTemp, 3) = dAggregateFuelCosts(iTemp, 3) + dDispatchSum(iTemp, 2, i) dCarbonPrice Next j Next i ReDim dREMIFuelData(2, 4) For i = 0 To iNumFuelTypes Select Case Sheet5.Cells(i + 2, 7) Case "NG" dREMIFuelData(0, 0) = dREMIFuelData(0, 0) + dAnnualSum(i, 3) For j = 0 To 3 dREMIFuelData(0, j + 1) = dREMIFuelData(0, j + 1) + dAggregateFuelCosts(i, j) Next j Case "RFO", "DFO"
107 dREMIFuelData(1, 0) = dREMIFuelData(1, 0) + dAnnualSum(i, 3) For j = 0 To 3 d REMIFuelData(1, j + 1) = dREMIFuelData(1, j + 1) + dAggregateFuelCosts(i, j) Next j Case Else dREMIFuelData(2, 0) = dREMIFuelData(2, 0) + dAnnualSum(i, 3) For j = 0 To 3 dREMIFuelData(2, j + 1) = dREMIFuelData(2, j + 1) + dAggregateFuelCosts(i, j) Next j End Select Next i sOutstring = "Year, Carbon Price, Fuel Code, MMBtu, Fuel Costs, VOM, Adders, Emissions Cost" Print #4, sOutstring For i = 0 To iNumFuelTypes sOutstring = Year(dtCurrDate) & "," & dCarbonPrice & "," sOutstring = sOutstring & Sheet5.Cells(i + 2, 7) & "," sOutstring = sOutstring & dAnnualSum(i, 3) & "," sOutstring = sOutstring & dAggregateFuelCosts(i, 0) & "," sOutstring = sOutstring & dAggregateFuelCosts(i, 1) & "," sOutstring = sOutstring & dAggregateFuelCosts(i, 2) & "," sOutstring = sOutstring & dAggregateFuelCosts(i, 3) Print #4, sOutstring Next i sOutstring = "Year, Carbon Price, Fuel Class, MMBtu, Fuel Costs, VOM, Adder s, Emissions Cost" Print #4, sOutstring For i = 0 To 2 sOutstring = Year(dtCurrDate) & "," & dCarbonPrice & "," Select Case i Case 0 sOutstring = sOutstring & "NG," Case 1 sOutstring = sOutstring & "Oil," Case 2 sOutstring = sOutstring & "Other(Electricity)," End Select sOutstring = sOutstring & dREMIFuelData(i, 0) & "," sOutstring = sOutstring & dREMIFuelData(i, 1) & "," sOutstring = sOutstring & dREMIFuelData(i, 2) & "," sOutstring = sOutstring & dREMIFuelData(i, 3) & ","
108 sOutstring = sOutstring & dRE MIFuelData(i, 4) Print #4, sOutstring Next i sOutstring = Year(dtCurrDate) & "," & dCarbonPrice & "," sOutstring = sOutstring & "All," sOutstring = sOutstring & dREMIFuelData(0, 0) + dREMIFuel Data(1, 0) + dREMIFuelData(2, 0) & "," sOutstring = sOutstring & dREMIFuelData(0, 1) + dREMIFuelData(1, 1) + dREMIFuelData(2, 1) & "," sOutstring = sOutstring & dREMIFuelData(0, 2) + dREMIFuelData(1, 2) + dREMIFuelData(2, 2) & "," sOutstring = sOutstring & dREMIFuelData(0, 3) + dREMIFuelData(1, 3) + dREMIFuelData(2, 3) & "," sOutstring = sOutstring & dREMIFuelData(0, 4) + dREMIFuelData(1, 4) + dREMIFuelData(2, 4) Print #4, sOutstring sOutstring = "Year, Carbon Price, Consumer Price" Print #4, sOutstring sOutstring = Year(dtCurrDate) & "," & dCarbonPrice & "," & dAggregateSum(1) / dAggregateSum(0) + dFixedCostAdjust Print #4, sOutstring sOutstring = "Year, Carbon Price, Production Cost" Print #4, sOutstring sOutstring = Year(dtCurrDate) & "," & dCarbonPrice & "," & dAggregateSum(1) Print #4, sOutstring sOutstring = "Year, Carbon Price, Exogenous Final Demand" Print #4, sOutstring sOutstring = Year(dtCurrDate) & "," & dCarbonPrice & "," & dAggregateSum(0) Print #4, sOutstring 'Check to build new generatio n If dAggregateSum(1) / dAggregateSum(0) + dFixedCostAdjust > dNewGenCost Then If 1 > 2 Then If Not Range("CheckBoxStatus") Then j = 0 Else j = UBound(NewPlant ) End If For i = 0 To j lGenCnt = lGenCnt + 1 ReDim Preserve aGenUnits(lGenCnt) ReDim Preserve dDispatchSum(iNumFuelTypes, 3, lGenCnt) aGenUnits (lGenCnt) = NewPlant(i) aGenUnits(lGenCnt).AvailDate = dtCurrDate aGenUnits(lGenCnt).RetireDate = DateAdd("yyyy", 50, dtCurrDate)
