Investments in Electricity

Competitive Electricity Markets and Sustainability
Edited by
Franois Lvque
Professor of Law and Economics, Centre of Industrial Economics (CERNA), cole des mines de Paris, France
Edward Elgar
Cheltenham, UK Northampton, MA, USA
Franois Lvque 2006 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical or photocopying, recording, or otherwise without the prior permission of the publisher. Published by Edward Elgar Publishing Limited Glensanda House Montpellier Parade Cheltenham Glos GL50 1UA UK Edward Elgar Publishing, Inc. William Pratt House 9 Dewey Court Northampton Massachusetts 01060 USA
A catalogue record for this book is available from the British Library Library of Congress Cataloguing-in-Publication Data Competitive electricity markets and sustainability / edited by Franois Lvque p. cm. Includes bibliographical references and index. 1. Electric utilities. 2. Competition. 3. Electric UtilitiesFinance. 4. Demand-side management (Electric utilities) I. Lvque, Franois, 1957 HD9685.A2C577 2006 333.793 23dc22 2006011132
ISBN-13: 978 1 84542 921 8 ISBN-10: 1 84542 921 4 Printed and bound in Great Britain by MPG Books Ltd, Bodmin, Cornwall
Contents
List of gures List of tables List of contributors Preface by Jean Syrota Acknowledgements 1 Investments in competitive electricity markets: an overview Franois Lvque PART I INVESTMENT IN GENERATION 21 54 vi vii viii xiii xv 1
2 Investment and generation capacity Richard Green 3 Generation technology mix in competitive electricity markets Jean-Michel Glachant PART II INVESTMENT IN TRANSMISSION
4 Problems of transmission investment in a deregulated power market Steven Stoft 5 Patterns of transmission investments Paul Joskow PART III COORDINATION BETWEEN INVESTMENTS IN GENERATION AND TRANSMISSION
87 131
6 Long-term locational prices and investment incentives in the transmission of electricity Yves Smeers 7 Compatibility of investment signals in distribution, transmission and generation Ignacio Prez-Arriaga and Luis Olmos Index
v
187
230
289
Figures
The determination of electricity capacity and prices 23 How the capacity mix aects revenues 27 Investment in England and Wales 42 Investment in Finland 43 Investment in Norway 44 Investment in Sweden 44 Investment in the United States 45 The determination of electricity capacity and prices 49 Present-day cost of generating electricity in the UK, 2003/04 69 3.2 CCGT cost of entry by country in Europe in 2005 70 3.3 Spark spread in Texas, 19992002 78 3.4 Finnish comparison of generation costs 80 4.1 Dening congestion rent and congestion cost 89 4.2 Cost to consumers compared with congestion cost and rent 90 4.3 Relationship of congestion to a transmission-cause reliability problem 92 4.4 A positive present value is not sucient 97 4.5 Lumpy technology may not exhibit returns to scale in the long run 101 4.6 Option rights reduce the feasible set of rights 111 4.7 Optimal investment in lump technology may be preferable 114 4.8 Optimal investment eliminates congestion 116 4.9 Investors should not capture full social benet 117 7.1 Process of computation of locational signals 252 7.2 Average L and G taris in Europe 265 7.3 L nodal taris in Europe 266 7.4 G nodal taris in Europe 266 7.5 Original and new L taris in Spain for the IEM-13 system 269 7.6 Original and new G taris in Spain for the IEM-13 system 270 7.7 Comparison between the transmission tari and the net inter-TSO payment for 17 European countries 271 7.8 Evolution of the energy price of several power exchanges belonging to Europex, January 2000 to November 2004 271 7A2.1 Proportionality principle in average participations 285
vi
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2A.1 3.1
Tables
3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 4.1 5.1 5.2 5.3 5.4 5.5 5.6 5.7 7.1 Generation capacity in the USA, 19902002 Generation capacity in California and Texas Generation capacity in Norway, England and Wales, Spain and Italy Generation fuel mix in the USA, 19982002 Generation capacity changes in the USA, 19902002 Generation capacity changes in California and Texas Generation capacity changes in England and Wales, Spain and Italy 1996 forecast costs of producing electricity, 2000 and 2015 2004 forecast costs of producing electricity, 2010 and 2025 Nuclear generation costs in the early twenty-rst century Nuclear generation costs in the 2003 MIT study Nuclear versus gas CCGT cost of capital analysis Three views of congestion Reliability upgrade projects: New England regional expansion plan 2004 Schedule of transmission network use of system generation charges, 2004/2005 Schedule of transmission network use of system demand charges and energy consumption charges, 2004/2005 E&W system operator incentive mechanism under NETA PJM interconnection charges: proposed Erie West HVDC Market window economic transmission projects in PJM as of November 2004 Examples of transmission congestion mitigated by reliability investments in PJM Impact of dierent factors on the total generation capacity needed to supply a 384-MW load, located close to a main consumption center, from two dierent locations, one close to the load center and the other close to an entry point for LNG Comparison of the cost savings involved in supplying a 384-MW load located close to a main load center 57 58 59 60 63 65 66 69 70 72 74 76 90 143 160 161 162 170 174 178
272 274
7.2
vii
Contributors
Jean-Michel Glachant is tenured professor and Head of the Department of Economics at the University of Paris Sud (France) where he created the Network Industry Research Group (GRJM). Prior to joining University Paris Sud in 2000, throughout the 1990s he was deputy director or director of the leading French institutional economics research centre (ATOM) at the Sorbonne University. His work focuses on the institutional economics of competitive reforms in the European network industry. His current work focuses on the creation of a single energy market in the European Union extended to 25 member states. He has advised the European Commission (DG Energy and DG Competition) on electricity reforms. He has been a member of the Economic Advisory Committee at the French Energy Regulatory Commission. He is a member of the board of the International Society for New Institutional Economics (ISNIE) and of the Faculty of the European School for Institutional Economics (ESNIE); and a partner of the Electricity Policy Research Group at the University of Cambridge, and of the European Energy Institute (EEI). He received his PhD in economics from the Sorbonne University. Richard Green is professor of economics at the University of Birmingham, UK and Director of the Institute for Energy Research and Policy. He has been studying the economics and regulation of the electricity industry since 1989, just before the industry in England and Wales was privatised. With David Newbery, he was responsible for the most inuential study of competition in the British electricity spot market. He has written two books, and more than 40 articles and book chapters, mostly on the electricity industry and its regulation. He is an associate editor of the Journal of Industrial Economics, and on the Editorial Board of the Journal of Regulatory Economics. He has spent a year on secondment to the Oce of Electricity Regulation, and has been a visiting Fellow at the World Bank Institute, the University of California Energy Institute and the Massachusetts Institute of Technology. He has been a specialist advisor to the House of Commons Trade and Industry Committee, and is on the academic advisory panel to the sta of the UKs Competition Commission.
viii
Contributors
ix
Paul Joskow is Elizabeth and James Killian Professor of Economics and Management at the Massachusetts Institute of Technology and Director of the MIT Center for Energy and Environmental Policy Research. He received a BA from Cornell University in 1968 and a PhD in economics from Yale University in 1972. Professor Joskow has been on the MIT faculty since 1972 and served as Head of the MIT Department of Economics from 1994 to 1998. At MIT he is engaged in teaching and research in the areas of industrial organization, energy and environmental economics, competition policy and government regulation of industry. He has published six books and more than 120 articles and papers in these areas. His papers have appeared in the American Economic Review, Bell Journal of Economics, Rand Journal of Economics, Journal of Political Economy, Journal of Law and Economics, Journal of Law, Economics and Organization, International Economic Review, Review of Economics and Statistics, Journal of Econometrics, Journal of Applied Econometrics, Yale Law Journal, New England Journal of Medicine, Foreign Aairs, Energy Journal, Electricity Journal, Oxford Review of Economic Policy and other journals and books. He is a Director of National Grid plc, a Director of TransCanada Corporation and a Trustee of the Putnam Mutual Funds. He previously served as a director of New England Electric System. He has served on the US Environmental Protection Agencys Acid Rain Advisory Committee and on the Environmental Economics Committee of the EPAs Science Advisory Board. He is a member of the Scientic Advisory Board of the Institut dconomie Industrielle (Toulouse, France) and the Scientic Advisory Board of the Conservation Law Foundation. He is a past-President of the International Society for New Institutional Economics and a Fellow of the Econometric Society and the American Academy of Arts and Sciences. Franois Lvque is professor of law and economics at cole des mines de Paris and visiting professor at the University of California at Berkeley. He is Director at Cerna, the research centre of the cole des mines de Paris in industrial economics. He has published several books on antitrust economics (Antitrust, Patents and Copyright, Edward Elgar, 2005; Merger Remedies in American and European Union Competition Law, Edward Elgar, 2003), on the economics of regulation (conomie de la rglementation, ditions La Dcouverte, 1999 and 2005; Transport Pricing of Electricity Networks, Kluwer Academic, 2003) and on the economics of intellectual property rights (Economics of Patents and Copyright, Berkeley Electronic Press, 2004). He is the author of 50 articles in the same areas. He has coordinated several large European research programmes on electricity reforms and energy policy.
Contributors
Lvque taught economics of natural resources at the cole des mines de Paris (198490), and environmental economics at cole des Hautes Etudes en Scince Sociles (EHESS) (19972001) and at Pavia University (19992002). In 1999 he created a new major in law and economics at the cole des mines. He has taught industrial economics at the cole des mines since 1996 and energy economics since 2004. He has also taught EU Competition Law at the Boalt School of Law, University of California at Berkeley, since 2002. He has regularly been commissioned by the French government, the Organization for Economic Cooperation and Development (OECD) and the European Commission to undertake consultancy and to participate in advisory committees. He founded Microeconomix, a Paris-based boutique specialising in the economic analysis of legal disputes. He is also a member of the French Environment Accounting Commission and of the Council on Intellectual Property. Luis Olmos was born in Madrid, Spain, in 1976. He received an Electrical Engineering degree and a PhD degree from the Universidad Ponticia Comillas (UPCO) in 2000 and 2006, respectively. Currently, he is a researcher at the Instituto de Investigacin Tecnolgica. His interests include areas such as the regulation of electricity markets and planning of power systems. He has worked on several aspects of the operation of power systems, such as the provision of ancillary services (load-frequency regulation). Currently, he is working on transmission pricing issues in the context of regional markets with a special focus on the problems of congestion management, sunk costs recovery, tariff design and grid expansion. Ignacio Prez-Arriaga was born in Madrid in 1948. He received a degree in electrical engineering from Comillas University, Madrid, Spain, and MS and PhD degrees in electrical engineering from the Massachusetts Institute of Technology (MIT), Cambridge, USA. He is Director of the BP Chair on Sustainable Development and Full Professor of electrical engineering at Comillas University, where he was Founder and Director of the Instituto de Investigacin Tecnolgica (IIT) for 11 years, and has been Vice-Rector for Research. For ve years he served as a commissioner at the Spanish Electricity Regulatory Commission. He is a life member of the Spanish Royal Academy of Engineering and Fellow of the Institute of Electrical and Electronic Engineers (IEEE). He is Director of the annual Training Course of European Energy Regulators at the Florence School of Regulation within the European University in Florence. He was the author of the White Paper on the Spanish electricity sector commissioned by the government in 2005. He has been principal researcher in
Contributors
xi
more than 40 projects and has published more than 100 papers in national and international journals and conference proceedings. He has worked and lectured extensively on power system dynamic analysis, monitoring and diagnosis of power system devices and systems, intelligent computer design of industrial systems, planning and operation of electric generation and networks, and regulation of the electric power sector. In this last topic he has been a consultant for governments, international institutions, industrial associations and utilities in more than 30 countries. His current research interests are centered on regulation of the electric power industry, the design of regional electricity markets and energy sustainability. Yves Smeers is the Tractebel Professor of Energy Economics at the Universit Catholique de Louvain in Belgium where he is aliated with the Department of Mathematical Engineering and the Center for Operations Research and Econometrics. He received an engineering degree from the Universit de Lige in 1967, and a degree in economics from the Universit Catholique de Louvain in 1969. He also obtained an MS degree in industrial administration and a PhD in operations research from Carnegie Mellon University in 1971 and 1972, respectively. His current research interests concentrate on computational equilibrium models and risk management in restructured electricity and gas systems. His experience in the area extends from operational to strategic market simulation models. He has published extensively in the area and acted as project leader on many projects for the European Commission, the World Bank, the OECD and the Belgian government. He has also conducted various assignments for major European gas and electricity companies, as well as for regulators. He is currently scientic adviser at the Department of Strategy of Electrabel/Suez where he works on market simulation models and risk management. He has recently published several articles in Operations Research, the Journal of Network Industries, Networks and Spatial Economics and Utilities Policy. Steven Stoft is an economist and independent consultant with 12 years experience in power market analysis and design. He is the author of Power System Economics, Designing Markets for Electricity (IEEE, 2002) and also of many published article on electricity market design. He has advised PJMs Market Monitoring Unit since 1999, was an expert witness for Californias Public Utility Commission and Electricity Oversight Board (EOB) in their litigation over long-term contracts before the Federal Energy Regulatory Commission (FERC). Beginning in 2004, he has worked with ISO New England in designing their Investment Capacity Payment (ICAP) market and was their expert economic witness before FERC. He is also
xii
Contributors
working with the EOB on installed capacity markets for California. Previously he was a Senior Research Fellow at the University of California Energy Institute, worked on regulatory and restructuring issues at the Lawrence Berkeley National Laboratory and spent a year in the Oce of Economic Policy at FERC. He received his BS in engineering math and his PhD in economics from the University of California at Berkeley.
Preface
Jean Syrota, Chairman (20002006), Commission de Rgulation de lEnergie
At the beginning of the liberalisation of electricity markets during the 1990s, the electricity industry was in a situation of plentiful generation and transmission. Then in order to monitor the liberalisation process, much attention was paid to the regulatory mechanisms that encourage the actors to improve the eciency of their short-term operations and to deal with the technical complexity that was faced in substituting daily free energy markets for the traditional centralised dispatch without reducing the safety of operation or downgrading the economic eciency of the previously integrated systems. Accordingly, a large number of researchers and practitioners worked actively to dene consistent rules dealing with daily markets, redispatching, balancing mechanisms and ancillary services procurement in parallel with grid-use taris and ex ante congestion management methods. Despite the work done, improvements are still possible and desirable. Meanwhile, with the increase of the overall demand and the reduction of generation margins notably entailed by the retirement of the oldest generators, signicant spot price increases have been observed and the grid became congested more frequently, especially at interconnections between the European countries. This situation raised some new issues that were largely ignored in the rst phase of the process. Especially in a liberalised context, these issues are strongly related to the denition of appropriate incentives for investment in generation and transmission. In the electricity sector there are diculties that are not present elsewhere. As far as the grid is concerned, there is lumpiness in a number of components and in a signicantly meshed network, a very common situation in the real world, there are numerous externalities between individual lines, making it dicult or even impossible to compute their incremental individual benets. Moreover, even if generation and transmission are often complementary goods, there are also many situations where they are replaced, for example when grid improvements can reduce or delay the need for new generation and conversely.
xiii
xiv
Preface
Thus generation and transmission are theoretically coupled and the vertically integrated structure that prevailed for a long time was supposed to allow a joint optimisation. When transmission and generation are properly unbundled, as required by European law, the joint optimisation can no longer be achieved, but some kind of coordination must be found to maintain an acceptable level of eciency without hindering the competition between generators. Additional diculties come from the length of time necessary to achieve new generation and transmission infrastructures with the associated risks for security of supply in the case of late commissioning. The regulator must assess to what extent coordination mechanisms are compatible with the requirements for the achievement of a fair and ecient competition in supply. To deal with such complex issues, the French Regulatory Commission for Energy (Commission de Rgulation de lEnergie: CRE) found it useful to promote academic research by prominent specialists of the liberalisation process in electricity markets. This book is the result of that research. The CRE is indebted to the authors for their very interesting analyses of the investment decision-making process and of the possibility of reconciling the perhaps diverging interests of two major parts of the electric industry. I am sure that these analyses will also be of interest to those who want to better understand the multiple aspects of the ongoing liberalisation process.
Acknowledgements
The origin of the book is based on discussions between the authors who met in January 2005 in Paris at the Commission de Rgulation de lElectricit, the French Regulatory Electricity Authority. The authors are grateful to the CRE and especially to Michel Massoni for nurturing the debates. The book would not have been written without their support. The usual disclaimers apply.
xv
1.
Investments in competitive electricity markets: an overview

Franois Lvque
INTRODUCTION
1.
Over the course of the past 20 years, most of the countries in the OECD have engaged in a competitive opening of their electricity markets. The incumbents were stripped of their legal monopolies, wholesale markets were formed, and dedicated organisations assumed management of the transmission grid. Large consumers acquired the ability to choose their electricity supplier. This opening to competition brought about a profound change in terms of the investment in both generation and transmission. Decisions concerning the construction of new power plants, in particular the timing and the technology mix (that is, the proportion of hydro electricity, nuclear, thermal and so on) now depend on decentralised initiatives of investors, and not on public authorities. As to transmission, which remained a monopoly, the reinforcement and expansion of high-tension power lines are no longer directly controlled by the generators. System operators have greater leeway for initiative. Depending on the specic case, they can sell nancial transmission rights, submit investment programmes to the regulatory authority, or invest as they see t. In a word, investments in an electricity system that is open to competition will no longer be coordinated by the same mechanisms as in the past. The planning that enabled a monopolistic and vertically integrated producer to adjust base- and peak-load capacities, as well as generation and transmission capacities, has been replaced by a series of decentralised decisions partly based on prices. This new decision set which involves many agents and combines market signals with regulation must be understood in detail. A thorough understanding is necessary to reveal to what extent, and under what conditions, competitive opening will result in an investment level that is consistent with the public interest. Only this will allow identication and evaluation of solutions to situations of investment shortfall or oversupply such as those we have seen arise on several occasions (for example, underinvestment in interconnection capacity in California, and
1
Competitive electricity markets and sustainability
overinvestment in independent gas-powered plants in the United States during the 1990s.) This is the spirit in which this book was prepared. This rst chapter contains seven sections. Following this introduction, Section 2 reviews the new terms of investment in generation and transmission. Sections 3 and 4 address investment in generation and in transmission, respectively. Section 5 resumes the discussion of the interface between investments in generation and transmission that we briey began in Section 2. Section 6 provides a preview of some essential points that the co-authors of this book raise in subsequent chapters, and that were not mentioned in the preceding sections. Finally, Section 7 concludes.
2.
THE ISSUE
Ideally, an optimal level of investment in the electricity system would involve joint optimisation of investments in generation and transmission. In fact, the goal is to minimise the cost of electricity to consumers. From an economic perspective, generation and transmission are complementary goods; if the price of one decreases, the quantity sold of the other increases. The mechanism underlying this phenomenon is simple: consumers are sensitive only to the total price of electricity since they do not consume the generated electricity and the transmission service separately. Consequently, if the price of a KWh falls, ceteris paribus, they will consume more electricity and demand a greater quantity of the transmission service. Consequently, investments in generation and transmission complement each other. Sometimes, however, investments in generation and transmission are substitutable. For example, in an isolated region with limited interconnection with the grid, a rise in local demand can be satised by either reinforcing the line or building a new power plant within the zone. If both investments occur simultaneously, then neither will be protable. When both activities are combined within a single rm, joint optimisation of investments is deemed self-evident, since the stockholder or manager maximises overall prots. In an electricity system that is open to competition, the visible hand of the manager fails to ensure coordination between generation and transmission. Transmission is separated from generation in one way or another (that is, accounting, managerial or legal unbundling) in order to ensure that rival generators have equitable terms of access to the grids. According to Steven Stoft, this new situation opens the door to strategic behaviour on all sides. In order to provide for future investments in transmission, the transmission system operator (TSO) must be informed of future investments in generation. Conversely, to plan these investments in
generation, producers require forecasts of the TSOs future investments in the grid. To escape from this deadlock, one of these stakeholders must draw rst by revealing its intentions and proceeding with the investment. However, the rst to invest becomes hostage to the other, since it is impossible to move a power plant, or pylons, without forfeiting the bulk of their value. This is the classical economic problem of the hold-up occasioned by stranded costs. The upshot is generalised underinvestment: each party, knowing that it may be taken hostage ex post, reduces investments ex ante. Thus, we cannot apply the idealised rule for investment in transmission, which would have the system operator (SO) plan investment by optimising transmission and generation as a function of future demand and then lay the power lines in the hope that the market will induce generators to invest according to plan.1 None the less, it is necessary to avert a profusion of waste by nding some way to coordinate investments in generation and transmission. Various instruments, such as nancial transmission rights and a zonebased rate structure for the grid, have been proposed in the recent economic literature. These are described and discussed in the chapters by Steven Stoft, Yves Smeers, and Ignacio Prez Arriaga and Luis Olmos. Before examining them more closely, it will be useful to examine the optimisation of transmission and generation separately. Although this simplies the issue considerably, these two issues taken individually are far from trivial. We shall now examine how to optimise the utilisation and size of the grid when generating capacity is optimal, and how to optimise the utilisation and volume of generating capacity when the grid is optimal.
3.
INVESTMENT IN GENERATION
Apart from grid constraints, what obstacles must the market mechanism contend with to yield a socially ecient level of investment in generation, that is, a level that satises the users needs at the lowest cost? The optimal investment in electricity generation is precisely determined by the theory, which addresses both total capacity and its distribution among power plant types. These latter, in fact, dier in terms of both variable costs, which are usually linked to the price of fuel, and xed costs, which essentially reect expenditures on construction. For a nuclear power plant, the former are low and the latter very high; for a gas turbine this is inverted. Consequently, nuclear plants should be used throughout the year to meet base-load requirements, while gas turbines should only be called on to meet peak-load demand at times of the year when there are spikes in demand.
In Chapter 2, Richard Green presents a simple model of optimal levels of generating capacity comprising only these two types of plants. Emphasising a graphical approach, he demonstrates how to identify the load-duration curve for the 8,760 hours in a year and how to translate it into an hourly price curve. Naturally, the highest price is found when demand is greatest. As this demand exceeds available capacity, the equilibrium price is not set at the marginal cost of the last unit generated, but rather at a higher level equal to the marginal opportunity cost of consumption (that is, above which the last consumer prefers to forgo rather than consume). The gap between these two marginal costs thus allows the peak-load plant that operates for the shortest period during the year to cover its costs. Note that this result contradicts the conventional wisdom that the electricity market is incapable of ensuring that plants xed costs are covered. This confusion arises from an overly hasty equating of the equilibrium price with the marginal cost of generation. In the presence of congestion, as during extreme peaks in this case, the shortage must be managed and resources allocated to those economic agents on whom the lack of access imposes the greatest cost. Furthermore, as Green reminds us, economic theory demonstrates that if the peaking plant that is used least covers its total costs, and if the allocation among the various means of generation is ecient, then all other plants can cover their total costs with market prices that are based on marginal costs. We note that investors clearly had no doubts regarding the ability of electricity markets to render new investments in generation protable. In the United States, as in England, there was even talk of a boom in the construction of new power plants, in particular those based on combinedcycle gas turbine (CCGT) technology. At the end of his chapter, Green examines trends in gross and net investment (the latter accounts for the decommissioning of old plants) and of the capacity margin in those two countries, as well as in Finland, Norway and Sweden. He particularly notes two phenomena. First, the capacity margin is shrinking. This result is consistent with the expected and desired results of the electricity system reforms, in the sense that the previous regime was characterised by excess capacity attributable to cost-plus regulation. Second, beyond a certain threshold, the shrinking of the capacity margin serves as a trigger to stimulate the resumption of investment. In his chapter, Jean-Michel Glachant also draws attention to the shift in the energy mix towards gas-based electricity generation. He measures and comments on it in the case of several US states, England, Italy and Spain. This evolution is consistent with developments in the relative performance of the dierent technologies, as the total cost of CCGTs has fallen below that of nuclear technology.
The preceding economic model assumes that there is no uncertainty in terms of demand.2 However, consumers reactions to price changes are very poorly understood. Except in the case of certain large consumers, who adjust their consumption to variations in the real-time prices on the spot market or accept compensation for forgone consumption, information on the price sensitivity of demand is inadequate. Most consumers are not confronted with hourly, or even daily, uctuations in the price of electricity. Their consumption is measured on a monthly or quarterly basis, and they are charged a rate per KWh that is independent of the hourly distribution of their consumption. Shielded thus from real-time price volatility, they have no need to hedge against the risk of high prices. Furthermore, most domestic consumers cannot be disconnected individually. And yet, there is no reason to believe that residents of residential neighbourhoods will face the same opportunity cost of not consuming. However, since they are all hooked into the same distribution network, creating a market of interruptible contracts cannot be readily envisaged. Consequently, there is no mechanism for revealing households willingness to pay during peak hours. Note that the underlying problem of short-term price inelasticity of demand did not originate with the opening of electricity systems to competition. Under the previous arrangement, estimates of the value of electricity lost in the event of a service interruption (value of loss load, or VOLL) were simulated by the planner in order to decide when generation capacity needed to be boosted. When the cost of the new investment was lower than the benet of the averted service interruption VOLL multiplied by the reduction in risk of blackout attributable to the increased capacity (loss of load probability, or LOLP) the investment was deemed worthwhile. To x an order of magnitude, if VOLL is 10,000 per MWh, then the public interest is served by the construction of a power plant that will reduce the risk of interruption by approximately ve hours over the course of a year. Today, with electricity systems that are open to competition, VOLL can also serve as a reference value. For example, during critical periods, a system operator may decide to purchase power at a price equal to VOLL. In this event it is acting in the name of, and on behalf of, consumers. However, it is quite unusual for the regulatory authorities to authorise such an astronomical price on the spot market, even during critical periods. The very potential of prices to reach that level provides a powerful incentive to generators to withdraw some capacity from the market so as to drive up the price that is, to exercise their market power during periods of tension between supply and demand. Thus, for reasons of social acceptability and market power, the spot price is often capped by regulation at a level far below the VOLL. This type of intervention inevitably distorts the
market signal towards underinvestment, and the plant with the shortest period of operation during the year can no longer cover its xed costs. The entire cascading structure for covering the xed costs of the various plants collapses. When real-time market prices are capped, undercutting investments, it becomes necessary to invoke other instruments to provide economic agents with a signal for the optimal capacity level. One elegant approach is based on the notion that generators do not supply a single good, electricity, but rather two goods, energy and capacity. Consumers value two services, the power itself when they want to watch television or turn on a light, and also an option value for being able to do this at any time. From this perspective, generators should be compensated for the capacity they supply regardless of their utilisation. In practice, two systems have been implemented: obligation capacity and capacity payments. In the former, retailers (suppliers to the end-users) are obliged to maintain a capacity that exceeds their expected peak load. To meet this requirement, they acquire purchasing rights from generators on a capacity market created for that purpose.3 In principle, the required capacity level must be determined by comparing VOLL with the cost of the supplementary obligation capacity. Provision must also be made to penalise retailers for failure to comply with the mandatory supplementary capacities imposed on them. We observe that this penalty establishes a de facto ceiling on the capacity market; retailers will prefer paying it to buying capacity at a higher price. Consequently, the amount of this ne must be linked to the cost to generators of making capacity available. In the United States, the utility PJM (PennsylvaniaNew JerseyMaryland) has enforced this type of obligation capacity market for several years. The required level represents about 20 per cent of peak load and the penalty corresponds to the xed costs of a peaking plant ($7.4 per MWh). During the 1990s, the English Pool established a capacity payments system. Here, the compensation to generators for the capacity they supplied was directly integrated into the electricity spot price. Unlike under the previous system, there was no dedicated capacity market on which supply and demand met directly. The capacity payment is also determined from VOLL. In the case of England, it was set equal to VOLL minus the higher of the station bid and marginal price (SMP), this dierence being multiplied by LOLP. Whether the selected system is obligation capacity or capacity payments, it is essential to bear in mind that the signals sent to investors originate at least as much from public authorities as from private agents. On the side of the invisible hand of the market: all the decentralised consumption and generation decisions that propel the evolution of the price; on the side of
the visible hand of public intervention: identifying and setting the price cap and calibrating VOLL. We shall see this hybridisation recur in the case of investments in transmission.
4.
INVESTMENT IN TRANSMISSION
Like other network infrastructures, electricity transmission grids present technical and economic characteristics that are quite challenging from the perspective of resource allocation. Like highways and airport runways, electrical transmission lines are congested. As a result, use of this infrastructure by one agent may degrade the quality of service available to another. In economic jargon, this is known as a negative externality. In the case of electricity, congestion may even result in the complete collapse of the system. If the current is not cut, the lines may stretch and melt! Again, like in the case of highway and airport infrastructures, investment occurs in discrete units, leading to discontinuous jumps in capacity. To expand a highway or an airport, a lane or a runway must be added in a single stroke. Smaller, fractional investments are impossible. In electricity, the line type for the high-voltage grid cannot be modulated by a single KV at a time. For example, either 220 or 400 KV must be chosen. Similarly, the gauge of the cable is not available in increments of a millimetre the choice is limited. These two technico-economic characteristics, congestion and indivisibility (or lumpiness), are sometimes evoked in defence of misguided concepts. First misconception: investment must proceed until congestion is eliminated. In fact, if it were necessary to reinforce electricity transmission lines to the point that their capacities would be able to carry any and all transactions between generators and consumers at all times, the grid would be bloated and astronomically expensive. If, during a single hour in one year, a plant that is remote from a consumption zone is 10 per MWh cheaper than a local, more expensive generator, and if one MW of that generation cannot be transmitted to the consumers because of an inadequate line rating, then that line is congested during that hour. The cost of this congestion is 10 per year. Clearly, adding one MW of capacity to that line would be much more expensive! Eliminating all congestion would only make sense if grid construction costs were nil. Obviously, this is not the case, and consequently the economically optimal level of congestion is not zero. In fact, it is found at the point at which the cost of reinforcing the grid is equal, at the margin, to the savings it makes possible, that is, electricity that can be bought from farther away at lower cost. The second misconception is that investment should be undertaken as soon as the new line construction project is protable. It may, indeed, be preferable to wait and
opt for a much more protable project later one which will add far greater capacity at a single stroke. Stoft uses a numerical example to illustrate how it could be better to construct a 1,000 MW line in two years than a 600 MW line today. This is attributable to the lumpiness of the investment, which does not allow demand growth to be matched by an increase in lockstep in generation. As with any infrastructure, it is worthwhile to distinguish between ecient use and ecient size of the network. In the rst case, capacity is treated as a given. Economic optimisation is thus a matter of allocating its use to the economic agents who value it most highly. The theory reveals that the key to accomplishing this lies in setting the access price equal to the short-term marginal cost. In electricity, this cost has two components. The rst is due to ohmic losses that make it necessary to inject more electricity than is withdrawn at the other end of the line. The second component is congestion, which makes it impossible to replace local, high-cost electricity with less-expensive power from a more distant plant. Note that both of these elements of the marginal cost can be expressed as a function of the price of the transmitted power itself, for example in /kWh. This allows us to establish an equivalence between the marginal cost of transmission and the marginal cost of generation. Between two local competitive markets, the equilibrium transmission price will equal the dierence in marginal production costs, so that a buyer will be indierent between buying from a seller who is closer but sells at a higher price and one who is farther away and sells more cheaply. The energy pricing system that corresponds to setting electricity transmission fees equal to the short-term marginal cost is called nodal pricing, or marginal locational pricing. These terms reect the fact that the electricity price is dierent at each node of the network. It also varies across time since demand, and by extension congestion, uctuates between the nodes. For example, the systems operator of PJM, the largest electricity market in the United States, computes the price at the 3,000 nodes several times per hour. The issue of ecient network size is an issue of optimal investment. The goal is to achieve the equilibrium size, that is, expand capacity to the point at which marginal cost rises above the benet yielded by continuing. In electricity, we have seen that this benet amounts to displacing local generation with more remote, cheaper generation. The distinction between ecient use and ecient investment arises because of a discrepancy between short- and long-term marginal costs the former being lower than the latter and between marginal and average costs the former again being the lower. These gaps are explicable in terms of contingencies as well as by the presence of lumpiness and economies of scale. For historical reasons, the current network is far from its optimal size. As Paul Joskow points out, the electricity transmission system we have
inherited today reects historical institutional arrangements, the limits of corporate activity, political boundaries and historical patterns of urban and industrial development. He states: We can change the institutions but we cannot erase the existing infrastructure in place at the time sector liberalization reforms are implemented but only change it gradually over time. As a rule of thumb,4 networks that predate the competitive opening are bloated. Governmental, and especially regulatory, intervention favoured capital expenditures and provided for broad margins of safety to accommodate growing demand and counter the risk of blackouts. Furthermore, investment in tiers is incompatible with the notion that installations erected for a 20- or 30- year lifespan can reect the optimal network size during each year. Inevitably, it will be under- or oversized, depending on the timing. Once again, overinvestment wins out because of economies of scale (that is, the greater the investment in capacity, the lower the cost of capital per unit of capacity). The essential result of the realities described above and the discrepancies they give rise to is that a price equal to the short-term marginal cost ensures eciency in use, but does not fully cover the investment expenditures necessary to construct an optimally sized grid. In other words, the nodal pricing system does not compensate the xed costs of investments in transmission. As Prez-Arriaga and Olmos emphasise, [C]ost recovery by nodal energy prices typically does not exceed 20 percent of total transmission costs. This consequence keeps the market from operating eciently. For Joskow, Transmission networks do not and will not evolve through the workings of the invisible hand of competitive markets. We note that, if there were no gap between the short-term marginal cost and the average cost, then a decentralised mechanism leading to an optimal level of investment might have been feasible. Such a mechanism has been conceptualised. The underlying principle is to allocate transmission rights that yield congestion rents to the owners of each line as they are generated. In his chapter, Stoft describes this mechanism of which we have provided a bare outline here in detail, establishing the link between the level of congestion rents, lumpiness, and economies of scale. Decentralised investments in transmission lines (called merchant lines by convention) are thus conned to modest growth. This conclusion recurs in the contributions of Joskow, Stoft, and Prez-Arriaga and Olmos. These authors envisage merchant lines only as a complement to investments regulated by public bodies. To cite Joskow again: Most transmission investment projects are being developed today and will be developed in the future by regulated entities. Or, according to Prez-Arriaga and Olmos: [R]egulated investment . . . must play a predominant role in the future development of almost every realworld transmission network.
10
5.
. . . AND BACK TO THE COORDINATION BETWEEN INVESTMENTS IN GENERATION AND TRANSMISSION
In a perfect world, in which demand reacts to the price of electricity and competitive local markets are linked by incrementally extensible transmission lines, the combination of nodal electricity prices and nancial transmission rights ensures a decentralised joint optimisation of investments in generation and transmission. Prices exactly cover the costs of the ecient mix, in terms of both the generation technologies and the distribution between power lines and power plants. Consequently, from a theoretical perspective, perfect coordination of investments in generation and transmission in an electricity regime that is open to competition is not impossible. The problem resides in the unrealistic nature of the assumptions all of which are needed to generate this result. Indeed, in our imperfect electrical world, consumers willingness to pay is not known, some generators possess market power, and transmission technologies feature lumpiness and economies of scale. And yet, the theory is not ready for the scrapyard. On the contrary, it suggests solutions for approaching the optimum and minimising market failure. In light of the failure of nancial transmission rights to cover the xed costs of transmitting, other methods must be envisaged and implemented to complement signals of short-term grid use with long-term signals to drive investment. A rst theoretical method is suggested by Smeers. It is based on designing a rate structure that captures several components. The investment model he elaborates succeeds in inducing an optimal level and location of generating capacity as well as in providing an incentive to the TSO to eciently manage congestion and develop infrastructure despite the fact that investments are indivisible. Smeers draws on the work of ONeill et al. (2004) who expand the denition of goods, energy in our case, to their spatial dimension. His model is more in keeping with the institutional environment prevalent in Europe than that in the United States. Network management is performed by an owneroperator of the infrastructure who integrates dispatching, maintenance and renewal of the infrastructure. Unlike in the situation in which ownership and dispatching are separated, here it is necessary to ensure that the system operator does not curtail investments in order to increase revenues by creating congestion. In the Smeers model, the system operator receives instructions from the regulator concerning how to set the long-term component of the price. It also receives monetary transfers as an incentive to select an appropriate grid conguration. The regulator is assumed to know electricity generators costs, consumers willingness to pay and the set of possible network
11
congurations. Finally, markets are competitive and all agents including the systems operator take prices as given. The work of Prez-Arriaga and Olmos also deals with pricing that is based on several components, combines short- and long-term signals, and covers xed costs. However, their procedure takes a more operational approach. Like Smeers, they focus on the European context. A numerical application of their model of long-term transmission costs has been computed for all European grids. From a practical perspective, it allows levels to be set for payments between system operators for use of the grid in other member countries. We observe that the work of Prez-Arriaga and Olmos is more relevant to cost allocation than to optimisation. It can be summarised as follows. To ensure that all transmission costs are covered, a second component of revenues must be added to the fee structure based on nodal electricity prices. Two cases can be distinguished. The rst deals with highly integrated networks: consumption centres and generation units are more or less evenly distributed throughout the territory, and no systemic congestion is foreseeable at any specic location. In this case, there is no need for localisation signals, especially since the beneciaries of investments in transmission would be dicult to identify and allocating individualised costs and benets impracticable. The other element of the price, to cover xed costs, must be computed by applying the Ramsey rule (that is, the size of the mark-up is inversely related to the consumers price elasticity of demand). In the second scenario, the additional component must capture as nearly as possible the costs and benets to the grid of decisions relating to the siting of the new plant or large energy consumer. Among the several available algorithms that are based on some measure of electricity use, Prez-Arriaga and Olmos recommend simple and robust schemes that are based on the average network use and, in those circumstances when it is essential to send to new network users signals reecting their responsibility on new network reinforcements, they propose some new ideas on how to modify standard algorithms to achieve this purpose. Joskow emphasises the importance of consistency in the organisation of energy markets and the institutions that govern transmission. Organizing power markets and transmission institutions as if a clear separation exists inevitably leads to serious problems. He analyses two cases: one on each side of the Atlantic. The aforementioned PJM is characterised by a system operator who does not own the grid. The grids are owned by electricity utilities that are vertically integrated in generation, distribution, and wholesale and retail operations. However, it is the PJM system operator who runs the day-ahead and balancing markets. It also operates the capacity market. Load-serving entities are, in fact, subject to capacity obligations computed on the basis of their monthly peak requirements. As Joskow explains, these
12
supply requirements play an important role in the process of investment in transmission and in providing siting incentives to generators. The other case he examines is the Anglo-Welsh system. Until March of 2001, the wholesale market was organised into a mandatory pool. Generators were compensated separately for power and for capacity. An energy-only market followed with the implementation of the NETA (New Electricity Trading Arrangements). We observe that the price on this market is not capped. The system operator, National Grid Company (NGC), is integrated. It functions as the system operator, oversees maintenance of the grid, and makes the investments. The transmission price is regulated by Ofgem (Oce of Gas and Electricity Markets) and includes an element that depends on location. Generators in the north of the country pay more than those in the south. In matters of investment, NGC is bound by obligations specied in the network code and by various standards. To comply with them, it conducts studies based on regional demand and supply estimates. When a violation of a standard is identied, NGC determines which investment projects should proceed. Their size determines whether they require approval from the regulator. Joskow emphasises how well the Anglo-Welsh system has performed since the mid-1990s. He considers this to be the most successful experience in market liberalisation anywhere in the world.
6.
OVERVIEW OF THE BOOK AND SYNOPSIS OF THE CONTRIBUTIONS
In addition to this introductory chapter, the book consists of three parts. Each of these comprises two chapters, where pure theory alternate with practical application or empirical study. Part I (Chapters 2 and 3) is devoted to investment in generation, while Part II (Chapters 4 and 5) addresses investment in transmission. Part III (Chapters 6 and 7) examines the issue of coordination between investments in generation and transmission. Greens contribution (Chapter 2) deals with the theoretical mechanisms that determine the choice of the level and mix of electricity generation capacity. A broad outline of these mechanisms was briey presented above. We shall underscore some of the contributions of this chapter in more detail. Green reminds us that, in the nal analysis, investments in generation are not only about increasing capacity to satisfy growing demand. Even with constant demand, new capacity is required to replace plants that are inecient owing to technological obsolescence and those that are at the end of their lifespan. It is just as important to examine the economic determinants of plant decommissioning as of new construction, especially since some plants can be mothballed before being denitively shut down.
13
They can be called on to meet exceptional needs. We note that there is a certain parallelism between decommissioning an old plant and commissioning a new one: both actions are irreversible. In the presence of demand uncertainty, this implies that it may sometimes be preferable to delay the decision rather than act immediately, since time may yield better information. Thus, investment is triggered, not when the price rises above marginal cost, but when it exceeds the marginal cost plus the option value. Conversely, decommissioning of a plant occurs when the price falls below marginal cost minus the option value. Greens contribution also discusses the cyclical character of investments in electricity generation. He notes that, in contrast with other commodities, the possibility of keeping plants in reserve and committing to long-term contracts should smooth the cycles. These latter operate in two ways. First, they mitigate the uncertainty facing entrants by allowing them to x a price or a margin of their sales price over the cost of fuel. Second, long-term contracts can function as a sort of coordination mechanism for investments in generation a mechanism that is starkly lacking after the transition from a monopolistic to a competitive market structure. In Chapter 3, Glachant presents a descriptive and applied economic portrait of the changes to the technology mix induced by the competitive opening. Did the reforms to the electricity sector have an impact on the choice of generation technologies? Does competition create new incentives that are biased towards certain technological developments? Or, conversely, does competition marginalise certain technologies that prospered in the context of a regulated industry? Drawing on extensive data, Glachant observes that, in the United States as in many European countries, electricity reforms were accompanied by a technology shift towards generation with CCGTs. To a lesser extent, an expansion of renewable energies can also be detected. On the other hand, the construction of new nuclear reactors came to a halt and the amount of electricity generated by this technology is declining. The conventional explanation for this dual trend is as follows: liberalisation created competition among technologies, allowing the eciency of gas to come to light, while the reforms also put an end to government subsidies to the nuclear option. Rather than any simple intrinsic superiority of gas, or the withdrawal of government support from research and development into nuclear technology, Glachant demonstrates that nuclear power is handicapped by much higher capital costs than those of electricity generated from gas. This dierential is attributable to much greater nancial risks associated with the choice of nuclear technology. Construction costs and the operational performance of these plants (particularly capacity availability and lifespan) are imprecise and highly variable. Glachant also reveals that capital intensity, the size of the minimum
14
unit of capacity, randomness in the construction schedule owing to antinuclear mobilisation, and the absence of any correlation between the price of the fuel and the price of electricity, are all factors that increase the riskiness of the investment. According to a MIT study that was extensively reviewed by Glachant, this set of factors gives gas an edge over nuclear in terms of the gearing rate (40 percent equity, versus 60 percent for nuclear) and a lower-yield requirement for these funds (8 versus 15 per cent). As illustrated by the credit arrangement of the Finish nuclear project TVO, the yield to investments in nuclear power is undoubtedly to be found in longterm contracts between generators and future buyers, reducing the risks and, by extension, the cost of capital. In Chapter 4, Stoft applies a pedagogical approach to elements of the economic theory that shed light on investments in transmission and the obstacles that undermine market eciency. Here the reader will nd denitions of essential concepts, such as congestion (or redispatching) costs, congestion rent and the cost of congestion to load. These costs are uncorrelated and should not be confused. Stoft also takes care to distinguish between two concepts that are often linked because they both underlie xed costs and violate a basic assumption of the invisible hand of the market; to wit, the convexity of the cost function. These concepts are returns to scale and lumpiness. As a nal pedagogical item, Stoft debunks two misconceptions that are currently in vogue: it is not true that the level of congestion should be reduced to zero; and it is not true that market power is required to recover xed costs. In his contribution, Stoft compares three dierent approaches to investment in transmission: the traditional planning approach, the merchant approach and the performance-based regulation (PBR) approach. He discusses the last of these approaches at length. Here the reader who is unfamiliar with the theory of incentives applied to natural monopoly will nd developments that shed light on the underlying principles of the price cap and on the dilemma confronting the regulator seeking to encourage the system operator to cut costs while not leaving an excessively high rent that will penalise consumers. Stoft points out two major diculties associated with establishing incentive regulation in the case of the electricity transmission grid. The rst is linked to the length of the delay in benets accruing to the investor. The timeframe of these investments may, in fact, look as follows: considerable sums must be committed over several years, which are followed by several more years during which the return is nil, or minimal, and only 10 or 15 years after the beginning of the project does it truly begin to pay o. The second diculty arises from the tight linkage between the investment and security of supply (reliability). In the United States, most of the major blackouts that occurred during the past 35 years were attributable to problems with trans-
15
mission rather than generation. In light of this, regularly pruning trees growing beside power lines, updating computer systems and installing linetrip relays are all essential. Thus, incentive mechanisms that cover activities other than the construction or reinforcement of power lines are needed. In the words of Stoft, PBR for transcos will be useful for shorter-term incentives, but it cannot yet be relied on to solve the long-term investment problems. Since the development of merchant lines is bound to be constrained by issues surrounding the recovery of xed costs, as we saw above, there is, in the nal analysis, no alternative for government authorities but to pursue traditional regulation. In Chapter 5, Joskow sketches out, in some detail, the various existing institutional arrangements that govern operation of the grid, inform the regulatory framework and provide incentives to invest in transmission. He demonstrates how these arrangements depend on the historical, economic and physical characteristics of the network and examines their performance. Joskows contribution is too rich to be summarised here. For example, it contains an exhaustive list of the various components of the network that play a role in reinforcing its capacity. Economic models tend to focus too exclusively on the construction of new lines, for there are many other ways to reinforce a network. Too illustrate, from Joskows list: new relays and switches, reconductoring existing lines, and new remote monitoring and control equipment. He also proposes a classication system for dierent types of investment and discusses it in detail. We also draw your attention to two original observations by Joskow. The rst pertains to the gulf between the viewpoints of economists and engineers on the subject of investments in transmission. The models of the former have little in common with the manner in which investments in transmission are actually programmed and developed, or in how the associated services are priced. They do not account for the engineering reliability criteria on which engineers base decisions to reinforce a network. Of course, there cannot be two disjoint types of investment, one based on economic calculations and the other on reliability. Joskow vigorously argues that these two approaches need to be reconciled. The second observation concerns the distinction between inter- and intra-SO transmission grid investments. Each TSO will rst tend to deal with congestion issues on its own grid independently, and then facilitate residual economic exchanges with other grids. This policy results in congestion being pushed across borders and in reduced economic eciency. Joskow suggests that inter-TSO investment opportunities can be addressed more eectively through interconnected zones using the same reliability criteria and standards of evaluation, as well as by integrating wholesale markets and harmonising pricing practices across countries. He strongly recommends the creation of regional transmission operators.
16
In Chapter 6, Smeers addresses a knotty and so-far unsolved problem nding price signals that will motivate system operators to invest optimally and allow them to recover their costs. In other words, is it possible to decentralise investment decisions when they are lumpy? The solution suggested by Smeers is a price comprising several components and incorporating access and congestion fees. The reader unfamiliar with optimisation models, in particular non-linear models, may benet from reading the introduction (Sections 1 and 2) and the discussion (Sections 8 and 9) of this chapter, where the authors approach and results are summarised in a nontechnical manner. One result that merits comment here deals with European regulation of interconnections. It stipulates that rates must comply with three principles: economic eciency, cost causality and nondiscrimination. Smeers begins by building a model with no linkage between agents localisation decisions and the structure of the network. Thus, his model does not respect the principle of cost causality. Nevertheless, the proposed price structure is ecient because it is based on price discrimination. It is well known in economics that price discrimination provides an economically optimal way for xed costs to be recovered. Next, Smeers introduces cost causality, which allows discrimination to be reduced but not eliminated. This can be accomplished without endangering the balanced budget of the system operator, but only at the cost of partially sacricing the goal of economic eciency. The prohibition on price discrimination must be juxtaposed with the loss of social surplus it entails. When subsidies to investments in interconnections are precluded, an arbitrage between the allowed level of discrimination and the tolerable amount of economic loss becomes necessary. Nothing in the European texts or discussions provides for this arbitrage. In Chapter 7, Prez-Arriaga and Olmos address the same issue as Smeers, long-term siting signals and covering the xed costs of transmission networks. However, they take a dierent approach their perspective is practical and their process operational. This compels them to make certain concessions, notably in adopting cost-allocation methods that sometimes owe more to accounting than to economics, and also in simplifying the physical functioning of the grid. Their contribution nicely rounds out the preceding contributions. In addition, they examine how locational signals that are derived from the existence of the transmission network dierences in energy prices due to losses and congestion, plus transmission charges with locational dierentiation compare numerically among themselves and also with other non-electrical locational signals, such as potential charges for the use of gas infrastructures or dierences in the eciency of thermal power plants because of the altitude over sea level.
17
It is recognised that the agents who make the decisions on transmission investments strongly depend on the specic regulatory paradigm that is adopted in each country: system operators, regulators, coalitions of network users and merchant investors alone or in dierent combinations can be the responsible parties. Accordingly, the economic signals that may provide incentives to make correct decisions on new transmission investments depend of the adopted regulatory paradigm. Although all the considered paradigms are useful ones, not all of them would result in a well-developed network. Under a competitive regulatory framework it is essential for the successful development of both generation and transmission to minimise the uncertainty that the decisions of generators create for the network planner and, conversely, that complete and reliable estimates of future transmission conditions be facilitated to generators by the system operator. Several regulatory instruments can be applied to reduce the unavoidable level of uncertainty that surrounds the decision making process of generators and transmission planners. Prez-Arriaga and Olmos remind us that there is more to the electricity network than the high-voltage transmission grid. The structure and renewal of the distribution grid must also be considered. However, these two components of transmission full dierent functions. Consequently, regulatory approaches and investment criteria must dier as well. A series of practical considerations are proposed in this contribution in order to ensure compatibility of signals for investments in transmission and distribution.
7.
CONCLUSION
After reading this introductory chapter, the reader may be amazed at the length of the road to be travelled on the way to ideal investment conditions. It should not be forgotten that this dicult task springs from a very ambitious goal. Investment is an issue of dynamic economics. This is more complicated than problems of static eciency, and the corresponding economic tools are less robust. Moreover, in this case the duration of investments is measured, not in years, but in decades. Seeking to know the optimal generating and transmission capacity of the electricity system is no less ambitious than attempting to build the cities of tomorrow and design the network of highways and byways that will link them. We must accept that the ideal of an electrical utopia will elude us, but instead we can elaborate principles of urban and land-use planning that will make decentralised decisions more ecient. Such is the hope of this undertaking.
18
NOTES
1. Note that application of this idealised rule not only runs up against the opportunism of generators. It also assumes that the SO (or the competent regulator) acts in the public interest, is able to forecast future energy demand, and is able to dene precisely the optimal level of capacity (that is, the number, type and location of plants) and the grid conguration that will satisfy that demand eciently. 2. It also assumes risk neutrality of investors. Risk aversion leads to underinvestment in peaking plants some of which are only protable, in principle, if they operate several hours per year on average. 3. If they are vertically integrated, they can arrange this supply internally. 4. With the notable exception of interconnections between countries. In Europe, these were built for security rather than business considerations. The opening to competition and burgeoning trade soon made their inadequacy clear.
REFERENCE
ONeill, R.P., P.M. Sotkiewicz, B.F. Hobbs, M.H. Rothopf and W.R. Steward Jr (2004), Ecient market-clearing prices in markets with non convexities, European Journal of Operations Research, 164 (1), 26985.
PART I
Investment in generation
2.
1.
Investment and generation capacity