109 dFixedGenCost = dFixedGenCost + aGenUnits(lGenCnt).FixedCost sOutstring = dCarbonPrice & "," & dtCurrDate & "," & aGenUnits(lGenCnt).SumCap & "," & aGenUnits(lGenCnt).PlantName Print #2, sOutstring Next i End If 'Reset Arrays ReDim dDispatchSum(iNumFuelTypes, 3, lGenCnt) ReDim dAnnualSum(iNumFuelTypes, 3) ReDim dAnnualSumVintage(1, iNumFuelTypes, 3) ReDim dAnnualSumType(iNumDEPTypes, 3) ReDim dAggregateFuelCosts(iNumFuelTypes, 3) ReDim dAggregateSum(3) End If dElapsedRunTime = dElapsedRunTime + 1 Next iDateCnt lAggOutRow = lAggOutRow + 1 Next dCarbonPrice Close #1 Close #2 Close #3 Close #4 Application.Stat usBar = False Application.DisplayStatusBar = oldStatusBar Application.ScreenUpdating = True End Sub
110 Subroutine BubbleSortGenStack The generation stack is sorted by costs for each hour of the simulation by this routine. Because the number of items is r elatively small (roughly 300 400), there is little computational efficiency lost in using a straightforward Bubble sort for the procedure. Sub BubbleSortGenStack(lSortIdx As Long) 'Bubble sorts array in ascending order on index lSortIdx Dim lArrSize As Long Dim lArrDepth As Long Dim i As Long Dim j As Long Dim bDone As Boolean Dim dTemp As Double lArrSize = UBound(dGenStack, 2) lArrDepth = UBound(dGenStack, 1) Do bDone = True For i = 0 To lArrSize 1 If dGenStack(lSortIdx, i) > dGenSta ck(lSortIdx, i + 1) Then 'Bubblesort in ascending order For j = 0 To lArrDepth dTemp = dGenStack(j, i) dGenStack(j, i) = dGenStack(j, i + 1) dGenStack(j, i + 1) = dTemp Next j bDone = False End If Next i Loop Until bDone End Sub
111 Lookup Functions The code makes extensive use of look up functions to turn text input into numerical indices for processing. These routines perform look ups for indices relating to the type of generating unit, the type of fuel, the price of fuel, or the type of electricity generating plant defined by the Florida Department of Environmental Protection. Public Function PlantTypeLookup(sPrimeMover As String, sFuelType As String) As Int eger Dim i As Long i = 1 Do If Sheet5.Cells(i + 1, 1) = sPrimeMover And Sheet5.Cells(i + 1, 2) = sFuelType Then Exit Do i = i + 1 Loop PlantTypeLookup = i End Function Public Function GetFuelIndex(sFuelType As String) As Long 'Gets the fuel type index from the Inputs sheet Dim i As Long i = 2 Do If Sheet5.Cells(i, 7) = sFuelType Then Exit Do i = i + 1 Loop GetFuelIndex = i End Function Public Function GetFuelPriceIndex(sFuelType As String) As Long Dim i As Long Dim bMatch As Boolean i = 2
112 Do If Sheet2.Cells(1, i) = sFuelType Then bMatch = True Exit Do End If i = i + 1 If IsEmpty(Sheet2.Cells(1, i)) Then Exit Do Loop If bMatch Then GetFuelPriceIndex = i Else GetFuelPriceIndex = 0 End If End Function Public Function DEPTypeLookup(sDEPType As String) As Integer Select Case sDEPType Case "Cogen" DEPTypeLookup = 0 Case "WTE" DEPTypeLookup = 1 Case "LF" DEPTypeLookup = 2 Case "IPP" DEPTypeLookup = 3 Case "FMPA" DEPTypeL ookup = 4 Case "CO OP" DEPTypeLookup = 5 Case "FPL" DEPTypeLookup = 6 Case "PE" DEPTypeLookup = 7 Case "GP" DEPTypeLookup = 8 Case "TECO" DEPTypeLookup = 9 Case "WTP" DEPTypeLookup = 10 Case "Hydro" DEPTypeLookup = 11 Case "New" DEPTypeLookup = 12 End Select End Function
113 Sub SetGenStack This subroutine is run once for each month of the simulation. Since fuel prices only change monthly, the fuel that each unit utilizes will stay the same for the entire simulation month. Th is routine checks in service and out of service dates to determine which units are available for the mo nth, and selects the cheapest fuel alternative. Sub SetGenStack(dtCurrDate, dCarbonPrice) 'Sub sets the Gen stack for the particular month, utilizing t he cheapest fuel alternative for each unit 'GenStack description (4xN array): 'Generator ID is element 0 'Generation cost/MWh is element 1 'Generator fuel type index is element 2 'Generator fuel burn/MWh is element 3 'Generator emissions/MWh is element 4 Dim i As Long Dim j As Long Dim k As Long Dim iDateIdx As Integer Dim dFuelCost As Double Dim dMinCost As Double Dim iMinFuel As Integer iDateIdx = DateDiff("m", Sheet2.Cells(2, 1), dtCurrDate) j = 0 For i = 0 To lGenCnt iMinFuel = 0 If dtCurrDate >= aGenUnits(i).AvailDate And dtCurrDate < aGenUnits(i).RetireDate Then ReDim Preserve dGenStack(4, j) dGenStack(0, j) = i If aGenUnits(i).NumFuels > 1 Then 'Loop through fuel alternatives and pick cheapest dMinCost = 1E+30 For k = 1 To aGenUnits(i).NumFuels If aGenUnits(i).Fuel(k 1).PriceIndex = 0 Then dFuelCost = 0 Else dFuelCost = Sheet2.Cells(iDateIdx + 2, a GenUnits(i).Fuel(k 1).PriceIndex) End If
114 dFuelCost = dFuelCost + aGenUnits(i).Fuel(k 1).Adder dGenStack(2, j) = aGenUnits(i).Fuel(k 1).TypeIndex dGenStack(3, j) = aGenUnits(i).HeatRate dGenStack(4, j) = aGenUnits(i).DEPCapUnit aGenUnits(i).HeatRate Sheet5.Cells(dGenStack(2, j), 8) aGenUnits(i).CCS / 1000 dGenStack(1, j) = (aGenUnits(i).HeatRate dFuelCost) + aGenUnits(i).VOM + (dGenStack(4, j) dCar bonPrice) If dGenStack(1, j) < dMinCost Then dMinCost = dGenStack(1, j) iMinFuel = k 1 End If Next k If aGenUnits(i).Fuel(iMinFuel).PriceIndex = 0 Then dFuelCost = 0 Else dFuelCost = Sheet2.Cells(iDateIdx + 2, aGenUnits(i).Fuel(iMinFuel).PriceIndex) End If dFuelCost = dFuelCost + aGenUnits(i).Fuel(iMinFuel).Adder dGenStack(2, j) = aGenUnits(i).Fuel(iMinFuel).TypeIndex dGenStack(3, j) = aGenUnits(i).HeatRate dGenStack(4, j) = aGenUnits(i).DEPCapUnit aGenUnits(i).HeatRate Sheet5.Cells(dGenStack(2, j), 8) / 1000 dGenStack(1, j) = aGenUnits(i ).HeatRate dFuelCost + aGenUnits(i).VOM + dGenStack(4, j) dCarbonPrice Else If aGenUnits(i).Fuel(0).PriceIndex = 0 Then dFuelCost = 0 Else dFuelCost = Sheet2.Cells(iDateIdx + 2, aGenUnits( i).Fuel(0).PriceIndex) End If dFuelCost = dFuelCost + aGenUnits(i).Fuel(0).Adder dGenStack(2, j) = aGenUnits(i).Fuel(0).TypeIndex dGenStack(3, j) = aGenUnits(i).HeatRate dGenStack(4, j) = aGenUnit s(i).DEPCapUnit aGenUnits(i).HeatRate Sheet5.Cells(dGenStack(2, j), 8) / 1000 dGenStack(1, j) = aGenUnits(i).HeatRate dFuelCost + aGenUnits(i).VOM + dGenStack(4, j) dCarbonPrice End If j = j + 1 End If Next i End Sub Sub CompareOutput of back casting.