Richard Green*
INTRODUCTION
This chapter considers the questions of how much generation capacity, and of what types, is required, and whether the market will provide it. It starts with a model in which there are two generating technologies, and shows how to obtain the optimal level of each kind, before focusing on the question of the overall level of capacity. Note that the discussion so far has been in terms of the level of capacity, rather than of investment. In an ideal world, this focus on capacity is appropriate, since net investment would equal the dierence between the desired level of capacity and the current level. This is dening net investment as the amount of capacity added to the industry, less the amount taken out of service, rather than by deducting any kind of depreciation charge from gross investment. The amount of plant taken out of service can be endogenous, of course, since far more plants are retired because they have reached the end of their economic life than because they have become physically incapable of (safely) generating any more electricity. In the real world, investment (and closure) decisions are more complicated than a simple comparison of actual and desired capacity would suggest. In particular, many capital-intensive industries are prone to capacity cycles, and there is a danger that the electricity industry could go down the same path. Investments in power stations are also typically lumpy and irreversible, and the theory of real options has potentially important implications for decisions of this type. This chapter considers these issues.
2.
A SIMPLE MODEL OF CAPACITY
It makes little sense to think of investment without thinking about the overall level of capacity in an industry. We therefore start with a model that develops the optimal level of generation capacity. The presentation here is graphical,1 while there is a mathematical version in the appendix. To keep the model as simple as possible, we shall assume that there are just two types
21
22
of generation capacity available to the industry. Peaking plants have relatively low xed costs per kW of capacity, but relatively high marginal costs per MWh generated. Open-cycle gas turbines are typical peaking plants, since they are relatively quick and cheap to build, but have low thermal eciency, and generally run on expensive distillate fuels. When only a small amount of energy is required, a diesel plant may be the most ecient option. In many systems, much of the peak demand is met by output from old plants. Their avoidable xed costs are relatively low, since the costs of building them have been sunk, and the choice is between closing the plant and keeping it open. Since the cost of stang, maintaining and insuring the plant, and paying fees to the owner of the transmission system can generally be avoided if the plant is closed, it would be wrong to say that these plants have no xed costs the term going forward costs can be used to describe costs that have to be incurred if the plant is kept open. Because the plants are relatively old and inecient, their marginal fuel costs will be high, and their age may also increase their variable operations and maintenance costs. Base-load plants, in contrast, have relatively high xed costs per kW, but lower marginal costs per kWh generated. These are generally new plants, and the key decision would be whether to build more of them, incurring capital costs. The ultimate example of a base-load plant would be a nuclear station, where very high capital costs are (sometimes) oset by low running costs the French seem to have achieved a favourable trade-o, the British not. Coal- and gas-red stations generally have lower capital costs than nuclear stations, but are also generally built in order to run at high load factors, at least when new. The top panel of Figure 2.1 (panel 1) thus shows how the total costs (in per MW over the year) of our two types of capacity depend upon the number of hours a year for which the station is operated. The vertical intercept shows the stations xed costs, while the slope of the line gives the cost per MWh. If the station is required to run for T* hours or less per year, then it is cheaper to use a peaking station, while if it is required to run for more than T* hours, then a base-load station has lower total costs. The bold, kinked, line gives the lower envelope of the two linear total cost functions, showing the ecient cost of meeting a demand lasting for any given number of hours. How can we tell how many hours a particular station will be needed for? Moving down to panel 2, this gives us a load-duration curve. The hours of the year are ranked in order of the demand for electricity, so that the hour with the highest demand is placed at the left-hand end of the panel, and the hour with the lowest demand at the right-hand end. The vertical axis then shows the demand in that particular hour. The demand during the T*th
23
/MW-year
(1) Peaking Base load
T* (3) GW GW D K B (2)
Hours/year
B /MWh PR CP CB B (4)
K D GW /MWh
T* (5)
Hours/year
Max demand
K D GW
T*
Hours/year
Notes: (1) Total costs by plant type. (2) Load-duration curve. (3) Reecting line. (4) Marginal cost and demand. (5) Price-duration curve.
Figure 2.1
The determination of electricity capacity and prices
24
highest hour is thus B GW. We take this to be a gross demand for electricity, including transmission losses and the amount of plant that has to be kept part loaded, or available at short notice, for reserve. Reading the graph the other way round, we could say that there are T* hours in which this gross demand for electricity is B GW or more. This therefore implies that if we want base-load plants to meet all the demands for electricity that last for T* hours or more of the year, then we should ensure that B GW are available. In the electricity industrys traditional model, this might have been the end of the process, since prices were often set, by a regulator or a self-regulating state-owned industry, to ensure that the industry was able to recover its average costs. Sometimes, however, prices were related to marginal costs (Electricit de France was a pioneer in this), and our framework can be used to obtain marginal costs. In a market setting, sucient competition implies that prices will equal marginal costs, increasing the importance of calculating them. Panel 3 is simply a reector, allowing us to move the capacities shown on the vertical axis of panel 2 to the horizontal axis of panel 4. If the marginal (operating) cost of base-load plant is equal to CB, then this will be the marginal cost of the industry whenever demand is equal to or less than the capacity of this type of plant assumed to be B GW at the optimal solution. At somewhat higher levels of demand, then the marginal cost will equal CP, the marginal cost of peaking plant. We must now discuss the total level of capacity. The load-duration curve in panel 2 has been drawn on the assumption that the price in the rst T* hours of the year is equal to CP, and that the price for the rest of the year is CB. At a price of CP, the maximum demand for electricity would be equal to D GW. If the industry had this much capacity, then that demand could be met in full, but the price would never exceed the marginal cost of the peaking plants, and those plants would not be able to cover their xed costs. This is clearly not sustainable in a market which investors are free to leave, and will not enter without a clear expectation that they will cover their costs, including an appropriate return on capital. What would happen if capacity were not sucient to meet a demand of D GW in full? This depends in part on the details of the market arrangements for the industry. The system operator may be able to reduce the amount of reserve plant that it is carrying, so that customers are not immediately aected, although this increases the risk of failures disrupting supply to larger numbers of customers. The cost of such a failure, multiplied by its probability, gives the value of additional generation at these times. With a shortage of capacity, generators would be able to bid up to this expected cost, and it would be rational for the system operator to pay it, even when it is greater than the generators marginal operating cost. It is
25
more likely, however, that the system operator will ask for load management, and will start to pay some customers to reduce their demand. Bids from these customers might set the price in a real-time market directly, or generators might raise their bids above their marginal cost, knowing that their competition now comes not from one another (since all are needed) but from the demand side. We discuss ways in which actual markets pay for capacity in Section 5. For the time being, we can assume that there is a relationship between the actual level of available capacity and the price that will be paid for electricity in the peak hour, and that this is shown by the downward-sloping line in panel 4. It shows that if there were D GW of capacity available, the price could be CP, which was the price assumed in drawing the loadduration curve with its peak demand at D GW. With less capacity, however, the price must rise to clear the market. If there is only K GW of capacity available, then the price must rise to PR in order to ration demand to the level of capacity. This can in fact be dened as the marginal cost of power at that time, set not by variable operating costs but by the opportunity cost of a consumer that has decided to reduce its demand. This is the price level shown at the top left of panel 5, which shows a price-duration curve. At the highest-demand hour, the price must be PR, but lower prices are possible in hours with lower demands. After P hours, the load-duration curve of panel 2 shows that demand at a price of CP has fallen to K GW, and this can of course be met in full by that level of capacity. In terms of the load actually served, the load-duration curve thus has a at segment at K GW, and the area above this represents unserved load, rationed either by price or some other means. Between hour P and hour T*, therefore, the price in a perfectly competitive market would be equal to CP, the marginal cost of the peaking plants which are at the margin, but not fully employed. From hour T* onwards, none of the peaking plants is needed, and the marginal cost falls to the marginal cost of the base-load plants, at CB. In a competitive market, this would also be the price. What about the total revenue received by any power station? For this, we need to return to panel 1. Remember that the marginal cost of a plant (reected in the level of the lines in panels 4 and 5) is given by the slope of its line in panel 1. Since the price between hour P and hour T* is equal to the marginal cost of peaking plants, the slope of a total revenue line (per MW per year) would be identical to the slope of the total cost line for peaking plants over this period. (The price is normally given in /MWh, while the axes of panel 1 are in /MW per year and hours per year, so that a line in this panel has a slope measured in /MWh.) Similarly, from hour T* onwards, the slope of a total revenue line would be given by the marginal
26
cost of a base-load plant, and this is also the slope of the ecient cost envelope over these hours. If the market price never exceeded CP, then the total revenue line would start from the origin and run parallel to the ecient cost frontier throughout its length. Peaking plants would cover their variable costs but make no contribution towards their xed costs. Base-load plants would make a contribution towards their xed costs, since the price exceeds their variable costs for the rst T* hours of the year, but there would be a shortfall equal to the xed costs of the peaking plants. This shows that if we had D GW of capacity, it would not be able to cover its costs in a competitive market system. With less capacity, however, the price will exceed CP for the hours with the highest levels of demand, and this will allow all the stations to make an additional contribution towards their xed costs. The slope of a total revenue line will be greater than the slope of the ecient cost envelope, and one possible line is shown rising from the origin to meet the ecient envelope at P hours. The slope of this line is given by the level of prices in panel 5. The implication is that the prices of panel 5 would be just sucient for a peaking plant to cover its xed costs, plus its variable costs of running for P hours. Since the load-duration curve shows that with a total capacity of K GW, all the peaking plants will need to run for at least this number of hours, this means that they would all be able to cover these costs. The total revenue line would then continue upwards at a rate of CP / MW per hour, which is the market price between hour P and hour T*, and also the slope of the ecient cost envelope. In other words, the total revenue line is now superimposed on the ecient total cost line. After hour T*, the slope of the total revenue line falls to CB /MW per hour, just like the slope of the ecient cost envelope. This shows that if the peaking plants with the shortest periods of operation are able to cover their total costs, and if the plant mix is ecient, then all the other plants will just be able to cover their total costs from market prices based on marginal costs. (See Box 2.1) We can summarise the links explained in Box 2.1 between the amount of capacity and the protability of that capacity. If there is too little capacity in total, then peaking capacity will be making supernormal prots, while if there is too much capacity overall, then peaking plants will be making losses. If we have the right amount of base-load capacity, then these stations will be making the same amount of prot (per MW-year) as peaking stations. If there is too much base-load capacity, then base-load stations will make less prot per MW-year than peaking stations. If there is too little base-load capacity, then these stations will make more prot per MW-year than peaking stations. In this simple model, there are three reasons for investment in new capacity. The rst is that the actual level of (a given type of) capacity is less than
27
/MW-year
T* /MW-year (2) Peaking Base load /MW-year (3)
Hours/year
Peaking
Base load
T* /MW-year (4)
Hours/year /MW-year Base load
T* (5)
Hours/year
Peaking
Peaking
Base load
T*
Hours/year
T*
Hours/year
Notes: (1) Optimal capacity mix. (2) Right total, too little base load. (3) Right total, too much base load. (4) Total too large, right base load. (5) Total too small, too much base load.
Figure 2.2
How the capacity mix aects revenues
28
BOX 2.1
WHAT IF THERE IS THE WRONG LEVEL, OR MIX, OF CAPACITY?
The protability of peaking plants depends on the total amount of capacity, rather than on the amount of peaking plant in the industry. To see this, consider what would happen if the total capacity remained at K GW, but that there was less than B GW of base-load capacity (and hence more peaking capacity than in the optimal plant mix). This is the situation shown in panel 2 of Figure 2.2. (Panel 1 repeats the top panel of Figure 2.1, showing the situation with the optimal level and mix of capacity.) For the highest-demand T* hours of the year, the situation is exactly the same as in Figure 2.1, and so the total revenue line rapidly rises to join the total cost line for peaking plants, and then runs on top of it. Things change after T * hours, however. With too little base-load capacity, some of the peaking plants will be running for more than T * hours, when demand is still too high to be met entirely by the reduced number of base-load plants. In those hours, the price will be equal to CP, and the total revenue line (in bold) continues with the same slope as before. Only once demand has fallen to the level of base-load capacity, at some point to the right of hour T *, does the price fall to CB, and the total revenue line atten to run parallel to the total cost line for base-load plants. Because the total revenue line is above the total cost line, all of the base-load stations are making a supernormal prot, over and above their costs. Because the lines are parallel, the prot earned by a base-load plant does not depend on how many hours it runs (as long as it runs in all the hours when the price is above CB). What If the Mix of Capacity is Wrong? We might have a situation with the right amount of capacity in total, but too many base-load plants (and hence too few peaking plants), for example. This is shown in panel 3 of Figure 2.2. Once again, the prices for the hours with the very highest demands are unchanged, and the total revenue line meets the total cost line for peaking plants while all of those plants are running. Base-load plants become marginal before the demand has fallen to B GW, however, and so the price falls to CB at a point to the left of hour T *. The total revenue line then runs parallel to, but below, the total cost line of base-load plants. The peaking plants are covering their
29
costs in full, but the base-load plants are making losses. Once again, in this simple example, the scale of the losses does not depend on exactly how many hours the base-load plant runs, within the range where the model shows them to be marginal. What If the Level of Capacity is Wrong? Panels 2 and 3 thus show that having the wrong mix of base-load and peaking plants does not affect the protability of peaking plants, as long as the industry has the correct overall level of capacity. If the industry has the wrong level of capacity, however, then the peaking plants prots will be affected. In panel 4, there are B GW of base-load plants, but total capacity is assumed to be between K GW and D GW. This means that there are some hours in which capacity is insufcient to meet demand in full at a price of CP, but fewer than P of them. The price exceeds CP during these hours, and so the total revenue line starts with a steeper slope, but the revenues (over and above variable costs) are not sufcient to cover the xed costs of a peaking plant. At the point where capacity is sufcient to meet demand in full at a price of CP, the total revenue curve becomes less steep, running parallel to, and below, the total cost line for the peaking plant. At hour T *, base-load plants become marginal, and the total revenue line attens again. This implies that both types of capacity make the same loss, measured in /MW per year. Even though we have the right amount of baseload capacity, the surplus of total capacity forces them into loss. What If the Level and Mix of Capacity are Wrong? Finally, panel 5 shows that it is possible for peaking capacity to make a prot, but for base-load capacity merely to break even, if we have the wrong amount of both total capacity and base-load capacity. In this case, a shortage of total capacity means that prices exceed CP for more than P hours, and so the total revenue line moves above the total cost line for peaking plants. It then runs parallel to that line for some hours, until demand falls to the level of base-load capacity, and those plants become marginal. This panel has been drawn so that this occurs just as the point where the total revenue line meets the total cost line for base-load plants. This is to the left of hour T *, which implies that there is too much base-load capacity. That would normally drive all the base-load plants into loss, because prices would be at CB for too much of the
30
year. As the panel has been drawn, this is exactly cancelled out by the higher prices in the very peak hours. It should be obvious that this is a rather special case. If there was a slightly lower proportion of base-load plants (within the given total capacity) then they would also be making a prot, while a slightly higher proportion (implying far too little peaking plant) would mean that the base-load plants were actually making losses.
the optimal level. The second is that some capacity has reached the end of its physical working life and must be replaced. The third reason is because changing relative costs make it economic to replace older capacity with a more ecient station. This could be the result of technological progress, and is more properly a matter for Chapter 4. The model implies that all three types of (optimal) investment will be protable for the rm that undertakes them. In a competitive market, we should therefore expect that rms would be willing to make these investments. When should capacity be reduced? In this model, if we have too much capacity, then it will not be able to cover its full economic costs including an appropriate return on capital. That is certainly a signal that no new investment should be made. In the absence of sunk costs, it is also a signal that capacity should be withdrawn from the market. Electricity generation, however, is subject to signicant sunk costs, in that power stations are capital intensive, have no real alternative use, their output can only be moved to a new market if transmission capacity is available, and the costs of physically moving a power station will generally be prohibitive. In the presence of sunk costs, it is no longer optimal to close a station as soon as it is unable to make a full return on the capital invested, as long as it is at least covering its variable costs. This drives a wedge between the prices at which entry becomes protable and those at which exit is sensible. The consequences of this wedge, and more generally of the need to make irreversible investments in an uncertain world, are considered in the next section.
3.
IRREVERSIBILITY AND UNCERTAINTY
We have just pointed out that electricity generation involves signicant sunk costs. The electricity industry is also subject to considerable uncertainty, over fuel prices and the level of demand. While fuel prices mostly aect the choice between dierent types of power station, the level of
31
demand feeds in to the total capacity required. Two types of demand uncertainty are relevant here. First, the short-term level of demand, relative to its trend, depends on the weather, and the state of any interconnected power systems net demand in an area will be higher if an adjoining area that normally exports to it is short of capacity, as California found to its cost when hydro-electric generators in the Pacic North West were short of water in 2000. Second, the trend itself is uncertain, depending on consumers reactions to prices, the level of economic activity and technological change. To some extent, short-term uctuations in demand can be hedged. A warm winter will reduce demand and lower prices relative to their expected level, but generators and retailers can sign contracts that x their revenues (for the contracted level of output) at the expected price level.2 A cold winter may be more of a problem, in case it raises demand above the level of available capacity, but contracts can once again largely insulate retailers from the nancial consequences of this uncertainty. Uncertainty over the trend level of demand creates more problems. If capacity is built to meet a need that does not arise, then prices may be depressed for several years, until demand rises or other plant is retired. Retiring plant, however, is an irreversible decision, and so generators will not lightly close a plant for good. Even mothballing a plant in the hope of reopening it later involves some costs, and so generators will not withdraw capacity from the market as soon as prices start to fall below their variable costs. The implication of this is that it is not sensible to make investments at the rst signs of an increase in demand, in case it is not sustained. Dixit and Pindyck (1994) consider the theory of irreversible investment under uncertainty. Their key point is that a rm that has not made an irreversible investment has the option of making the investment at some later time, and it gives up this (so-called real) option once it makes the investment. Since the rm is giving up the value of the option as well as the sunk cost of the investment itself, it is best to wait until the expected value of the investment exceeds the sum of its direct cost and its option value. Consider a rm which is considering whether to build a power station. In the rst period of its life, it can make a prot before xed costs of 100. In all subsequent periods, we expect the station to make a prot of 100 as well. The discount rate is 10 per cent, and so the value of this expected prot stream is 1,000 (we get a simpler present value if we ignore the fact that the station will not actually last for ever). If the cost of the station is 800, then building it now will lead to an expected prot of 200. On the traditional analysis, this is clearly the sensible decision. Assume, however, that our expected prot of 100 per period after the rst period is actually the average of a prot of 50, and of a prot of 150, each expected with a probability of 0.5. Furthermore, assume that if the
32
generator waits until the end of the rst period until it invests, it will know which of these will be the case. In the low-demand state, with a prot of only 50 per period, investing in the second period would not be protable a revenue stream of 500, discounted to the second period, is clearly less than the investment cost of 800, incurred at that time. In the high-demand state, investment is protable, since the investment of 800 brings in revenues with a present value of 1,500. The rms expected prot from a second-period investment is 350 there is a 50 per cent chance that it can make an investment with a prot of 700, while it will lose no money if demand is low and it does not invest. Discounting this expected prot to the rst period, we get an expected prot of 315. This is greater than the expected prot from investing at once, by some 115. This gure is the option value of waiting one period until the uncertainty is resolved. If the rm follows the traditional analysis and invests at once, it is giving up this option value, and hence the chance to wait and avoid the mistake of investing too early. This does not mean that waiting is always optimal. If the rm had to wait 6 periods until it knew whether demand was going to be high or low for the rest of time, then its expected prot from making the investment decision at this time will still be 350, but this must now be discounted by a factor of 1.16. This gives a present value of 350/1.16, or 197.57, which is less than the present value of investing now. Alternatively, assume that the expected value of revenues in each period is 140, with an equal chance of revenues of 70 and 210 after the rst period. In the second period, it would still be optimal not to invest if demand turns out to be low, and so investing only in the case of high demand brings expected prots of 650 (equal to 0.5 (2,100 800)). Discounting this to the rst period gives an expected prot of 590.91. If the rm invests at once, however, it gets an expected prot of 600, which is greater. The option value of waiting is now outweighed by the immediate prots to be gained. This leads us towards the kind of investment rule that Dixit and Pindyck derive investment is optimal if conditions are at the level where the current prot to be had from investing now outweighs the benet of waiting for more information. They express this as a trigger price investment rule the rm should invest only once the variable reaches a particular level. This is a feature of the stochastic processes that they use, in which the future path of a variable depends only upon its current level. A more complex stochastic process, in which the future depended on the path taken by the variable, would require a more complicated investment rule. Note that the trigger conditions for investment will generally be well above the minimal conditions at which investment would just be protable,
33
if the expected future conditions were sure to be realised. This is not only the result of a market-based system, however, for Dixit and Pindyck show that the social welfare-maximising decision rule is identical to the rule that would be adopted in a competitive market. Furthermore, at least for a range of problems, the competitive solution is to invest at the same trigger point as in the monopoly solution. The calculations involved are dierent a competitive rm has no option value of waiting, because if it does not make a protable investment, it can safely assume that some other rm will do so. With enough potential entrants, the price will never rise above the trigger for further investment, but random movements in demand can be expected to depress the price below this level. The trigger price must therefore be high enough for the rms to expect to just cover their costs if they invest at this price, realising that it will be a maximum and they will obtain a rather lower average price. This clearly implies that the trigger price must exceed their average costs. Interestingly, the perfectly competitive rm can perform some of its calculations as if there would be no further entry (Dixit and Pindyck, 1994, p. 291). Real options theory can also be used to determine the point at which capacity should leave the system. This can happen in two ways. First, capacity can be mothballed, with the possibility of returning it to service later. Mothballing generally involves an ongoing maintenance cost, together with costs when the plant is prepared for mothballing and then returned to service. Second, a plant can be scrapped, which is an irreversible decision for that plant, but may not preclude the possibility of making a completely new investment if conditions improve. As long as putting a plant in mothballs and then taking it out involves some xed costs, it should be obvious that it will not be optimal to mothball a plant as soon as the price falls below its variable costs (minus the cost of maintaining it in mothballs). Similarly, an even lower price is required to make total scrapping optimal. To the extent that withdrawing capacity from the market is likely to raise prices, this means that prices are likely to stay depressed for longer than if there were no sunk costs. On the other hand, the option to mothball a plant, once it has been built, reduces the potential losses from a period of depressed prices, and therefore raises the value of making the initial investment. This tends to encourage investment, compared to a situation in which mothballing is not possible. So far, we have assumed that decisions can be implemented as soon as they are taken. Bar-Ilan and Strange (1996) consider the impact of investment lags it normally takes several years to build a power station, and the pre-construction formalities, such as obtaining planning permission (zoning permits) can add considerably to this time.3 When these lags exist, uncertainty can actually encourage investment, in the sense of reducing the
34
trigger price at which it is optimal to start construction. The intuition is that for a given level of uncertainty and of the current price, the expected prot from starting operation in, say, four years time is greater than the expected prot from starting now. The price might rise over the intervening period, which would raise prots, while the downside risk from a price fall is limited by the option of abandoning the project. While waiting to start the investment still has a value, in that the rm learns more about the evolution of prices, it involves a cost, in that if prices rise strongly, the rm will not be able to take advantage of these immediately, and will have missed out on some potential prots. A higher level of uncertainty increases the expected value of the benet from high prices, while the losses from low prices are still capped by the option of abandonment. Bar-Ilan and Strange show that for some parameter values, the overall eect of an investment lag is to lower the trigger price at which investment is started to below the price that would trigger investment in a world of certainty. For a time to build of three or four years, however, which should be more than sucient for a combinedcycle gas turbine, uncertainty still raises the trigger price at which it is optimal to invest. Bar-Ilan et al. (2002) obtain similar results in a model which explicitly studies the problem faced by a (traditional) electric utility. The utility faces a cost of carrying excess capacity, and a (possibly dierent) cost of insucient capacity, and must decide when, and how much, capacity to build in order to meet its stochastically evolving demand. The utility must incur a xed cost each time it adds capacity, as well as the per-unit cost of the capacity. If construction lags are short, then uncertainty leads the utility to wait longer before adding capacity (that is, it will not invest until the level of demand is higher, relative to its existing capacity), although it will then add a larger increment of capacity. If construction lags are long, however, the utility may optimally choose to invest when it has a higher level of spare capacity (with growing demand, this eectively means investing earlier), and to build less capacity each time. Once again, however, a very long lag is required for the utility to nd it optimal to invest earlier than it would do in a world of certainty. Finally, we should perhaps ask how applicable the model of a stochastic electricity price is. The impact of the real options approach to investment depends upon the level of uncertainty if there is very little uncertainty, then option values are low. If the level of capacity adjusts quickly to the level of demand, then this would produce an electricity price that was always close to the cost of generation. In technical terms, the price would be strongly mean reverting. This again reduces the level of uncertainty, and hence the value of real options. The extent and speed of mean reversion is, in principle, an empirical question, and one that requires rather more data
35
than we have available, given the relatively recent spread of electricity markets across the world. Pindyck (1999), however, shows that while many fuel prices have been mean reverting over the long term, their short-term movements can be considered as Brownian motion. Furthermore, a model with mothballing and entry will tend to put a oor and a ceiling on the price of electricity, but the price can still vary over an economically signicant range. Our conclusion is that the real options approach does have lessons for investment in power stations. This conclusion will be strengthened when we consider the propensity of many industries to invest in waves, creating cycles of over capacity and low prices, followed by inadequate capacity and high prices.
4.
INVESTMENT CYCLES
Many capital-intensive industries are prone to investment cycles. At rst, the industry may be short of capacity and prices will be high. This acts as a signal to investors, who start to add capacity. In the absence of a coordination device, however, they are in danger of over-reacting too many investors read the high prices as a signal that their own investment will be protable, and somehow fail to take the likely actions of others into account. Once the new capacity comes on stream, it will depress prices. This will be sucient to halt most new investments, but the existing capacity is likely to stay in service. Scrapping decisions are irreversible and will not be taken unless the price falls suciently below the variable costs of staying in operation. Furthermore, in a competitive industry there can be coordination problems all rms can agree that some capacity should be taken out of service, but the rm that actually does so will incur a loss that it would prefer others to bear. Eventually, however, capacity will fall as plants are retired, or demand will rise. As the margin between capacity and demand narrows, prices will rise again, up to the point where investment is again perceived as protable. The industry is then in danger of repeating the cycle. Industries such as oil tankers and copper are well known for exhibiting these kinds of cycle (Hawdon, 1978; Brennan and Schwartz, 1985). There have also been fears that the electricity industry could be vulnerable to similar problems. Bunn and Larsen (1992, 1994) discuss investment in England and Wales under the Pool system, which linked prices explicitly to the degree of spare capacity. In each half-hour, a capacity payment equal to the calculated loss of load probability, multiplied by the value of lost load (see Box 2.2) less the bid-based price of energy, was paid to all scheduled generators, and a similar payment was made to all those generators that were available but not scheduled to operate. These payments rise
36
BOX 2.2
THE VALUE OF LOST LOAD AND THE LOSS OF LOAD PROBABILITY
The value of lost load was used by system planners as an estimate of the amount that consumers would be willing to pay in order to avoid a power cut, and hence as an input into deciding the value of additional capacity. It can be estimated by surveys, or by observing actual behaviour, such as the point at which customers on realtime tariffs start to reduce load, or the amount that companies are willing to spend on their own back-up generators. One important issue is that the value of lost load varies from customer to customer, and depends upon the length of an interruption the cost per kWh of a short failure may be signicantly less than the cost of a long failure. The loss of load probability is the probability that demand (including reserve) will exceed the available capacity, and that it will be necessary to shed load. This depends upon the level and variability of demand, and the level and reliability of capacity. Integrated over time, it gives the expected number of power cuts. A related concept is the expected amount of unserved energy the number of power cuts multiplied by their length and duration. Traditionally, system planners were required to minimise the cost of the system while keeping either the expected number of outages or the amount of unserved energy below target levels. The value of lost load gives the gross value of additional power when the system is short of capacity and customers are being rationed. Since the variable cost of production must be incurred to provide this power, the net value of additional power (and hence of the capacity that could produce it) is equal to the gross value, less the variable cost. The expected value of capacity is thus the value of lost load, less the plants variable cost, multiplied by the loss of load probability. This, using the plants bid in place of variable cost, is the unscheduled availability payment that was given to all plant that was available but not actually generating in the Electricity Pool of England and Wales. A plant that was scheduled to generate received a capacity payment equal to the value of lost load, minus the marginal price of power, the system marginal price, multiplied by the loss of load probability. In theory, these payments represented the true value of capacity in each half-hour, and the amount that consumers should pay towards its costs.
37
sharply when the margin of spare capacity falls, for the loss of load probability is very non-linear in spare capacity. If investors are over-sensitive to the current level of capacity payments, compared to a long-run equilibrium level, then too much capacity will be ordered during times of shortage when capacity payments are high, leading to a surplus of plant and negligible payments in a few years time. Ford (1999, 2002) discusses similar issues in the context of the United States, and particularly the western states. Writing before the California crisis, he predicts (1999) that if generators had to rely on the (now-defunct) Power Exchanges energy price for all of their revenues, then they would be able to cover their full costs only at times of relative shortage. The sensitivity of revenues to the level of spare capacity, coupled with delays in the permitting and construction of power plants, would lead to alternating periods of high and low prices. His model suggested that these could be largely eliminated by introducing a xed (in US$/MWh) capacity payment, paid to all generators. A payment of US$5/MWh, for example, would stabilise generators earnings and allow them to cover their full costs without requiring the Power Exchange price to rise in shortage conditions. In the long-run equilibrium, there would be little dierence between the average price with the capacity payments and without, although prices would be higher in the short run with the capacity payment scheme. These higher prices presumably encouraged sucient extra investment to create a cushion of spare capacity that allows the energy price to stabilise at the level of marginal cost, with little or none of the scarcity element shown to the top left of panel 5 of Figure 2.1. Ford (2002) argues that investors overbuilding is typically the result of not taking into account the eect of others investment decisions on the market price. They may well have accurate expectations about the price and how it will respond to their own investments, but are not able to keep track of other investors decisions during an investment boom. As a result, too much capacity is added, sowing the seeds for a following bust and future shortage Ford predicted that California could suer further shortages of power in 2007, despite the large amount of capacity added just after the crisis of 200001. Are electricity markets fated to suer from such cycles? Mothballing and long-term contracts may help to smooth them. In England and Wales, nearly 10 per cent of the industrys capacity has at times been mothballed, unavailable for short-term use (and freed from the requirement to pay transmission charges), but capable of being returned to service should market conditions change. Putting older plant into mothballs as a wave of new capacity enters the market can help to put a oor under prices. It should be noted, however, that since mothballing a plant and then returning it to
38
service involves sunk costs, the real options approach implies that this price oor will be below the level of the plants variable costs. Returning the plant to service can help to slow the increase of prices once demand rises relative to capacity. Since the costs of taking a plant out of mothballs are much less than those of a new plant, the price needed to trigger new investment will be higher than the price needed to return a mothballed plant to service. This implies that by the time investors are rationally contemplating new investment, there should be few mothballed plants to cushion further price rises.4 Long-term contracts may help to reduce investment cycles in two ways. First, they can reduce uncertainty for the entrants, allowing them to lock in a price, or a margin between their fuel costs and their selling price. This will lower the price needed to trigger investment. First, it can reduce the entrants cost of capital, by allowing higher gearing, since their cash ows are less uncertain. Second, less uncertainty reduces the value of waiting before investing, and lowers the trigger price for investment. The second way in which long-term contracts may help to reduce investment cycles is as a coordination device. The demand for long-term contracts comes from electricity retailers, and will depend on their estimates of the future demand for power. While one cause of overinvestment may be the independent decisions made by investors who are not fully aware of each others plans, the retailer knows its own demand for power. Investors who are unable to nd a retailer to contract with will not be able to take their project forward on that basis. This does not mean that they will have to abandon the project, however, because it is still possible to invest as a merchant plant, selling in shorter-term markets. With higher risks, they will need less gearing, which will raise their cost of capital. Furthermore, the higher uncertainty raises the trigger price at which investment becomes optimal. If retailers are signing a lot of long-term contracts, there may be enough capacity to keep prices below this trigger, so that merchant plants are neither needed nor observed. If retailers are reluctant to sign long-term contracts, however, then there would not be enough capacity without merchant plants. Prices will have to reach levels that can trigger investment in these plants, and without the coordination device of contracts, there is a danger that the reaction to high prices could be excessive. Why would retailers be reluctant to sign long-term contracts? In the past, such contracts acted as a hedge on their purchase costs and reduced their risks. If retailers face competition, however, then long-term commitments can increase their risks (Newbery, 2002). If the wholesale market price of power falls, then entrants to retailing can take customers from the incumbents, unless the incumbents reduce their prices in line with the wholesale price.5 If the retailers selling prices depend upon the current wholesale price, but their costs depend upon their long-term contracts, then their
39
prots become more risky. The best response in this situation is to reduce the proportion of their demand that they buy on long-term contracts. Green (2004) models this process, in a setting where the wholesale market is imperfectly competitive and the volume of contracts aects the mark-up over marginal costs, rather than the amount of capacity. He shows that it can lead to a quantitatively signicant impact on contracting and on wholesale market prices.
5.
PAYING FOR CAPACITY IN PRACTICE6
The analysis so far has been based on the idea that prices in the highestdemand periods can rise above the marginal operating cost of the peak generators, and that they will rise by more when capacity is relatively scarce. How do markets around the world put this idea into practice? The key distinction is between markets that are energy only and those in which there are explicit arrangements to pay for capacity. Australia has an energy-only market with a price cap set at A$10,000/MWh. If the system operator ever has to shed load because of a lack of capacity, the price is set to this level. At other times, the market price is set by the marginal generators bid. When capacity is scarce, but sucient to meet demand in full, generators are likely to raise their bids above their marginal operating costs, recovering part of their xed costs. In practice, most generators and retailers are likely to hedge most of their expected trades in advance. This stabilises the generators revenues, and minimises the retailers exposure to prices close to the cap. The price at which trades are hedged will depend on how often spot prices are expected to approach the cap, which will depend on the level of spare capacity relative to normal levels of demand. The level of the cap is critical to the success or failure of the scheme. If the cap is too low, generators will be unable to cover their costs unless power shortages are frequent, which is unlikely to be politically acceptable. If the cap is too high, the problem will not be one of cost recovery but of excessive prices. In many markets, a price cap has been imposed as a way of mitigating market power. The problem then is that one instrument is trying to do two tasks holding prices down when there is enough capacity but not enough competition, and allowing generators to recover their xed costs when capacity is short. Not surprisingly, this is too much. In many US markets, regulators have enforced relatively low price caps and recognised that they will need to nd other ways of recovering xed costs. In Australia, the government actually raised the price cap from A$5,000/MWh to A$10,000/MWh in 2002, in response to fears that the lower gure would not provide sucient incentives to generators to keep capacity available.
40
The calculations which Australian market participants make if they choose to hedge are similar in spirit to those which underlay the capacity payment in the former Electricity Pool of England and Wales. As mentioned above, this payment was based on the net value of lost load, multiplied by the loss of load probability. This is eectively the expected value of payments under a strict version of the Australian system, in which prices rose above marginal operating cost only when load was actually lost, and assuming that the price cap was set at the value of lost load. The Pool became discredited in practice (Oer, 1998), but in principle, capacity payments should have given the correct incentives for investment and closure decisions a plants expected capacity revenues (which could easily be hedged for at least a year in advance) depended on the value of the power shortages that it would prevent. The value of lost load and the loss of load probability were discussed in Box.2.2, above. The main alternative to an energy-only market is to have a market for energy and a separate market for capacity. The system operator, or a regulator, denes the amount of capacity that every retailer must have access to, relative to the load that it serves. Typically, this requirement for installed capacity will exceed the retailers expected peak demand, since the actual peak may be higher than expected, and not all capacity will be available at the peak time. In order to meet the requirement, generators can sell, and retailers buy, capacity credits. A retailer that does not have enough credits has to pay a penalty. This naturally caps the price that would be paid for capacity credits, and so the penalty should be related to the cost of making capacity available. To the extent that the average price of capacity credits will be less than the penalty, the penalty should exceed the xed cost of capacity; to the extent that generators will have other revenues, the penalty can be reduced. In the PJM (PennsylvaniaNew JerseyMaryland) market, the penalty is set to equal the xed costs of a peaking plant. An installed capacity market obviously provides incentives to make capacity available. If capacity is relatively plentiful, then retailers can buy the full amount of credits required, and the price will be the avoidable cost of keeping plants open. (This assumes that plants with insucient revenues from capacity credits would close.) If capacity becomes short, so that some retailers are facing the penalty, this will set the price of unsold capacity credits. The price of capacity credits can thus vary from the cost of keeping old plant open to the cost of building new plant as long as the penalty has been set at an appropriate level. If so, the market can send appropriate signals about investment and closure decisions investment should be protable when the market falls short of its installed capacity requirement. The level of the installed capacity requirement is therefore critical in determining the amount of capacity which the industry will be given incentives to
41
provide. It is an administrative decision, just like the administrative decision on the level of a price cap in an energy-only market. The advantage of choosing an installed capacity market is that the connection between the decision and the risk of power cuts should be clear, as opposed to the rather indirect link between a price cap and the resulting level of capacity. In practice, however, administrative decisions are common throughout electricity markets. System operators have to decide how much reserve to buy from day to day. Generators, and consumers able to reduce load at short notice, may form the supply side of an active market, but the demand side is missing. Even if consumers knew their willingness to pay to avoid having their own load cut o, there is a public good aspect to reserves, in that inadequate levels of reserves can result in a system-wide failure. If the market grows short of reserves, system operators will reduce their requirements to a minimum, raise prices and start to ration consumers through rolling blackouts, aiming still to avoid a system breakdown. It is not clear how prices are set in these circumstances. Economists need to learn more about how electricity system operations determine prices at peak times. Joskow and Tirole (2006) set out some of the key issues involved, but there is room for further study.
6.
INVESTMENT IN LIBERALISED MARKETS
We conclude this chapter with a brief look at investment patterns in a number of liberalised electricity markets. Figure 2.3 shows investment in England and Wales since the reforms of 1990. The top line shows the capacity margin the excess of generation capacity over peak demand (weather adjusted), as a proportion of that peak demand. This started the period at more than 30 per cent, above the planning margin of 28 per cent used by the nationalised Central Electricity Generating Board. With an apparently high level of spare capacity, we might expect little investment, but the second line in the gure clearly shows a stream of new investment during the 1990s, amounting to nearly half the peak demand. Most of this was in combined-cycle gas turbine (CCGT) stations, oering much greater thermal eciency and lower emissions than the industrys existing coaland oil-red stations. At the time of the restructuring, nearly 80 per cent of the industrys capacity was owned by just two generating companies, and these built some CCGT stations in order to modernise their portfolios and reduce their emissions of sulphur dioxide. Most of the new capacity was built by entrants to generation, however. The rst stations were mostly built by the regional electricity companies, the former distribution and retailing monopolies. The stations gave them some insurance against the market
42
Proportion of peak demand 0.3 0.2 0.1 0 0.1 1991/92
Capacity margin Gross investment Net investment
1994/95
1997/98
2000/01
2003/04
Source: National Grid Company (NGC).
Figure 2.3
Investment in England and Wales
power of the major generators, by diversifying their purchases, and allowed them to earn some unregulated income. Both these stations and a later wave of unaliated entrants expected to benet from the major generators market power, in that prices in the wholesale market remained above the levels needed to remunerate new entrants for most of the 1990s. The bottom line of Figure 2.3 shows that net investment has been well below gross investment in other words, many plants have been closed over the period. Most of these closures were by the major generators, making way for the new entrants and keeping plant margins from rising. Capacity margins actually declined steadily, but not excessively, through the 1990s. The small amount of net investment was not sucient to keep up with demand growth. The last major additions to capacity came in the nancial year 2000/01, and with little investment and some closures since then, the capacity margin has fallen sharply. The margin between wholesale prices and generators costs also fell in 2000. A period with little investment was an appropriate response to these prices, and wholesale margins have risen recently, reecting the tighter market. It remains to be seen when investment will recover, although two of the largest companies have recently announced projects. Figure 2.4 shows investment in Finland, which liberalised its market in 1996.7 There have been very few plant closures, so that gross and net investment are almost equal. There is a signicant amount of investment in the years immediately after liberalisation (which will have been planned before liberalisation took eect, but during a period when it was becoming
43
Proportion of peak demand 0.3 0.2 0.1 0 0.1 1992