1 15 Sub CompareOutput() Dim iSrcRow As Integer Dim iTgtRow As Integer Dim iOutRow As Integer Dim iTemp As Integer Dim bMatch As Boolean Dim sPlant As String Dim dGen As Double iSrcRow = 2 iOutRow = 3 Do Until IsEmpty(Sheet4.Cells(iSrcRow, 2)) sPlant = Sheet4.Cells(iSrcRow, 2) 'Check to see if plant has already been counted iTemp = 3 bMatch = False Do Until IsEm pty(Sheet11.Cells(iTemp, 1)) If sPlant = Sheet11.Cells(iTemp, 1) Then bMatch = True Exit Do End If iTemp = iTemp + 1 Loop If Not bMatch Then dGen = 0 iTemp = 2 Sheet11.Cells( iOutRow, 1) = sPlant Do Until IsEmpty(Sheet4.Cells(iTemp, 2)) If sPlant = Sheet4.Cells(iTemp, 2) Then dGen = dGen + Sheet4.Cells(iTemp, 6) iTemp = iTemp + 1 Loop Sheet11.Cells(iOutRow, 2) = dGen dGen = 0 iTemp = 2 Do Until IsEmpty(Sheet10.Cells(iTemp, 3)) If sPlant = Sheet10.Cells(iTemp, 3) Then dGen = dGen + Sheet10.Cells(iTemp, 5) iTemp = iTemp + 1 Loop Sheet11.Cells(iOutRow, 3) = dGen i OutRow = iOutRow + 1 End If iSrcRow = iSrcRow + 1 Loop End Sub
116 Subroutine GrowLoads This set of routines utilizes a base hourly electric load shape with particular peak demand and total energy demand characteristics, and reshapes it to be consistent with the peak demand and total energy demand forecast used in the simulation, the target c haracteristics. The routine first shifts the base shape so that its weekdays and weekends conform to the target year. The routine then searches for the shape parameter that satisfies: Note that at the peak hour of the year, the exponential term is equal to 1 and the peak hour simply grows at the same rate as demand. Every other hour of the year is scaled, so that peak demand can grow at 5%, say, and total energy demand can grow at 10%. The load shape will flatten itself out in this case. If total energy grows at less than demand, the load shape will stretch. Sub GrowLoads() 'Only needs to be rerun after changing load forecast 'Deleting all load data but the base year is not required before running but makes the output cleaner Dim iBaseYear As Integer Dim iTargetYear As Integer Dim i As Long Dim j As Long Dim dBaseLoad() As Double Dim dLoadShape() As Doub le Dim lCount As Long Dim iLength As Integer Dim iCount As Integer Dim iDay As Integer Dim iRow As Integer Dim dDemand As Double Dim dEnergy As Double Dim dDelta As Double
117 'Read and format the base shape lCount = 1 iBaseYear = Year(Sheet3.Cells(lCount, 1)) lCount = DateSerial(iBaseYear + 1, 1, 1) DateSerial(iBaseYear, 1, 1) iLength = lCount 24 1 ReDim dBaseLoad(iLength) iCount = 0 For i = 1 To lCount For j = 2 To 25 dBaseLoad(iCount) = Sheet3.Cells(i, j) iCount = iCount + 1 Next j Next i 'Read and output grown loads iRow = 12 lCount = lCount + 1 Do Until IsEmpty(Sheet8.Cells(iRow, 1)) iTargetYear = Sheet8.Cells(iRow, 1) Call ShiftLoadShape(dBaseLoad, dLoadShape, iBaseYear, iTargetYear) dDemand = Sheet8.Cells(iRow, 2) dEnergy = Sheet8.Cells(iRow, 3) 1000 Call GetShapeParm(dLoadShape, dDemand, dEnergy) iDay = (UBound(dLoadShape) + 1) / 24 iCount = 0 For i = 0 To iDay 1 Sheet3.Cells(lCount, 1) = DateSerial (iTargetYear, 1, 1) + i For j = 2 To 25 Sheet3.Cells(lCount, j) = dLoadShape(iCount) iCount = iCount + 1 Next j lCount = lCount + 1 Next i iRow = iRow + 1 Loop End Sub Sub GetShapeParm(dLoadShape() As Double, dTgtDem As Double, dTgtEng As Double) Dim dDelta As Double Dim iPower As Integer Dim i As Long Dim j As Long Dim iLength As Long Dim dLoadMax As Double Const ConvCrit = 1
118 iLength = UBound(dLoadShape) For i = 0 To iLength If dLoadShape (i) > dLoadMax Then dLoadMax = dLoadShape(i) Next i dDelta = 10 iPower = 2 Do While Abs(dTgtEng TotalEnergy(dLoadShape(), dLoadMax, dTgtDem, dDelta)) > ConvCrit iPower = iPower 1 dDelta = MinValDeviation(dLoadShape(), dLoadMax, dTgtDem, dTgtEng, dDelta, iPower) Loop For i = 0 To iLength dLoadShape(i) = dLoadShape(i) (dTgtDem / dLoadMax) Exp(dDelta Log(dLoadShape(i) / dLoadMax)) Next i End Sub Function MinValDeviation(dLoadSh ape() As Double, ByVal dLoadMax As Double, ByVal dTgtDem As Double, ByVal dTgtEng, ByVal dDelta As Double, iPower As Integer) As Double Dim dInitDev As Double Dim dLastDev As Double Dim dDev As Double Dim dWorkDelta As Double dInitDev = dTgtEng TotalEn ergy(dLoadShape(), dLoadMax, dTgtDem, dDelta) dDev = dInitDev dWorkDelta = dDelta Do While Sgn(dInitDev) = Sgn(dDev) dLastDev = dDev dWorkDelta = dWorkDelta Sgn(dInitDev) 10 ^ iPower dDev = dTgtEng TotalEnergy(dLoadShape(), dLoadMax, dTg tDem, dWorkDelta) Loop If Abs(dLastDev) < Abs(dDev) Then MinValDeviation = dWorkDelta + Sgn(dInitDev) 10 ^ iPower Else MinValDeviation = dWorkDelta End If End Function
119 Function TotalEnergy(dLoadShape() As Double, ByVal dLoadMax As Double, ByVal dTgtDem As Double, ByVal dDelta As Double) As Double Dim i As Integer Dim dTempSum As Double Dim iLength As Integer iLength = UBound(dLoadShape) For i = 0 To iLength dTempSum = dTempSum + dLoadShape(i) (dTgtDem / dLoadMax) Exp(dDelta Log (dLoadShape(i) / dLoadMax)) Next i TotalEnergy = dTempSum End Function Sub ShiftLoadShape(dBaseLoadShape() As Double, dTargetLoadShape() As Double, iBaseYear As Integer, iTargetYear As Integer) Dim dtStartBase As Date Dim dtStartTarget As Date Dim iDat eOffset As Integer Dim iBaseLength As Integer Dim iTargetLength As Integer Dim i As Long Dim iCount As Integer dtStartBase = DateSerial(iBaseYear, 1, 1) dtStartTarget = DateSerial(iTargetYear, 1, 1) iBaseLength = UBound(dBaseLoadShape) iTargetLength = (DateSerial(iTargetYear + 1, 1, 1) DateSerial(iTargetYear, 1, 1)) 24 1 ReDim dTargetLoadShape(iTargetLength) iDateOffset = Weekday(dtStartTarget) Weekday(dtStartBase) If iDateOffset > 3 Then iDateOffset = iDateOffset 7 If iDateOffset < 3 Th en iDateOffset = iDateOffset + 7 Select Case iDateOffset Case 0 iCount = 0 Case Is > 0 iCount = iDateOffset 24 Case Is < 0 iCount = iBaseLength + 1 iDateOffset 24 End Select For i = 0 To iTargetLength
120 If iCount > iBaseLength Then iC ount = iCount iBaseLength 1 dTargetLoadShape(i) = dBaseLoadShape(iCount) iCount = iCount + 1 Next i End Sub Sub test() Dim i As Long Dim j As Long Dim iLoadShapeIdx As Integer Dim dLoadShape(8759) As Double Dim dTemp As Double Dim dDemand As Double Dim dEnergy As Double Dim bEcon As Boolean bEcon = Not Range("CheckBoxStatus") dDemand = 49391 1.1 dEnergy = 246492002 1.05 iLoadShapeIdx = 0 For i = 1 To 365 For j = 2 To 25 dLoadShape(iLoadShapeIdx) = Sheet3.Cells(i, j) iLoadShapeIdx = iLoadShapeIdx + 1 Next j Next i End Sub