Source: Nordel.
1994
1996
1998
2000 2002/03
Figure 2.4
Investment in Finland
increasingly likely), and very little plant has been commissioned since then. Once again, capacity margins have been falling, but from a relatively high level. However, more plants will be needed as demand grows and old stations retire, and a consortium of electricity companies and large industrial consumers is building a 1,600 MW nuclear reactor (work started in the spring of 2005). This is believed to oer the lowest cost option for the base-load supplies of power that the industrial users need. Since the consumers are part of the consortium building the station, they will be insulated from the way in which market prices in Nord Pool depend on the level of rainfall. Figure 2.5 shows the situation in Norway, which has also had little investment. To some extent, this reects the lack of suitable sites for new hydroelectric schemes. A national debate on whether to build gas-red power stations, which would increase the countrys carbon dioxide emissions (but might reduce those of the world), has not yet reached a conclusion. Capacity margins fell sharply in the rst half of the 1990s, after liberalisation in 1991. They have risen since 1997, but this is because the peak demand has been falling, rather than because of signicant increases in capacity. As a hydro-dominated system, Norway will always require a relatively high margin of generation capacity over peak demand, because the average hydro-electric power station can only store enough water to operate for about half the year. If the average level of demand is 70 per cent of the peak demand, building enough hydro storage capacity to meet the annual demand for energy will imply enough generating capacity to meet 140 per cent of the peak demand. The level of trade among the Nordic countries
44
Proportion of peak demand 0.5 0.4 0.3 0.2 0.1 0 1992