121 Subroutine DefineNewPlant This routine defines the new plant that will be built by the simulation, if a new plant is necessary for reliability purposes. The new plant can either be determined endogenously, by least cost, or exogenously based on a user supplied configuration. Private Sub DefineNewPlant(NewPlant() As GenPlant, dCarbonPrice As Double) Dim dLeastCost As Double Dim lLeastCostUnitID As Long Dim lFuelIndex As Long Dim dEmi ssionsCost As Double Dim dTotalCost As Double Dim bEconomic As Boolean Dim bDone As Boolean Dim lArrSize As Long Dim tmpPlant As GenPlant Dim i As Long Dim j As Long bEconomic = Not Range("CheckBoxStatus") If bEconomic Then j = 0 dLeastCost = 1E+30 i = 2 Do Until IsEmpty(Sheet5.Cells(i, 13)) ReDim Preserve NewPlant(j) lFuelIndex = GetFuelIndex(Sheet5.Cells(i, 20)) dEmissionsCost = Sheet5.Cells(i, 21) Sheet5.Cells(lFuelIndex, 8) dCarbonPrice / 1000 dTotalCost = dEmissionsCost + Sheet5.Cells(i, 19) NewPlant(j).AvailPct = Sheet5.Cells(i, 14) NewPlant(j).PlantName = Sheet5.Cells(i, 13) NewPlant(j).SumCap = 1000 NewPlant(j).WinCap = 1000 NewPlant(j).DEPCapUnit = 1 NewPlant(j).DEPTypeCode = "New" NewPlant(j).HeatRate = Sheet5.Cells(i, 21) NewPlant(j).NumFuels = 1 ReDim NewPlant(j).Fuel(0) NewPlant(j).Fuel(0).Type = Sheet5.Cells(i, 20) NewPlant(j).Fuel(0).TypeIndex = GetFuelIndex(NewPlant(j).Fuel(0).Type) NewPlant(j).Fuel(0).PriceIndex = 0 NewPlant(j).VOM = dTotalCost Sheet5.Cells(i, 15)
122 NewPlant(j).LevelCost = dTotalCost NewPlant(j).CCS = Sheet5.Cells(i, 24) NewPlant(j).Fixe dCost = Sheet5.Cells(i, 25) NewPlant(j).SumCap i = i + 1 j = j + 1 Loop lArrSize = UBound(NewPlant) Do bDone = True For i = 0 To lArrSize 1 If NewPlant(i).LevelCost > NewPlant(i + 1).LevelCost Then 'Bubblesort in ascending order tmpPlant = NewPlant(i) NewPlant(i) = NewPlant(i + 1) NewPlant(i + 1) = tmpPlant bDone = False End If Next i Loop Until bDone Else i = 2 j = 0 Do Until IsEmpty(Sheet5.Cells(i, 13)) If Sheet12.Cells(i + 3, 2) > 0 Then ReDim Preserve NewPlant(j) lFuelIndex = GetFuelIndex(Sheet5.Cells(i, 20)) dEmissionsCost = S heet5.Cells(i, 21) Sheet5.Cells(lFuelIndex, 8) dCarbonPrice / 1000 dTotalCost = dEmissionsCost + Sheet5.Cells(i, 19) NewPlant(j).AvailPct = Sheet5.Cells(i, 14) NewPlant(j).PlantName = Sheet5.Cells(i, 13) NewPlant(j).SumCap = 1000 Sheet12.Cells(i + 3, 2) NewPlant(j).WinCap = 1000 Sheet12.Cells(i + 3, 2) NewPlant(j).DEPCapUnit = 1 NewPlant(j).DEPTypeCode = "New" NewPlant(j).HeatRate = Sheet5.Cells(i, 21) NewPlant(j).NumFuels = 1 ReDim NewPlant(j).Fuel(0) NewPlant(j).Fuel(0).Type = Sheet5.Cells(i, 20) NewPlant(j).Fuel(0).TypeIndex = GetFuelIndex(NewPlant(j).Fuel(0).Type) NewPlant(j).Fuel(0).PriceIn dex = 0 NewPlant(j).VOM = dTotalCost Sheet5.Cells(i, 15) NewPlant(j).LevelCost = dTotalCost NewPlant(j).CCS = Sheet5.Cells(i, 24) NewPlant(j).FixedCost = Sheet5.Cells(i, 25) NewPlant(j).SumCap
123 j = j + 1 End If i = i + 1 Loop End If End Sub Public Function PriceNewPlant(NewPlant() As GenPlant) Dim i As Long Dim j As Long Dim lArrSize As Long Dim dTemp As Double lArrSize = UBound(NewPlant) If Not Range( "CheckBoxStatus") Then j = 0 Else j = lArrSize End If For i = 0 To j dTemp = dTemp + NewPlant(i).SumCap / 1000 NewPlant(i).LevelCost Next i PriceNewPlant = dTemp End Function
124 Subroutine AddGenStack The generation stack utilized in the program is a complicated data element, and this routine is used to add generating units to the stack. It is called whenever a new plant is built endogenously Sub AddGenStack(dGenStack() As Double, NewPlant As GenPlant, PlantIndex As Long) Dim lArrSize As Long lArrSize = UBound(dGenStack, 2) + 1 ReDim Preserve dGenStack(4, lArrSize) dGenStack(0, lArrSize) = PlantIndex dGenStack(2, lArrSize) = NewPlant.Fuel(0).TypeIndex dGenStack(3, lArrSize) = NewPlant.HeatRate dGenStack(4, lArrSize) = NewPlant.DEPCapUnit NewPlant.HeatRate Sheet5.Cells(dGenStack(2, lArrSize), 8) / 1000 dGenStack(1, lArrSize) = NewPlant.LevelCost End Sub Sub RetrofitCCS(aGenUnits() As GenPlant, dCCSCost As Double, dCCSPct As Double, dCarbonPrice As Double) 'Process es the generation stack End Sub
125 Output Routines The code utilizes a number of routines that write output to flat files for post processing and diagnostic purposes. These routines write the hourly loads used in the simulation, the fuel prices, and the characteristics of the generating units that exist at the beginning of the simulation. Sub WriteInputLoad(sDir As String) Dim sOutfile As String Dim sLine As String Dim i As Long Dim j As Long sOutfile = "Input_Load.csv" Open sDir & sOutfile For Output As #1 i = 1 Do Until IsEmpty(Sheet3.Cells(i, 1)) sLine = Sheet3.Cells(i, 1) For j = 2 To 25 sLine = sLine & "," & Sheet3.Cells(i, j) Next j Print #1, sLine i = i + 1 Loop Close #1 End Sub Sub WriteInputFuelPrice(sDir As St ring) Dim sOutfile As String Dim sLine As String Dim i As Long Dim j As Long sOutfile = "Input_FuelPrice.csv" Open sDir & sOutfile For Output As #1 i = 1 Do Until IsEmpty(Sheet2.Cells(i, 1)) sLine = Sheet2.Cells(i, 1) For j = 2 To 28
126 sLine = sLine & "," & Sheet2.Cells(i, j) Next j Print #1, sLine i = i + 1 Loop Close #1 End Sub Sub WriteInputGenUnits(sDir As String) Dim sOutfile As String Dim sLine As String Dim i As Long Dim j As Long sOutfile = "Input_GenUnit s.csv" Open sDir & sOutfile For Output As #1 i = 1 Do Until IsEmpty(Sheet1.Cells(i, 1)) sLine = Sheet1.Cells(i, 1) For j = 2 To 23 sLine = sLine & "," & Sheet1.Cells(i, j) Next j Print #1, sLine i = i + 1 Loop Close #1 End Sub
127 LIST OF REFERENCES Ackerman Frank and Elizabeth A. Stanton. 2012 Climate Risks and Carbon Prices: Revising the Social Cost of Carbon Economics: The Open Access, Open Assessment E Journal Vol. 6, 2012 10. http://dx.doi.org/10.5018/economics ejournal.ja.2012 10 A Decomposition Analysis Using FUN Climatic Change 117 (2013): 515 530 Communications of the ACM Vol. 41, No. 8 (1998): 35 42. Price Elasticity of Renewable Energy Laboratory Blumsack, Seth. 2007. "Measuring the Costs and Benefits of Regional Electric Grid Integration," Energy Law Journal 28: 147 184 Rice University Chandley, John and William W. Hogan. Coase, Ronald H. Economica New Series, Vol. 4, No. 16. (Nov., 1937) : 386 405 Journal of Law and Economics Vol. 3 (Oct., 1960) : 1 44. ation, and Efficiency: California at Berkeley Analysis of Potential Impacts of CO 2 Emissions Limits on Electric Power Co sts in the ERCOT Region Benefit Cost Studies: Toward More Comprehensive Assessments of FERC Electricity icity Delivery and Energy Reliability, U.S. Department of Energy.