Source: Nordel.
1994
1996
1998
2000 2002/03
Figure 2.5 Proportion of peak demand 0.4 0.3 0.2 0.1 0 0.1
Investment in Norway
1992
Source: Nordel.
1994
1996
1998
2000 2002/03
Figure 2.6
Investment in Sweden
has risen over the 1990s, and Norway has made up for a decline in its capacity margins by importing increasing amounts of power in dry years. There has been very little investment in Sweden, shown in Figure 2.6. Again, capacity margins were very high at the start of the 1990s the country has a mix of hydro and thermal resources, so the appropriate margin will be
45
Proportion of peak demand 0.3 0.2 0.1 0 0.1 1992 1994 1996 1998 2000 2002/03
Source: US Energy Information Agency.
Figure 2.7
Investment in the United States
lower than for Norway. Shortly after liberalisation in 1996, a signicant amount of old capacity was retired, and capacity margins fell. As a member of Nord Pool, Sweden may be able to import power if its own supplies are inadequate, and so the desirable level of reserve capacity within its borders fell after 1996. To the extent that the Nordic countries share common weather patterns, however, the same shock that makes one country want to increase its imports may be aecting its neighbours. This will reduce, but by no means eliminate, the insurance benets of the multi-country system. However, transmission constraints can mean that capacity in another country is not able to respond to failures, and more localised reserve is needed. The Swedish system operator, Svenska Kraftnt, was concerned by the low margins, and has been paying for stations in a capacity reserve. This is intended as a temporary measure, but no permanent solution has yet been agreed. Plants entering the reserve (mostly being brought back into service after mothballing in 1998 and 1999) accounted for nearly three-quarters of the capacity added between 2001 and 2003. Finally, we come to the United States, shown in Figure 2.7. Note that the gures given are nationwide averages, and the picture in individual regions may be quite dierent in particular, low capacity margins in the western states contributed to the California crisis of 200001. On the national scale, however, capacity margins were high in the early 1990s, and there was accordingly little net investment. The lack of investment lasted until 1998, while growing demand eroded the margin of spare capacity. Investment took o at the end of the decade, however, with new capacity equal to over 20 per cent
46
of peak demand added between 1999 and 2002. Three-fths of this was CCGT capacity, and most of the other plants were peaking combustion turbines. Building combustion turbines is generally the quickest way to deal with a shortage of capacity, while the high thermal eciency of CCGTs has made them the technology of choice in many markets. In the mid-1990s, most states were debating whether to create electricity markets, and this regulatory uncertainty caused most investors to wait until policy had been decided. Once the uncertainty was resolved, a combination of low capacity margins and the technological opportunity presented by CCGTs made a sharp rise in investment optimal. In many markets, the CCGTs were expecting to replace stations burning the same fuel (natural gas) but at a thermal eciency one-third lower. It is likely, however, that the dramatic increase actually seen went well beyond the optimal level of investment. Electricity companies saw their share prices soar at the end of the 1990s, and investment decisions were made in the atmosphere of a stock market bubble that subsequently collapsed. The incumbent utilities chose not to close the older plants that the entrants were expecting to displace, and the price of natural gas rose sharply, depressing the stations load factors as those burning other fuels became more competitive. As in England and Wales, a number of the new plants have suered from nancial problems. How can we sum up these experiences? Clearly, each country is aected by its own specic background, making generalisation dangerous. Liberalisation has often been accompanied by a reduction in capacity margins, but these margins have not generally fallen to a level that posed a danger to security of supply.8 When capacity margins become very low, investment has followed. A high capacity margin is generally a signal that little investment is needed, and so we should not be worried about the examples that combine little investment with high margins. It is worth noting, however, that we still have relatively little experience, given the long asset lives of power stations, and so it is early to draw rm conclusions. In particular, we do not yet know how willing investors will be to re-enter the markets of the UK and the US when new capacity is needed, given the losses some of them have suered over the last few years.
7.
CONCLUSIONS
This chapter has presented a theoretical model showing how the optimal level and mix of capacity can be derived for an electricity industry without uncertainty. In the model, investors will receive normal prots when the industry has the correct level of capacity, while a surplus will depress their revenues, sending a signal that some capacity should leave the market. If
47
there is a shortage of capacity, which is the more worrying case for our purposes, then prices will rise and existing plants will make supernormal prots, which should act as a strong signal for new investment. In other words, the market is capable of sending the correct signals for the socially optimal level of capacity. The real world is not certain, of course. The real options approach to investment shows that it is rational to delay investment in the face of uncertainty, whether the investor is a private monopolist, a competitive rm or a welfare-maximising social planner. If demand evolves stochastically, then investment should be held back until it is clear that the new capacity will really be needed, given the extent of sunk costs in the electricity industry. This implies that the price at which investment becomes attractive is higher than the price needed for the plant to just cover its costs, since there is a risk that the price will fall back in future. Similarly, it is not rational to close plants as soon as the price falls below the level of their variable costs. A prot-maximising rm and a welfare-maximising social planner should react to uncertainty in exactly the same way, implying that there is no market failure. As long as the market mechanism reveals the true value of capacity at times when there is a shortage, investors should receive the correct signals. How will investors respond to these signals? We do not yet have enough experience to draw denitive conclusions. In the markets shown in the previous section, there does seem to be an inverse correlation between investment levels and the margin of spare capacity, which is desirable. Capacity margins seem to be stabilising at lower levels in liberalised markets than at the start of the 1990s, which is also probably desirable holding too much spare capacity is expensive. The US probably had an excessive level of investment at the end of the 1990s, feeding through to a rapidly rising capacity margin, and nancial problems for many investors, in the early years of this century. The UK experienced some similar problems on a smaller scale. Overall, the experience is mixed. When judging the ability of markets to coordinate investment, however, we must remember that the alternative is not the outcome of a perfect social planner, but of imperfect regulation. On that basis, market-driven investment in electricity would seem to have an acceptable record, and prospects.
APPENDIX 2A
A MATHEMATICAL MODEL OF CAPACITY AND PRICING
Let us consider the problem of nding the welfare-maximising levels of capacity, and period-by-period outputs. The dual variables to this problem are the electricity prices in each period. Assuming that a competitive
48
electricity market has rules that will set these prices, we can then rely on the fundamental theorem of welfare economics to show that these are the quantities of output and capacity that the competitive market would produce. We shall work in discrete time, with T time periods, and I types of plant. Period 1 has the highest demand, and T the lowest, working from left to right of the load-duration curve in panel 2 of Figure 2A.1. Plant type 1 has the lowest operating costs, and type I the highest, with the others ranked in order, working up the stack of capacity in the same panel. The operating cost of plants of type i is ci per unit of output per period, and the xed cost (for the whole planning horizon) is di per unit. Without loss of generality, we only list capacity types that will be used in equilibrium, which has to imply that d1 d2 . . . dI. The output from plant type i in period t is qi, which must be less than its capacity of ki. The gross benet from consuming an output of Qi in period t is Bt(Qi). Our problem is thus:
Max W
qit, ki
Bt
t i
qit
i t
ci qit
i
di ki. (2A.1)
s.t.
qit
ki
i, t
Alternatively, we can write it as a Lagrangean, which gives us: Max L

qit, ki t
Bt
qit
ci qit
diki
it (qit
ki ). (2A.2)
If we then dierentiate with respect to our choice variables and the Lagrangean multipliers, we obtain: L qit Bt L kit L
it
qit
ci di ki.
it
(2A.3)
it
(2A.4) (2A.5)
qit
This gives us a rst-order condition for output: 0 qit Bt qit ci 0. (2A.6)
it
49
/MW-year
T* (3) GW GW D K B +D B (2)
Hours/year
B /MWh PR CP CB B (4)
K D GW /MWh
T* (5)
Hours/year
Max demand
K D GW
P T* s T* T*+u
Hours/year
Notes: (1) Total costs by plant type. (2) Load-duration curve. (3) Reecting line. (4) Marginal cost and demand. (5) Price-duration curve.
Figure 2A.1
The determination of electricity capacity and prices
50
With it always (weakly) positive, this states that the output from plant type i will be positive if the marginal benet of consumption equals or exceeds its variable cost. Furthermore, whenever the plant is producing, it will equal the dierence between the marginal benet of consumption and the plants variable cost. From the second set of derivatives, we have:
it
di.
(2A.7)
This tells us that the sum of those dierences (between the marginal benet of consumption and its variable cost, while the plant is operating) will equal its xed cost. In other words, the total cost of the plant is equal to the marginal benet of its output, summed over all the periods in which it is running. The third set of derivatives reminds us that the output from each plant type cannot exceed its capacity: 0
it
kit
qit
0.
(2A.8)
At times when the plant is not running at full capacity, it is zero relaxing the capacity constraint in those periods would not raise social welfare. These equations can be solved for the optimal capacities and pattern of output. In a market system, the market price in each period should be equal to the marginal benet of consumption, Bt ( i qit ) . We can get a better feel for the pattern of outputs and capacities if we introduce some notation for the number of periods of operation over which two (adjacent) plant types have equal total costs: ti di ci di
1
ci .
(2A.9)
We then know that we want plant of type i to be marginal at time ti, so that all the plant of type i 1 should stop running before this time. That allows us to calculate the total capacity requirement of types 1 to i, as equal to the level of output that will equalise the marginal benet of consumption at time ti with its marginal cost:
i
B ti
j 1
kj
ci
iti.
(2A.10)
Note that this formulation includes the Lagrangean multiplier for the marginal type of plant, since that plant type might be capacity constrained. In terms of panel 4 of Figure 2A.1, we might be on one of the vertical segments of the industry marginal cost curve. If we ignore that possibility, then
51
the requirement is that the marginal benet of running all the capacity up to and including type i should just equal the variable cost of plant type i. We determine the amount of peaking capacity (type I) in a slightly dierent way. We need to nd the time, tr, at which the marginal benet of consumption just equals the variable cost of peaking capacity, assuming that all the plant types are running at capacity:
I
B tr
j 1
kj
cI.
(2A.11)
We can then obtain an equation for the amount by which the marginal benet of consumption exceeds the variable cost of peaking capacity, in each period up to period tr. This is the net benet of the marginal unit of capacity in that period. The sum of these net benets should equal the xed cost of peaking capacity, dI:
tr t 0 I
Bt
j 1
kj
cI
dI.
(2A.12)
Solving equations (2A.11) and (2A.12) together gives us the solution for the total amount of capacity, and hence the amount of peaking capacity. One simplication in the main text is that the load-duration curve has been drawn as a smooth curve, despite the apparently large drop in price that occurs at T*. Such a fall in price would normally lead to an increase in demand, of course. However, the load-duration curve has to be monotonic, by construction. Figure 2A.1 has been drawn to show what would actually happen. We would expect some demand curves to pass through the vertical section of the marginal cost curve in panel 4, and the price at those times is between the variable cost of base-load plant and the variable cost of peaking plant. In our mathematical notation, it is between zero and (CP CB) in those hours. If we had enough peaking plant that some of it was still running until just before hour T*, then the base-load plant would already have recovered its xed costs by this time. With several periods just after T* in which the price exceeds its variable cost, the base-load plant would earn supernormal prots. Equation (2A.7) shows that this is not a feature of the optimal solution. Instead, we need to have less peaking plant, and more base-load plant (within the same total capacity), so that the price starts to fall below CP before hour T* (at, say, T* s), and is still above CB until some point after hour T* (say, T* u). This gives us the downward-sloping segment in the price-duration curve of panel 5. Because we are moving down a vertical
52
segment of the marginal cost curve in these hours, the level of output is constant, and we therefore have a horizontal segment in the load-duration curve, from T* s to T* u, at a height equal to the (new) level of baseload capacity, B D. In practice, however, most electricity industries contain so many plants that the marginal cost curve has few large discontinuities, giving a smooth load-duration curve.
NOTES
* Support from the Commission de Rgulation de lnergie is gratefully acknowledged. I would like to thank the Department of Applied Economics, University of Cambridge, for its hospitality. I have beneted from helpful comments by Franois Lvque, JeanMichel Glachant, Paul Joskow, Steven Stoft and participants at a seminar at the CRE. This model draws on the presentation in Stoft (2002), although he does not put the diagrams on a single page. Depending upon the market arrangements, these could be contracts for physical delivery, or nancial hedging contracts for dierences around a spot market price. The pre-construction stages obviously involve some sunk costs, but as long as the company does not sign binding contracts until these stages are completed, it is the length of the construction period that mainly determines the lag between committing to the irreversible investment and being able to earn revenues from a power station. If all plants were identical, it would not be optimal to build a new plant while an existing plant was in mothballs, but it is conceivable that investment in an ecient new plant could be protable while an inecient old plant was best kept out of the market (but not yet retired). The incumbent may be protected by switching costs, so that few customers will switch unless the incumbent is signicantly more expensive than its rivals, but even in this case, the incumbents optimal retail price varies with the current wholesale price. This section owes a lot to comments from Steven Stoft, although he should not be held responsible for the way in which I have interpreted them. The data on peak demands, obtained from Nordel, are not adjusted for weather, and a three-year moving average has been used to smooth out uctuations in Figures 2.46. It is far from clear that the power cuts in California in 2001 were caused by a physical shortage of capacity, as opposed to a shortage of capacity being oered to the market (Blumstein et al., 2002).
1. 2. 3.
4.
5. 6. 7. 8.
REFERENCES
Bar-Ilan, A. and W.C. Strange (1996), Investment lags, American Economic Review, 86 (3), 61022. Bar-Ilan, A., A. Sulem and A. Zanello (2002), Time-to-build and capacity choice, Journal of Economic Dynamics and Control, 26 (1), 6998. Blumstein, C., L. Friedman and R.J. Green (2002), The history of electricity restructuring in California, Journal of Industry Competition and Trade, 2 (12), 938. Brennan, M.J. and E.S. Schwartz (1985), Evaluating natural resource investments, Journal of Business, 58, 13557.
53
Bunn, D. and E. Larsen (1992), Sensitivity of reserve margin to factors inuencing investment behaviour in the electricity market of England and Wales, Energy Policy, 20 (5), 42029. Bunn, D. and E. Larsen (1994), Assessment of the uncertainty in future UK electricity investment using an industry simulation model, Utilities Policy, 4 (3), 22936. Dixit, A.K. and R.S. Pindyck (1994), Investment Under Uncertainty, Princeton, NJ: Princeton University Press. Ford, A. (1999), Cycles in competitive electricity markets: a simulation study of the western United States, Energy Policy, 27, 63758. Ford, A. (2002), Boom and bust in power plant construction: lessons from the California electricity crisis, Journal of Industry, Competition and Trade, 2 (12), 5974. Green, R.J. (2004), Retail competition and electricity contracts, CMI Working Paper EP33, University of CambridgeMIT Institute. Hawdon, D. (1978), Tanker freight rates in the short and long run, Applied Economics, 10 (3), 20317. Joskow, P.L. and J. Tirole (2006), Reliability and competitive electricity markets, Rand Journal of Economics, forthcoming. Newbery, D.M. (2002) Problems of liberalising the electricity industry, European Economic Review, 46, 91927. Oer (1998), Review of Electricity Trading Arrangements: Proposals, July 1998, Birmingham: Oce of Electricity Regulation. Pindyck, R. (1999), The long-run evolution of energy prices, Energy Journal, 20 (2), 127. Stoft, S.E. (2002), Power System Economics: Designing Markets for Electricity, Chichester: Wiley.
3.
Generation technology mix in competitive electricity markets

Jean-Michel Glachant
INTRODUCTION
1.
Competitive electricity reforms began to emerge at the beginning of the 1990s. We now have a certain distance, especially in Europe and the United States, for observing behaviour in terms of investments in generation within a competitive framework. The oldest of these reforms currently have 15 years of experience, as in England (in eect since 1990) and Norway (since 1991). The rst European Directive on creating a single electricity market (Eu,1996) was adopted a decade ago and enacted as of 1997 (like in Spain) and 1999 (like in Italy). In the United States, the beginning of the competitive reforms can be traced to the creation of independent system operators (ISOs) between 1995 and 1997 covering Texas, California and PJM (PennsylvaniaNew JerseyMaryland), but a true opening of competitive wholesale markets cannot be said to antedate 1998. Did these competitive reforms of the electricity industry have an impact on the choice of generating technologies? Do the new competitive pressures create an incentive for producers to select new methods for generating electricity (like the combinedcycle gas turbine: CCGT) and to abandon old technologies adopted under their previous status as utility monopolies, these former technologies favoured by policies and government subsidies and nanced by a guaranteed sales price? Would there be a new trend that the most capital-intensive technologies (for example, nuclear power) would be avoided? What link can we trace from actual competitive market reform to actual generation technology choice? Chapter 2 discussed the economic theory of capacity investment and its cycles, and the actual patterns in liberalised markets; this chapter is a more empirical study of actual generation technology choice. We shall address the potential impact of the competitive reforms on the choice of production technology in three stages.1 The rst stage (Section 2), examines whether the main competitive reforms in the United States and Europe fostered an evolution in the volume and technology of investments in
54
Generation technology mix
55
generation. If a technology change occurs with low volumes of investment we may doubt its signicance. Therefore we have to establish whether these reforms were indeed often accompanied by high levels of investment, and whether a technological change particularly favoured new gas-based CCGT technology. Having veried both the importance of the investment wave and of the technological shift, we have to ask, in a second stage (Section 3), why this technology change occurred in that wave of capacity investment. By looking at the traditional method of comparing electricity generation costs, levelised cost methodology, we shall nd that it provides a simple economic explanation for this technological shift by establishing the CCGT system as the least expensive among existing technological alternatives. However, in contrast to cost analysis performed in the UK and the US, the main French analysis of the cost of electricity generation always nds in favour of nuclear technology. Then, in Section 4, the third and nal stage is dedicated to understanding why cost analysis can diverge that much. In particular, we shall focus on the economic determinants that have been accounted for in the comparative analysis of technology costs under a competitive framework, taking as a benchmark the cost study made at the Massachuselts Institute of Technology (MIT) in 2003. As already suggested by Chapter 2, we shall see that the core of this debate on the actual costs of new plants is on evaluating the eects of competitive power markets and of vertical disintegration of generation on the cost of capital. Since nuclear power is extremely capital intensive, two to three times more than its alternative technologies, it is much more sensitive to the way the nancial market and the banking industry actually take into account the risks and uncertainty of generation investment. Section 5 concludes.
2.
ONE DECADE OF INVESTMENT IN GENERATION IN A COMPETITIVE FRAMEWORK
The multitude of US and European experiences in competitive electricity reforms is characterised by a vast range of timings and a broad diversity of modalities. These reforms do not all present the same investment prole or technological choice. However, two broad traits can be seen to prevail. On the one hand, when these reforms were accompanied by investments in generation, very large volumes of new capacity were created (up to 30 or 40 per cent). This simple fact means that it has been a wave of signicant investment. On the other hand, the technology principally used in this
56
important wave of competitive investments is relatively new: gas-based systems (up to 90 per cent of new capacity). This shows that the new wave of investment is characterised by a kind of technology shift. A Large Volume of Investment in Capacity The evolution of generating capacity in the United States between 1990 and 2002 shows a pronounced change in pace during the recent competitive period of 19982002. During the eight years preceding the pivotal year of 1998, total production capacity increased by only 44 GW, or about 5 per cent, while during the four years following 1998 this capacity rose by 155 GW about 20 per cent. This represents a genuine leap forward in generation capacity (Hunt, 2002). However, since a number of US states either did not embark on these competitive reforms, or did so belatedly, there is some room for doubt concerning the link between the evolution of total volume of capacity and the reforms. This is why we must look at the distribution of these capacities between the electricity utilities (being the traditional regulated companies from before the reforms) and the independent power producers (the IPPs). These independent generators, though they may have been created before the competitive wave of the late 1990s and owned by the utilities, none the less represent one of its key features, because such generators work outside the traditional regulatory framework that has been applied to utilities for decades.2 In the eight years following 1990, total capacity owned by the utilities declined by 7 GW (1 per cent) while that owned by the IPPs increased by 28 GW ( 300 per cent, either by construction or by acquisition from the utilities). However, during the four years after 1998, which was the true start of the competitive era, the capacity owned by the utilities fell substantially, by 132 GW (18 per cent), while that owned by the IPPs shot up 271 GW ( 700 per cent, either by construction or by acquisition as new subsidiaries of the utilities). The strong growth in investments during the early competitive period, 19982002, which appears to be without precedent in terms of volume, is thus particularly characterised by the actions of the IPPs. They not only accounted for the bulk of investments in new capacity, but also substantially dismantled the installed base of power plants managed by the utilities. When these four exceptional years ended in 2002, the IPPs found themselves owning a generating capacity of nearly 310 GW, or over half as much as the traditional utilities (597 GW) (Table 3.1). It is true that these IPPs can be subsidiaries of the utilities, particularly from other states. But this does not negate the special role they have played in this massive wave of recent investment in generation.
57
Table 3.1
Generation capacity in the USA, 19902002

Nameplate capacity (gigawatts) 781 735 9 825 44 728 7 37 28 980 155 597 131 308 271
Producer type 1990 Total industry Utilities IPP 1998 Total industry Change total industry 19901998 Utilities Change Utilities 19901998 IPP Change IPP 19901998 2002 Total industry Change total industry 19982002 Utilities Change Utilities 19982002 IPP Change IPP 19982002
Source: Own calculations data from IEA.
If we now concentrate on only two of the most extensive competitive reforms in the United States, California and Texas (representing a total of nearly 130 GW in capacity in 1998), we again nd evidence of intense investment in generation ( 20 GW overall) between the years 2000 and 2002. The increase in capacity attains 10 per cent in California, and 30 per cent in Texas where we observe a boom in capacity (Table 3.2). In Europe, two countries also witnessed considerable investments in generation under the competitive reforms (totalling approximately 45 GW). First, in England and Wales the equivalent of 40 per cent of the initial capacity was added during the rst decade of the reform, while Spain added nearly 30 per cent in ve years. Italy also created nearly 6 per cent in new capacity in the rst four years. Only Norways capacity had scarcely changed ( 3 per cent) 10 years after the competitive reform, (Table 3.3). A Profound Change in Technology: The Dash for Gas While in the rst years of the competitive era in the United States, the growth of capacity investment was unprecedented, the evolution of the fuel
58
Table 3.2
State
Generation capacity in California and Texas

Installed capacity (gigawatts) 54.3 54.1 54.1 57.2 59.5 5.2 78.2 79.9 86.8 94.0 101.7 23.5 Annual variation (%)
California Year 1998 Year 1999 Year 2000 Year 2001 Year 2002 Total change 19982002 Texas Year 1998 Year 1999 Year 2000 Year 2001 Year 2002 Total change 19982002
Sources:
0.6 0 5.7 4.0 9.6
2.2 8.6 8.3 8.2 30
Various years: EIA-DOE; PJM reports.
mix too presents a break during this post-1998 competitive period (Tables 3.4 and 3.5) Previously, between 1990 and 1998, coal and nuclear technologies dominated generation in the United States with 56 per cent (in 1990) and 54 per cent (in 1998), while gas (gas only or dual fuel-oil and gas3) accounted for about one-quarter (23 and then 27 per cent). During the rst competitive period, 19982002, gas-based systems jumped to 38 per cent of capacity (of which approximately 20 per cent was gas only), while coal and nuclear fell well below half at 45 per cent. In fact, in absolute value, coal and nuclear generation capacity remained unchanged over these four years of technological change, with 338 GW and 105 GW, respectively. It is rather the expansion of gas ( 113 GW) and dual fuel ( 38 GW) that opened the technological shift. New capacity in these two gas-based technologies represented 97 per cent of incremental output in the United States between 1998 and 2002. The comparative evolution of the capacity and fuel mix of the utilities and the IPPs is also very signicant (see Tables 3.4 and 3.5). In 1990, the IPPs generated 90 per cent of their output from hydro and other renewable energy sources. Between 1990 and 1998, increased capacity of the IPPs in the two gas-based technologies (gas only and dual fuel) represented 49 per cent of total growth in these technologies in the United States, while the corresponding number for the utilities was only 11 per cent4. Consequently, by 1998 the two gas technologies accounted for 52 per cent of the IPPs
59
Table 3.3 Generation capacity in Norway, England and Wales, Spain and Italy
Country Norway Year 1991 Year 2002 Change 19912002 Italy Year 1998 Year 2002 Change 19982002 England & Wales Year 1990 Year 2000 Change 19902000 Capacity additions 19902000 Capacity closures 19902000 Capacity mothballed in 2000 Spain Year 1998 Year 2003 Change 19982003 Installed capacity (gigawatts) 27.1 28.0 0.9 75.0 79.2 4.2 63.9 71.6 7.7 25.0 17.3 5.4 49.1 63.6 14.5 Variation (%)
3.3
5.6
12 39 7 8
29.5
Sources: Various years: Statistics Norway and NVE; British Electricity Association; GRTN; REE.
installations, while hydro and other renewables amounted to only 28 per cent. However, at 10 GW, the gas only capacity of the IPPs was only onequarter that of the utilities. Between 1998 and 2002, the increase in the IPPs capacity in the two gas technologies represented 99 per cent of the total growth in these technologies in the United States. While the utilities additional gas only capacity amounted to only 8 per cent of that total, their capacity in dual-fuel fell by 16 per cent (the utilities converted or sold to the IPPs some 24 GW of their dual-fuel capacity). After this short period of intense technological change, the IPPs gas only capacity was 99 GW in 2002 and nearly twice as large as the utilities gas only capacity. Overall, 55 per cent of the total generation capacity of the IPPs consists of gas or dual fuel, as opposed to 26 per cent in the case of the utilities. As a result, the rst years of the competitive era opened a sweeping and rapid technological switch to gas in the United States, and it is actually closely linked to the competitive reform. The technological evolution of two big
60
Table 3.4
Year
Generation fuel mix in the USA, 19982002

Energy source Energy mix (%) Generations capacity (megawatts) 337,800 45,300 82,100 142,100 0 104,800 95,500 16,400 900 824,900 320,600 42,800 42,400 124,000 0 104,800 91,100 2,200 200 728,100 6,600 800 Capacity change 19982002 (%)
Producer type
1998 1998 1998 1998 1998 1998 1998 1998 1998
Total electric power industry Total electric power industry Total electric power industry Total electric power industry Total electric power industry Total electric power industry Total electric power industry Total electric power industry Total electric power industry Electric generators, electric utilities Electric generators, electric utilities Electric generators, electric utilities Electric generators, electric utilities Electric generators, electric utilities Electric generators, electric utilities Electric generators, electric utilities Electric generators, electric utilities Electric generators, electric utilities Electric generators, IPPs Electric generators, IPPs
Coal Petroleum Natural gas Dual red Other gas Nuclear Hydroelectric Other renewables Other
41.0 5.5 10.0 17.2 0.0 12.7 11.6 2.0 0.1 100.0%
Total industry (year 1998) 1998 1998 1998 1998 1998 1998 1998 1998 1998 Coal Petroleum Natural gas Dual red Other gas Nuclear Hydroelectric Other renewables Other
44.0 5.9 5.8 17.0 0.0 14.4 12.5 0.3 0.0 100.0%
Total utilities (year 1998) 1998 1998 Coal Petroleum
17.7 2.1
61
Table 3.4
Year
(continued)
Energy source Energy mix (%) Generations capacity (megawatts) 10,300 9,000 0 0 3,200 7,300 0 37,200 338,200 43,200 195,000 180,200 2,200 104,900 96,300 18,800 800 978,800 260,600 25,800 54,600 99,200 0.1 4.6 137.4 26.8 ** 0.2 0.9 14.9 12.3 18.8 18.7 39.7 28.8 20.0 Capacity change 19982002 (%)
Producer type
1998 1998 1998 1998 1998 1998 1998
Electric generators, IPPs Electric generators, IPPs Electric generators, IPPs Electric generators, IPPs Electric generators, IPPs Electric generators, IPPs Electric generators, IPPs Total electric power industry Total electric power industry Total electric power industry Total electric power industry Total electric power industry Total electric power industry Total electric power industry Total electric power industry Total electric power industry Electric generators, electric utilities Electric generators, electric utilities Electric generators, electric utilities Electric generators, electric utilities
Natural gas Dual red Other gas Nuclear Hydroelectric Other renewables Other
27.6 24.4 0.0 0.0 8.5 19.7 0.0 100.0%
Total IPP (year 1998) 2002 2002 2002 2002 2002 2002 2002 2002 2002 Coal Petroleum Natural gas Dual red Other gas Nuclear Hydroelectric Other renewables Other
34.5 4.4 19.9 18.4 0.2 10.7 9.8 1.9 0.1 100.0%
Total industry (year 2002) 2002 2002 2002 2002 Coal Petroleum Natural gas Dual red
43.7 4.3 9.2 16.6
62
Table 3.4
Year
(continued)
Energy source Energy mix (%) Generations capacity (megawatts) 100 67,400 88,000 1,000 0 596,700 66,900 15,400 98,800 70,100 0 37,500 7,200 11,900 100 307,900 Capacity change 19982002 (%) ** 35.6 3.5 54.3 18.1 915.8 1,900.5 861.3 671.6 ** ** 130.0 62.4 * 728.0
Producer type
2002 2002 2002 2002 2002
Electric generators, electric utilities Electric generators, electric utilities Electric generators, electric utilities Electric generators, electric utilities Electric generators, electric utilities Electric generators, IPPs Electric generators, IPPs Electric generators, IPPs Electric generators, IPPs Electric generators, IPPs Electric generators, IPPs Electric generators, IPPs Electric generators, IPPs Electric generators, IPPs
Other gas Nuclear Hydroelectric Other renewables Other
0.0 11.3 14.7 0.2 0.0 100.0%
Total utilities (year 2002) 2002 2002 2002 2002 2002 2002 2002 2002 2002 Coal Petroleum Natural gas Dual red Other gas Nuclear Hydroelectric Other renewables Other
21.7 5.0 32.1 22.8 0.0 12.2 2.4 3.9 0.0 100.0
Total IPP (year 2002)
Note: (**): one cannot do a division by 0. A third producer category (CHP: combined heat & power) is not presented. Source: Own calculations data from IEA.
competitive reforms in the United States (California and Texas) thus corroborates the overall American data. In the additional capacity that came online, around 30 GW was powered by natural gas. In California and Texas virtually all post-1998 capacity addition was in gas-based technologies (Table 3.6).
63
Table 3.5
Generation capacity changes in the USA, 19902002