128 Organization Science Vol. 23 No. 5 (September October 2012) : 1264 1281 Fa American Economic Review Vol. 97, No. 4 (2007) : 1250 1277 Federal Energy Regulatory Commission, Form 1 (Annual Report of Major Electric Utility) Federal Energy Regulatory Commission, Form 714 (Annual Electric Control and Planning Area Report) Florida Reliability Coordinating Council 2012 Regional Load and Resou rce Plan To Business E Commerce NBER Working Paper 8017. 2 Abatement in the Presence of Journal of Environmental Economics and Management 39 (1) : 1 38 ing Paper Journal of Political Economy 94(4) : 69 1 719. Econometrica Vol. 47, No. 1 (1979), pp. 153 161. Independent System Operator: Avoiding the Separat The Electricity Journal December 1995 : 26 37. Institute for Econom ic Research, Helsinki Cost of Carbon for Regulatory Impact Analysis Support Document February 2010
129 IPCC, 2007: Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, M.L. Parry, O.F. Canziani, J.P. Palutikof, P.J. van der Linden and C.E. Hanson, Eds., Cam bridge University Press, Cambridge, UK Utilities Policy 13(2005): 95 115 Joskow, Paul L Center for Energy and Env ironmental Policy Research Working Paper 05 012, August 2005. Ma rginal Effects of the Price for Carbon Dioxide: Quantifying the Effects on the Market for Electric Generation in Florida The Electricity Journal Vol. 23 No. 4 (2010): 73 78 Restructuring the U.S. Electric Power Sector: A Review of Recent Kwoka, John, Michael Pollitt, and Sanem Sergici. Journal of Regulatory Economics 38(2010) : 86 109 Journal of Financial Intermediation 5 (1996) : 25 5 283. Milgrom, Paul and John Roberts. 1992. Economics, Organization, and Management Englewood Cliffs, NJ: Prentice Hall. Run Relations and Short Energy Economics 31(2009) : 503 509 National Climatic Data Center, NOAA. Historical Climatological Series 5 1 and 5 2 Newcomer, Adam, Seth A. Blumsack, Jay Apt, Lester B. Lave, and M. Granger Morgan. 2008. Environmental Science and Technology Vol. 42 No. 9 ( 2008 ): 3139 3144 PJM Transmission System Operator. University, October 1980
130 Nordhaus, William D. 2011. from the RICE University, October 2011 The Journal of Finance Vol. LI, No. 2 (1996) : 579 611 Based Approach to Short IEEE Transactions on Power Syst ems 5(4), ( November 1990 ): 1535 1550 Potential Effects of Proposed Climate Change Policies Energy Market Rose, Nancy L., and Paul L. Joskow. The RAND Journal of Economics Vol. 21, No. 3 (1990) : 354 373. Identification and Inference for Econometric Models: Essays in Honor of Thomas J. Rothenberg, J.H. Stock and D.W.K. Andrews (eds), Cambridge University Press, Chapter 5 Examining the Effects of Deregulatio n on Retail Electricity Prices Cornell University Working Paper, February 2006. Tol, Richard S.J. 2005. The Marginal Damage Costs of Carbon Dioxide Emissions: An Energy Policy Vol. 33 (2005) : 2064 74 United States Department of the Census Population by State 1990 2009 Annual Energy Outlook, 2008 2013 Market a Form 423 ( Monthly Cost and Quality of Fuels for Electric Plants Data ) 2008 2013 United formation Administration Form 860 ( Annual Electric Generator Report ) 2008 2013 Form 861 ( Annual Electric Power Industry Database ) 2008 2013
131 United States Department of Energy Energy Information Administration. State Energy Data System Database United States Department of Energy Energy Information Administration. EIA Form 861 Database. Ene rgy and Security Act of 2009, H.R. 2454 in the 111 th Act in the 111 th United States Government Accountability Office, 2008. U.S. Senate United States Securities and Exchan
132 BIOGRAPHICAL SKETCH Ted Kury is director of Energy Studies for the Public Utility Research Center (PURC) at the University of Florida. He is responsible for p romoting research and outreach activities in energy regulation and policy. He develops research strategies that inform the academic community and practitioners on emerging issues and best practices and serves as an expert resource for regulatory profession als, policymakers, and service providers in Florida and around the world. change policies and serves on the steering committee of En ergy He also collaborates with faculty at other universities around the state as part of the Florida Energy Systems Consortium, a consortium recently created by the governor to leverage assists in the coordination In collaboration with the World Bank staff, he designs curriculum and leads sessions of the PURC/World Bank International Training Program on Utility Regulation and Strategy He also de velops advanced courses and customized training courses in energy regulation. Previously, Mr. Kury was a senior structuring and pricing analyst at The Energy Authority in Jacksonville, Florida where he developed proprietary models relating to the manageme nt of system wide cash flows at risk, including the quantification of portfolio risk related to both physical utility and financial assets. He also built custom software packages to quantify cross commodity risk, asset valuation, and optimization of natura l gas storage with dynamic programming.
133 Mr. Kury began his career in energy as a senior economist at SVBK Consulting Group in Orlando, Florida. Some of his duties included participating in legal proceedings relating to the deregulation of electric markets and establishment of tariffs and helping municipal electric, natural gas, and water/wastewater utilities develop retail rates. Mr. Kury e arned M.A. and B.A. degrees in e conomics from the State Un iversity of New York at Buffalo, a nd his Ph. D. in e conomics from the University of Florida in 2013.