Mix of the Capacity change As % of the change (in %) (in megawatts) industry change 100.0% 16.8 20.4 49.0 42.1 7.4 100.0% 27.2 156.7 67.5 138.1 50.5 100.0% 22.7 2.2 35.5 30.2 0.0 5.4 43,700 7,300 8,900 21,400 18,400 3,200 6,400 1,700 10,000 4,300 8,800 3,200 28,600 6,500 600 10,100 8,600 0 1,500 100.0% 100.0 100.0 100.0 100.0 100.0 14.6 23.7 112.5 20.2 47.9 100.0 65.4 88.6 7.1 47.4 46.9 0.0 n.a
Producer type Capacity change total industry 19901998 (Industry capacity change in coal 19901998) (Industry capacity change in petroleum 19901998) (Industry capacity change in gas 19901998) (Industry capacity change in dual 19901998) (Industry capacity change in nuclear 19901998) Capacity change total utilities 19901998 (Utilities capacity change in coal 19901998) (Utilities capacity change in Petroleum 19901998) (Utilities capacity change in gas 19901998) (Utilities capacity change in dual 19901998) (Utilities capacity change in nuclear 19901998) Capacity change total IPP 19901998) (IPP capacity change in coal 19901998) (IPP capacity change in petroleum 19901998) (IPP capacity change in gas 19901998) (IPP capacity change in dual 19901998) (IPP capacity change in nuclear 19901998) (IPP capacity change in other renewable 19901998)
64
Table 3.5
(continued)
Mix of the Capacity change As % of the change (in %) (in megawatts) industry change 100.0% 0.3 1.3 72.9 24.6 0.1 100.0% 45.6 12.9 9.3 18.9 28.4 100.0% 22.3 5.4 32.7 22.5 13.9 1.7 154,700 400 2,100 112,800 38,000 200 131,500 60,000 17,000 12,200 24,800 37,300 270,800 60,300 14,600 88,600 61,000 37,500 4,600 100.0% 100.0 100.0 100.0 100.0 100.0 85.0 15,326.9 816.9 10.8 65.2 21,199.2 175.0% 15,398.5 703.5 78.5 160.5 21,299.2 n.a
Producer type Capacity change total industry 19982002 (Industry capacity change in coal 19982002) (Industry capacity change in petroleum 19982002) (Industry capacity change in gas 19982002) (Industry capacity change in dual 19982002) (Industry capacity change in nuclear 19982002) Capacity change total utilities 19982002 (Utilities capacity change in coal 19982002) (Utilities capacity change in petroleum 19982002) (Utilities capacity change in gas 19982002) (Utilities capacity change in dual 19982002) (Utilities capacity change in nuclear 19982002) Capacity change total IPP 19982002 (IPP capacity change in coal 19982002) (IPP capacity change in petroleum 19982002) (IPP capacity change in gas 19982002) (IPP capacity change in dual 19982002) (IPP capacity change in nuclear 19982002) (IPP capacity change in other renewable 19982002)
Note: A third producer category (CHP: combined heat & power) is not presented. Source: Own calculations data from IEA.
65
Table 3.6
State
Generation capacity changes in California and Texas

Capacity changes (gigawatts) 0 0 0 6.3 1.4 0 0.3 5.2 0 0.1 0.3 22.8 1.2 0 1.1 23.1 Changes fuel mix (%) 0 0 0 121 27 0 6 100 0 0.4 1.3 98.7 5.2 0 4.8 100
California Nuclear 19982002 Coal 19982002 Oil 19982002 Gas 19982002 Dual fuel 19982002 Hydro 19982002 Renewable & others 19982002 Total capacity change 19982002 Texas Nuclear 19982002 Coal 19982002 Oil 19982002 Gas 19982002 Dual fuel 19982002 Hydro 19982002 Renewable & others 19982002 Total capacity change 19982002
Sources: Various years: EIA-DOE.
In Europe, the technological evolution of the three competitive reforms that had invested in generating capacity, England and Wales, Spain and Italy, mirrors that of the United States. Overall, over 32 GW of gas-based generating capacity was built in these three countries. This exceeds their total expansion in capacity ( 27 GW). The dominant technology of new British and Italian development is gas, at between 75 and 95 per cent of the total capacity change. In Spain, a strong programme of support for renewable energies (especially wind power) has limited new gas capacity to 43 per cent of the total capacity change (Table 3.7).
3.
ECONOMIC DETERMINANTS OF THE NEW DOMINANCE OF GAS-BASED TECHNOLOGY
The fact that the fuel mix could change subsequent to a major upheaval like the competitive reforms was an open question before the reforms. However, given the technological changes seen after liberalisation and privatisation in the airline industry (with the expansion of the hub and spokes model)
66
Table 3.7 Generation capacity changes in England and Wales, Spain and Italy
Country England & Wales Nuclear 19902000 Coal 19902000 Oil 19902000 Others 19902000 Gas 19902002 Total capacity change 19902000 Spain Nuclear 19982003 Coal 19982003 Gas & dual fuel 19982003 Hydro 19982003 Renewable 19982003 Total change 19982003 Italy Gas & dual fuel 19982002 Hydro 19982002 Renewable 19982002 Total capacity change 19982002
Sources:
Capacity changes (gigawatts) 1.2 9.6 4.2 1.8 22.1 7.7 0.2 0.8 6.2 0.1 7.2 14.5 3.8 0.7 0.7 5.2
As % of the total capacity change 15 124 54 23 286 100 1.4 5.6 42.7 0.7 49.6 100 73.0 13.5 13.5 100
British Electricity Association (2001); GRTN; REE various years.
and in the telecom industry (with the expansion of digital technology and wireless network) (Bailey et al., 1985; Vickers and Yarrow, 1988); an energy technology change was not seen as impossible. Notably coal, as a heavily subsidised fuel in Europe, was seen as going to suer. None the less, predicting the sequence of changes that would follow the competitive reforms with any precision was rendered dicult because of the wide array of possible ineciencies under the former regime where cost and technologies were submitted to public regulation and public energy policy, and resulted in prices being imposed to customers through franchised monopolies (Joskow and Schmalensee, 1983; Beesley and Littlechild, 1994). Several ineciencies might have distorted the choice of generation technologies under the old regime of regulated production dierently (Joskow and Schmalensee, 1983, chs 7 and 12). On the purely theoretical front, Averch and Johnson demonstrated in 1962 that regulated rms could have a rational preference for the most
67
capital-intensive technologies because their Cost based price regulation was giving a guaranteed rate of return (ROR) to their invested capital. While the standard microeconomics was still asserting that the marginal pricing rule was able to rationally frame the management of existing monopoly (Mishan, 1968; Turvey, 1968; Turvey and Anderson, 1977), further economic literature showed that economic theory has not improved much in the four last decades of the twentieth century in explaining how the regulated framework inuences the electrical utilities in choosing their generating technologies (Pollitt, 1995; Ishii, 2004; Ishii and Yan, 2004). In practice, however, many countries had entrenched a fuel mix by promoting national combustibles (such as coal and lignite) through providing subsidies on their prices, or nancing a national variant within a given technology (such as nuclear) (Newbery and Green, 1996). In each of the US states, regulators have also used a variety of means to inuence the choice of their utilities. In particular, US regulators have the right to set regulated taris ex post (that is, after the facts) by giving or not giving their approval to the evolution of the cost of generation (for example, oil and gas prices could vary considerably from one period to another, without being totally passed through by the regulator) (Joskow and Schmalensee, 1986; Joskow, 1989, 2003). Quite apart from these general distortions, which can logically be laid at the feet of imperfect regulation, or the regulators preferences, or government energy policy, empirical studies have also discovered a whole universe of hidden ineciencies of all kinds. These have been concealed as much by dierences in construction costs as by the range of the operational performances of the main technologies (nuclear or fossil fuel), even within a single country, and even when account has been taken of the impact of the age and size of the plants and the precise characteristics of the fuel (Joskow and Rose, 1985; Joskow and Schmalensee, 1987; Joskow, 2002; Wolfram, 2003). Even within this imprecise framework, it is reasonable to expect that extending competitive mechanisms and eliminating government funding would impose a strong competitiveness constraint on producers, and thus promote the adoption of those technologies that truly are the most ecient in generation (Littlechild, 1994; Pollitt, 1995). The spread of gasbased technologies observed in Great Britain as of 199293 is thus attributable to the fact that this emerging technology proved to be the cheapest at the point in time when the competitive reforms were coming into their own (Newbery and Green, 1996; Newbery, 2000; Bower, 2004b). In the United States, the same wave, occurring later, marked both a technology shift and the end of regulatory uncertainty (Hunt, 2002; Ishii, 2004; Ishii and Yan, 2004).
68
The Comparative Competitiveness of Gas and Coal The economic benets of new gas-based generating technologies have been demonstrated many times and in various places, usually with the traditional method for comparing electricity generation costs known as levelised cost methodology. This methodology has been in use for decades and predates the competitive reforms. It was, and continues to be, used as much by investors as by institutions and economists. Based on an estimate of the various types of costs (capital costs, operating and maintenance costs, fuel costs) and annual energy generation over the lifespan of a typical plant for a given energy system, it discounts the ows of costs and energy prices using reference rates (traditionally, 5 and 10 per cent). The result is a comparable series of electricity generation cost levels that allow a ranking by economic merit to be established for choosing between alternative technologies (DGEMP, 1997 and 2003; IEA, 1999, 2002 and 2003a; EIA-DOE, 2000 to 2004; RAENE, 2004). The new competitiveness of gas relative to coal was clearly of great economic importance in countries such as the United States and Great Britain, where coal dominated the fuel mix (60 per cent coal in British electricity capacity in April 1990; 44 per cent coal capacity for US utilities in 1990 and 1998).5 This competitive edge of gas was simultaneously substantial and enduring, since the most recent study of Britains Royal Academy of Engineering in 2004 continued to describe CCGT plants as the leastexpensive method available for generating electricity (Figure 3.1). In the United States, studies by the Department of Energy (EIA-DOE) in 1996 predicted economic dominance for gas that would persist beyond the forecast horizon (2015) and featured a cost advantage of over 20 per cent in the average scenario (approximately $8 per MWh). For the rst time in a long while, these same forecasts, conducted in 2004, foresee that a greater competitiveness for coal could emerge between 2010 and 2025 (with a cost advantage of about 2 per cent, or $1.2 per MWh, $0.48 having been absorbed by higher network development costs explain this inclusion of network costs). Conrming this recent change, the February 2005 forecasts foresee one-third of the 200525 coming new plants in the United States being coal red with about 90 GW new capacity installed and 1,000 TWh of supplementary coal generation in 2025. However, new gas plants are still considered as providing about 60 per cent of the expected 280 GW generation investment for 200525 (Table 3.8 and 3.9). Of course, such studies are based on numerous assumptions, variants and scenarios that cannot be discussed in detail here. The economic dominance of gas over coal is thus relative, not absolute. There are indeed several regions in which coals competitiveness was never challenged by that of gas. In

8 7 Cost of generating electricity (p/kWh) 6 5 4 3 2 1 0 Standby generation cost Cost of generating electricity
69
Note: PF 5 pulverised fuel; CFB circulated uidised bed combustion; IGCC integrated gasication combined cycle; OGCT open cycle gas turbine; BFB bubblinguidised-bed combustion. Source: RAENG (2004).
Figure 3.1 Table 3.8
Item Capital O&M Fuel Total Heat rate
Source: EIA-DOE (1996).
ire PF d C oa l-f i C red F C B oa lIG fire C d G C as O -fire C d G G T as C -fire C d G T N uc le ar Po ul try -li t BF ter O B w ns in h d o fa re rm O w ffs in h d o fa re rm W av m e ar & in e
oa
l-f
Present-day cost of generating electricity in the UK 2003/04 1996 forecast costs of producing electricity, 2000 and 2015
2000 Conventional pulverised coal 26.41 10.72 13.58 50.72 9.840 Advanced combined cycle 2015 Conventional pulverised coal Advanced combined cycle 7.00 4.82 24.38 36.20 5.687
1994 mills per kilowatthour 11.24 26.18 4.82 10.72 22.35 7.42 39.41 44.32 Btu per kilowatthour 7.300 8.142
70
Table 3.9
Costs
2004 forecast costs of producing electricity, 2010 and 2025

2010 Advanced coal Advanced combined cycle Advanced coal 2025 Advanced combined cycle 12.33 1.36 37.91 2.78 54.38
Capital Fixed Variable Incremental transmission Total
33.77 4.58 11.69 3.38 53.43
2002 mills per kilowatthour 12.46 33.62 1.36 4.58 32.95 11.74 2.89 3.26 49.65 53.20
Source: EIA-DOE (2004).

50 45 Cost of new entry (Euro/MWh) 40 35 30 25 20 15 10 5 0 CCGT Belgium CCGT UK CCGT Spain CCGT Germany Fixed cash cost CCGT Italy Fuel cost Variable Capital costs
Source: Credit Suisse First Boston (2004).
Figure 3.2
CCGT cost of entry by country in Europe in 2005
European studies conducted by investment banks, such as those of the Crdit Suisse First Boston, cost calculations are individualised country by country to better account for local conditions (Figure 3.2). For the same CCGT technology and the same forecast entry into service in 2005, a dierence of $5 per MWh (or 13 per cent) is found to exist between Belgium and Italy.
71
The future competitiveness of these technologies will also depend on the carbon price, trading in which has only begun in Europe. In scenarios developed in Oxford by John Bower, the price of a tonne of CO2 is the key to the relative competitiveness of gas and coal, but also to that of existing power plants and investment in new British plants. With the price of carbon at 15 /CO2 ton, old nuclear energy is seen as having a marginal cost in 200812 smaller than 15/MWh, old CCGT plants around 25, and old coal plants around 31 permitting them to temper the interest of investing in new CCGT plants (Bower, 2004a). The Comparative Competitiveness of Gas and Nuclear Power Since the beginning of the US and British dash for gas, many studies using the levelised cost methodology have established that nuclear power has become the most expensive base-load thermal technology. The same conclusion is found in studies by investment banks, such as those from the Crdit Suisse First Boston in 2004 (approximately 50 per cent production cost gap to the detriment of new investment in nuclear power). Thus in the United States, during the past 10 years the DOEs Annual Energy Outlook has not foreseen any resumption of investment in nuclear power, regardless of the timeframe considered (19952025). Similarly, in its most recent World Energy Investment Outlook (IEA, 2003b) the International Energy Agency does not foresee any new investment in nuclear power in the United States or Europe, except in France and Finland, throughout the 30-year period, 200030. A MIT study in 2003 set the cost spread of nuclear, gas and coal at 25 to 50 per cent (MIT, 2003). In France the comparative analysis of the competitiveness of gas versus coal has made considerably less of an impression than the position of gas relative to nuclear energy. In a 1997 study by the Ministry of Industry (DGEMP, 1997), new CCGT technology successfully undercut the nuclear domination of the base load by assigning a price of 0.032 per kWh to the latter (for a series of 10 nuclear reactors) compared to 0.029 for CCGT under several scenarios (a fall in the price of gas, increased thermal eciency, lower construction costs). This reversal of technological prospects is quite abrupt relative to 1993 (only four years earlier), when the median cost advantage of French nuclear energy over coal (the only alternative envisaged at the time) was 30 per cent. However, to date all French studies have consistently demonstrated the economic superiority of nuclear technology over those based on coal or gas in the median scenarios. The most recent study of reference costs by the Ministry of Industry (DGEMP, 2003), describes a nuclear Pressurized Water Reactor (PWR) operating at 28.4 per MWh in 2015 (using 2001 prices, or 29.9 at 2004 prices), that is 20 per cent cheaper than
72
Table 3.10 Nuclear generation costs in the early twenty-rst century (per MWh)
Belgium Ampere 2000 30 Finland 2001 24 France DGEMP 2003 28.4 UK RAE 2004 33.8 USA MIT 2003 $ 67
Source: Areva (2004).
CCGT (34.5). Since French calculations for gas and coal tend to yield approximately the same results as studies from other countries, this dierence is entirely attributable to how the French ministry analyses the costs of nuclear power. Several other recent studies (Santaholma, 2003; AREVA, 2004; RAENG, 2004) (See Table 3.10), whether they concur with the French conclusions or not, have revealed that the discrepancies result from methodological dierences and reect divergences in the understanding of the impact of the competitive reforms on the comparative economic advantages of various generation technologies.
4.
THE COST OF CAPITAL AS A KEY ECONOMIC DETERMINANT OF TECHNOLOGY CHOICE IN A COMPETITIVE FRAMEWORK
It is common practice to rely on the traditional levelised cost methodology to evaluate the relative competitiveness of alternative technologies for generating electricity. This assumes, at least implicitly, that the costs and benets of technologies can be computed with no regard for the context in which they are to be implemented (regulated monopoly or competition). This working hypothesis, even when implicit, is surprising. Indeed, the economic analysis of regulated monopolies suggests that their generation costs could have been poorly understood or controlled, especially since the risks inherent in the choice of technology and capacity were not borne by the producers, but rather by consumers (Joskow, 2000). Conversely, in this new competitive framework these risks are borne by the producers themselves, a priori, and make up an essential element of the constraints guiding their behaviour (Joskow, 2003). One essential feature of competitive electricity regimes is that these new constraints are activated by the absence of guarantees on the demand addressed to each rm and the level or evolution of market and input prices. These unconstrained movements in volumes and prices, which could lead to extreme volatility, introduce a
73
new risk for producers. That volatility is more consequential for generators that are not vertically integrated and sell their output mainly or entirely in wholesale markets with neither established customer base nor contractual price guarantee (like Merchant Plants, which are IPPs that have not contracted for their output). Furthermore, in a competitive electricity generation regime, producers have access to neither government subsidies nor government capital (both of which are State Aid in the EU competition framework). Therefore, the cost of capital for investments in generation becomes one of the producers key decision variables (MIT, 2003), particularly given the high level of capital intensity of the electricity industry (IEA; 2003b). Costs and Risks Specic to Nuclear Power in a Competitive Regime The traditional framework for economic analysis of production costs, the so-called levelised cost methodology, encounters a major snag when comparing very dierent technologies in the new competitive framework. To rank these technologies by cost, this method assumes that it is easy, or at least not too dicult, to translate the risk proles of the various technologies into discounted cost levels. In the traditional methodological framework, a rst, simple way of expressing the technological component of the risks is to conduct sensitivity analysis. Thus, a sensitivity analysis of nuclear powers vulnerability to fuel cost variations, as in the 2004 Royal Academy of Engineering study, conrms that nuclear technology insulates the generation process from randomness aecting the fuel. The cost of generating gas-red plants is shown to be sensitive up to 30 per cent to their fuel costs, with extreme gas price scenarios making nuclear power cheaper than CCGT (RAENG, 2004). However, since the principal costs and risks of nuclear power lie outside of fuel costs, it remains to adapt the sensitivity methodology to this technology. The MIT study published in 2003 features a very interesting adaptation of this methodology to account for the new competitive environment in the United States that will frame any potential investment in American nuclear power. The reference cost for nuclear power in the MIT study is very high ($67 per MWh), compared with 28.4 in the DGEMPs French study published the same year. None the less, the MIT study indicates how this high reference cost could be reduced in the United States by a concerted series of voluntary actions (Table 3.11). Two primary levers for lowering costs are identied. First, the total cost of nuclear power could be reduced by $12 per MWh (18 per cent) if construction costs ($2,000 per kW in the reference scenario) could be brought down to $1,500 (bringing them below the 1,700 of the Royal
74
Table 3.11
Nuclear generation costs in the 2003 MIT study

With reduced construction cost (25%) With reduced construction time (5 years 4 years) 53 $ 2 $ ( 3%) With reduced O&M cost (down to 13 $ / MWh) 51 $ 2 $ ( 3%) With reduced cost of capital (down to gas or coal cost of capital) 42 $ 9 $ ( 13%)
Total reference cost ($ / MWh)
67 $
55 $ 12 $ ( 18%)
Source: MIT (2003).
Academy and near the 1,300 level of the DGEMP study for a series of 10 reactors at the exchange rate 1 $1.2). Second, total costs could also be reduced by $9 per MWh (13 per cent) if the cost of capital in nuclear power could be brought to the same level as capital costs for gas and coal technologies. The underlying economic reasoning is particularly relevant for our purpose. It underlines that private investors run much higher risks when choosing nuclear technology over gas or coal. Furthermore, investors cannot mitigate their own nuclear plant construction or management hazards by improving their knowledge or exploiting any series eect. Also, the competitive electricity market is much more unpredictable and volatile than the old regulated monopoly market. Merchant plants (independent generators) are particularly dependent on their wholesale market prices exposure. Thus, in terms of risk taking, nuclear power runs completely counter to the choice of sensible private investors. In the MIT study this is true for several reasons, all of which are adding their negative impact on the investors decision.
The real construction costs of a nuclear plant are poorly dened, since they have not been the subject of a controllable experiment for a very long time and were dicult to control in the past. The capital intensity of a nuclear plant is three to four times greater ($1,500$2,000/kW) than that of a CCGT plant, and the unit size of a plant is two to four times higher. Thus, the unit cost of the minimal investment could be 10 to 15 times greater. Moreover, if reduced construction costs are attained by investing in multiple reactors (often 10 units), the minimum size of a nuclear investment programme is 10 to 16 GW (depending on the reactor technology), somewhere between $15 and $30 billion. The timeframe for building a nuclear plant is inevitably long (at least ve years in a best-case scenario, versus two years for a CCGT plant),
75
but also uncertain. This makes the timing of entry into the market considerably more random, while the per-unit generating capacity is also much greater than for a gas-red plant. There is no possible correlation between the evolution of the market price of electricity and that of the principal components of the cost of nuclear power. This contrasts with the case of gas, for example, while this correlation, though imperfect, is eective (a substantial increase in the price of gas will have a positive impact on the wholesale price of electricity). Finally, the operational performances of nuclear power plants are also varied and stochastic. This is particularly true for the capacity factor, where a minimum factor of 85 per cent is required for nancial equilibrium of the nuclear project. In England the ve advanced gascooled reactors (AGRs) units operated from 1990 to 1995 at a mean load factor of 61 per cent, with the best unit at 78 per cent and the worst at 36 per cent (Mac Kerron, 1996). Another source of randomness aects operating and maintenance costs, with a $5 per MWh gap existing between current US mean performance, at $18 (fuel included), and its best quartile, at $13.
Each of these specic components of the risk associated with nuclear power are detrimental to the choice of this technology by an investor. The MIT study expresses this risk gap with a structure and cost of capital that diers from that of gas-based technology. For a CCGT plant, the American investor needs only to supply 40 per cent of funding from equity (compensated at a rate of 12 per cent) and can borrow the remaining 60 per cent (compensated at 8 per cent). For a nuclear plant, the investor needs to supply 50 per cent of funding from equity (compensated at a rate of 15 per cent) and borrow the remaining 50 per cent (at the same 8 per cent rate). This analysis of cost of capital specic to nuclear power plants is a major source of the cost dierences between MIT and other studies (the Belgian, British, Finnish and French shown in Table 3.10). However, a recent French study (IGF-CGM, 2004) has shown that the current nancial management of Electricit de France (EdF) now counts on a 13.7 per cent yield to equity in its nuclear generation. This is particularly striking because all existing French nuclear assets have been nanced in the former framework of a regulated monopoly backed by the nancial guarantee of the French state. This suggests that a new nuclear investment which would rely purely on market-based private nancing could not expect a lower return. Then a lower average cost of capital cannot come from a lower yield to equity and will crucially depend on the amount and rate of long-term debt or longterm bonds that the nuclear lender can nd in the market. None of the
76
Table 3.12
Belgium ampere 2000 All capital at 5% discount rate
Nuclear versus gas CCGT cost of capital analysis

Finland 2001 All capital at 5% discount rate France DGEMP 2003 All capital at 8% discount rate UK RAE 2004 All capital at 7.5% discount rate USA MIT 2003 CCGT at 9.6% (40% equity at 12 %) (60 % debt at 8%) Nuclear at 11.5% (50% equity at 15 %) (50% debt at 8%)
non-MIT studies distinguish between a nuclear and a non-nuclear investment, between a rate of return on equity and on debt (Table 3.12). In conclusion, the 2003 MIT study suggests that competitive reforms have a real impact on the choice of generating technology and that they particularly weigh in on the choice between nuclear and gas. They primarily impact on the cost factor most characteristic of nuclear power its capital intensity. Nuclear technology is extremely capital intensive, requiring $2,000/kW in construction costs in the United States (apart from the cost of capital and interest over the construction period) and yielding an annual energy volume of 7.4 MWh with median annual sales of approximately $320. The sensitivity of nuclear power to investors nancial behaviour is thus not comparable to that of other technologies. If, in addition to this core characteristic, nuclear technology also features a large number of uncertainties in terms of costs and yields, the US capital markets cannot deal with it on the same footing as the other technologies. However, since the functioning of capital markets is neither unied internationally nor stable over time, the MIT study does not allow any precise conclusions to be drawn concerning the nancing of future nuclear investment in France. The 8 per cent borrowing rate and 15 per cent yield to stockholders used by the MIT study may appear too high given current nancial conditions in France. A recent French report already mentioned (IGF-CGM, 2004) places the cost of debt at under 5 per cent (with a good credit rating), while the new standard for yield to equity adopted by EdF is currently 13.7 per cent. However, there is no more guarantee that EdF will be able to easily nd long-term loans (30 years and more) or the required equity. The former period of unlimited nancing for nuclear power at the bond-market rate with the guarantee of the French government is a thing of the past. A 50/50 nancing split between equity and debt will constitute a challenge for the French nuclear champion, which would need to mobilise about 40 billion in equity to recreate its current nuclear base,
77
after 2017 the foreseen date of the rst plant closure (Glachant and Finon, 2005) at the current French ministrys construction costs. Another recent ocial report to the French minister of the economy, nances and the industry suggested that the actual amount of equity at EdF is probably nil, or even negative, for the scal years 2004 and 2005 (Roulet, 2004). Comparison of Costs and Risks Specic to Nuclear Power and Gas in a Competitive Regime If nuclear generation technology appears so sensitive to the competitive reform, then gas technology is as well, but in a dierent fashion. The bursting of the American electricity bubble in 200203 revealed that the massive ow of IPP investment in gas-based generation capacity did not represent a sustainable nancing or investment regime (Joskow, 2003). Clearly, both the IPPs and their bankers had underestimated the risks of independent generation remunerated exclusively on the wholesale markets terms and unable to rely on either a portfolio of end clients or a long-term sales contract. A retroactive computation of the so-called spark spread (that is, the gap between the wholesale price of electricity and the fuel cost of the gas necessary to generate it) reveals the evolution of producers net revenues and how the IPPs experienced the oundering of the wholesale electricity market under the impact of the new overcapacity. In Texas, CCGT capacity increased by 23 GW between 1999 and 2002 while the on-peak spark spread decreased from $23 to only $6 a MWh (Figure 3.3) to face about $11 capital costs and about $5 operating and maintenance (O&M) costs! More generally, during the quarters following the Enron bankruptcy in the autumn of 2001, the business model consisting of a merchant plant to generate electricity and trading platforms to sell it on the wholesale market, which typied the rst period of the US competitive reform, collapsed in the United States. The stock market values of the main pioneers of the reforms (AES, Williams, Calpine, El Paso, Mirant, Reliant, Dynegy and so on) fell 90 to 95 per cent between the spring of 2001 and the spring of 2003, resulting in a $130 billion loss in stock value (excluding Enron), while their credit ratings fell to between BB and B. Securities representing their numerous bond issues were negotiated at below face value, by 15 to 75 per cent. New ows of bank and bond nancing into the US electricity sector fell by 70 per cent (that is, US$30 billion) between 2001 and 2002 (de Luze, 2003). Consequently, many of the IPPs assets have ended up in the hands of their bankers, while any future nancing of IPPs that do not have sales contracts for their output appears to be out of the question. Thus, the business model of the American electricity sector is reverting to one of vertical integration between generation and sales, either in the
78
US$/MWh $25.00 $23.11
ERCOT On-Peak Spark Spreads
$22.33
$20.00
$15.00 $11.93 $10.00 $6.15 $5.00
$0.00 1999 Source: de Luze (2003). 2000 2001 2002
Figure 3.3
Spark spread in Texas, 19992002
traditional form of a utility, or in the surrogate form of an IPP covered by a long-term sales contract (Joskow, 2003; CERA, 2004). Newbery (2000) showed how this vertically integrated competitive model diers from a pure vertical unbundling framework when it comes to investing in generation. The entry of new generators in England and Wales has been favoured in the dash for gas by the hedging eect of the long-term contracts (15 years) these generators signed with the regional distribution monopolies (the regional electric utilities (RECs)). None the less, despite the severity of the nancial correction that struck the IPPs in the United States, their stockholders and their bankers, no shift in technology choice has been induced. Gas-based generation has remained the norm. It is not the technology itself, but rather the business model that has shifted in the United States. The US competitive reforms are evolving towards dierent forms of vertical integration between generation and the nal sale, with no technological correction for the excesses of the gas bubble until the rise of gas prices prompts such a change. The main correction needed after the US gas bubble was a capacity adjustment to dry the existing overcapacity (see Chapter 2, this volume). The constancy of technological choices after a bubble and a shock of this magnitude may surprise. Under precisely these conditions, nuclear
79
technology remains a candidate for a radical rethinking of cost comparison methodologies that would allow the option value it represents to be captured. The fact that nuclear costs are not correlated with the standard market costs has to be properly valued. Several interesting academic studies have been working along these lines. Newbery et al. (2004) seek to substitute Monte Carlo simulations for the levelised cost methodology to compare the complete risk proles of alternative technologies (coal, gas and nuclear), and then generate a measure of the option value of nuclear power as a hedge against higher gas and coal prices. Gollier et al. (2004) propose a measure of the option value of a new, and much more modular, nuclear technology in which the investor could decide to successively build 1, 2, 3 or 4, 300 MW modules and where the investment cost diminishes with each additional module. But none of these studies calls for any link to an existing or coming investment. Thus, the current Finish example of a group of electricity distributors and large industrial clients co-nancing a nuclear reactor (to be built by the French rm Areva) remains an isolated empirical instance to account for. According to Santaholma (2003), President of the Finnish Energy Industries Federation, this investment could represent 10 per cent of the capacity used during peak hours in Finland. While this project has the status of a national project (publicly supported by the government, parliament and labour unions), it is nanced by TVO, a cooperative of local utilities and large industrial consumers. It will resell at cost price a volume prorated to reect each participants share in the investment. Given the three assumptions that total equipment costs will be 1,750 per kW (including all cost of capital xed at 5 per cent), that reactor operation is expected a capacity factor of 91 per cent (or 8,000 h annually), and that the included O&M fuel cost will be 10.2 per MWh, the mean cost per MWh is announced at 24, versus 32 for a CCGT plant all values for year 2000 (Figure 3.4). This Finnish example illustrates that, within the framework of a very long-term contract (40 years) nanced by the future energy buyers and transferring all risks to these buyers along with a very low cost of capital, it is possible to end up with very low values for capital cost and other costs (construction and O&M cost uncertainty being transferred to the energy buyers). In some ways, the cost analysis associated with the nuclear investment in Finland expects to reproduce the conditions of the French nuclear programme of the 1970s to 1980s within the framework of the existing European competitive reform. Today, in France, other paths, more or less similar, are being explored to bind industrial clients to EdF with long-term contracts (IGF-CGM, 2004) or to establish durable links between other electricity concerns in Europe
80
60 50 40 30 20 10 0
Elspot price 2000 Finland Elspot price 2001 Finland
Fuel costs O&M costs Capital costs Elspot prices

18.4 3 7.2 22.8 14.9 13.9 7.6
Nuclear Coal
10
Euro/MWh
17.1 23.7 7.4 1.5 5.3

Gas
15.8 8.2 6.5 10.2

Peat
40.1
13
Wood Wind operation hours = 2200 h/s
Real interest rate = 5.0 % November 2001 Prices
Operation hours = 8000 hours/year R. Tarjanne & K. Luostarinen 12.2.2002 Generation costs without investment grant and electricity tax rebate
Source: Santaholma (2003).
Figure 3.4
Finnish comparison of generation costs
and Frances nuclear capacity (Glachant and Finon, 2005). It is hoped that the dire predictions emanating from the United States regarding the structural handicap impeding nuclear technology in a competitive investment environment can be countered by the vertical integration of generation with nal sales or using long-term contracts, the construction of a large number of reactors and plants by very big concerns, the accumulated expertise of a world nuclear leader, very favourable credit ratings for this type of borrower and long-term debt (up to 30 years), as well as consideration of the option value of nuclear power against a possible rise in the price of fossil fuels and of carbon emission permits. The actual results of a new French nuclear case will be tested as soon as prototypes of the new nuclear reactor EPR are built in France (in the coming years), but large-scale closure of existing nuclear plants is not foreseen before 2020 and could be delayed for more than a decade.
5.
CONCLUSION
We have observed that the investment phases of the current competitive reforms in the electricity sector have been accompanied by a strong preference for gas-based generation technology, in particular for combined-cycle (CCGT) plants: 150 GW of capacity have been built in the United States and 32 GW in England, Spain and Italy.
81
Many studies attribute the new dominance of this gas technology to its lower cost compared to coal and nuclear technologies. These studies assume that the electricity reforms only accelerated the recognition of this greater eciency of gas and forced the hands of generators by introducing competition and curbing government subsidies or other state aids. This analysis of the reversal of the ranking of total coal and gas technology costs was broadly accepted during the 1990s and could, in fact, be overturned again in the United States due to the current gas prices. However, the ranking by economic merit between CCGT and nuclear technologies remains contested in France by the ministry of industry and by Britains Royal Academy of Engineering. The MIT study has allowed two causes of the divergence in the cost analysis of nuclear to be clearly identied. First, the baseline costs of nuclear technologies (construction and dismantling costs, O&M costs) and their operational performance (in particular the availability of plants and their lifespan) remain imprecise and highly variable. Second, characteristics that are intrinsic to investment in nuclear (especially: xed R&D costs, capital intensity per kW, reactor and plant minimal size, construction time, lack of correlation between input costs and the price of electricity) increase the risks assumed by the investor in a competitive framework. While nuclear power has a potential value as a hedge against the current rise of fossil fuel prices and of carbon emission costs, the setbacks experienced by the UK nuclear generator (British Energy, which had to be saved by direct state aid) demonstrate that the commercial survival of nuclear power in a competitive environment is not assured even when plants are already built. It is therefore understandable that in a competitive environment the cost of capital for investments in nuclear power could be driven up relative to that for gas or coal, especially when the investment is not to be made in an old and large vertically integrated utility which has a long record of excellent nuclear performance. All this undercuts the competitiveness of nuclear technology which is at least three times more capital intensive than gas. In the United States, the bursting of the CCGT investment bubble and the nancial crisis confronting independent power producers, as well as the bankruptcy of Enron and the collapse of the merchant plant trading model, have not succeeded in rehabilitating nuclear technology on a pure market base. They have, rather, exacerbated the typical nuclear problems by raising the cost of capital to all electricity generation projects. In Europe, however, or at least in Finland and France, nuclear professionals claim to be able to fully manage the risks and costs specic to nuclear power. All they need now is for capital markets to share that analysis. The privatisation of the French nuclear champion EdF could provide an occasion for these markets to give their rst feedback.
82
NOTES
1. I shall deal separately with the particular cases of peak-load technologies and renewable energy. Electricity generation is characterised by the coexistence of several technologies responding to dierent parts of the load; notably a specic technology can be used to generate at peak times (like a gas turbine having no combined cycle to recycle the heat produced by the ignition of the fuel). Renewable energy is another particularity in technology choice, since the growth of renewable technologies adoption still depends on subsidisation by the public energy policy and is not driven by independent decisions taken by the electrical companies. 2. Note that in the United States expressions like independent generators or independent power producers could refer to really new entrants as well as to new subsidiaries created by the old utilities under the umbrella of new laws. 3. Until 1995 gas plants in the United States favoured the old gas technology existing before the CCGT. 4. The balance was accounted for by a third type of actor: combined heat and power producers CHPs. 5. In turn, the French Ministry of Industry has recognised the competitiveness of the new CCGT in the 1997 update of its Reference Costs for Power Generation representing a break with the position taken in 1993. CCGT replaced coal as the traditional thermal energy benchmark. It dominated the mid-base and, in some scenarios, even the base-load generation.
REFERENCES
Areva (2004), Competitiveness and Sustainability: the Respective Trumps of Nuclear and Gas Power Generation, Paris. Averch, H. and L. Johnson (1962), Behavior of the rm under regulatory constraints, American Economic Review, 52, 105269. Bailey, E., D. Graham and Kaplan D. (1985), Deregulating the Airlines, Cambridge, MA: MIT Press. Beesley, M. and S. Littlechild (1994), Privatization: principles, problems and priorities, in M. Bishop, J. Kay and C. Mayer (eds), Privatization and Economic Performance, Oxford: Oxford University Press, pp. 1531. Bower, J. (2004a), Energy Trading Solutions in a Carbon Free World, Oxford Energy Institute, June. Bower, J. (2004b), Price impact of horizontal mergers in the British generation market, in D.W. Bunn (ed.), Modelling Prices in Competitive Electricity Markets, Chichester: Wiley, pp. 99120. British Electricity Association (2001), Electricity Industry Review, London, January. CERA (2004), North American Electric Power Watch, Cambridge, MA: CERA, Winter. Crdit Suisse First Boston (CSFB) (2004), Electricity Handbook, Boston: CSFB, March. de Luze, G. (2003), Investment in power generation: a bankers perspective, IEA/NEA (International Energy Agency and Nuclear Energy Association) Workshop on Power Generation Investment in Liberalised Electricity Markets, Paris, March. Direction Gnrale de 1Energie et des Matires Premires (DGEMP) (1997), Cots de rfrence de la production dlectricit, Ministre de lIndustrie, Paris.
83
Direction Gnrale de 1Energie et des Matires Premires (DGEMP) (2003), Cots de rfrence de la production lectrique, Ministre de lconomie, des Finances et de lIndustrie, Paris. EIA-DEO (Energy Information Agency and Department of Energy) (19962004), Annual Energy Outlook, Paris. (EU) European Union (1996), Common Rules for the Internal Market in Electricity, Directive 96/92/EC of the European Parliament and the Council of 19 December 1996. Glachant, J.-M. and D. Finon (2005), A competitive fringe in the shadow of a state owned incumbent: the case of France, Energy Journal, Special Issue European Electricity Liberalisation, 26, 181204. Gollier, C., D. Proult, F. Thais and G. Walgenwitz (2004), Comment valoriser la modularit dans les projets dinvestissement lectronuclaire?, IDEI & CEA (Institut dconomie Industrielle, Toulouse, and Centre dEnergi Atomique) February. GRTN (Gestore Rete Transmissione Nazionale) (200004), Statistical Data and Annual Reports, Rome, Italy. Hunt, S. (2002), Making Competition Work in Electricity, Chichester: Wiley. IEA (International Energy Agency) (1999), Electric Power Technology. Opportunities and Challenges of Competition, Paris. IEA (International Energy Agency) (2002), Security of Supply in Electricity Markets. Evidence and Policy Issues, Paris. IEA (International Energy Agency) (2003a), Power Generation Investment in Electricity Markets, Paris. IEA (International Energy Agency) (2003b), World Energy Investment Outlook 2003, Paris. IGF-CGM (Inspection Gnrale des Finances & Conseil Gnral des Mines) (2004), tude sur la formation des prix de llectricit dans un march concurrentiel, Rapport au Ministre de lconomie, des Finances et de lIndustrie, Paris, September. Ishii, J. (2004), Technology adoption and regulatory regimes: gas turbine electricity generators from 1980 to 2001, CSEM Working Paper, 128, Center for the Study of Energy Markets, University of California Energy Institute, Berkeley, CA. Ishii, J. and J. Yan (2004), Investment under regulatory uncertainty: U.S. electricity generation investment since 1996, CSEM Working Paper, 127, Center for the Study of Energy Markets, University of California Energy Institute, Berkley, CA. Joskow, P. (1989), Regulatory failure, regulatory reform and structural change in the electric power industry, Brookings Papers on Economic Activity: Microeconomics, 125298. Joskow, P. (ed.) (2000), Economic Regulation, Cheltenham, UK and Northampton, MA, USA: Edward Elgar. Joskow, P. (2002), Electricity sector restructuring and competition: a transaction-cost perspective, in E. Brousseau and J.-M. Glachant, The Economics of Contracts. Theory and Applications, Cambridge, MA: Cambridge University Press, pp. 50330. Joskow, P. (2003), The dicult transition to competitive electricity markets in the US, Cambridge Working Papers in Economics and CMI Working Paper, 0328, University of CambridgeMIT Institute. Joskow, P. and N. Rose (1985), The eects of technological change, experience and environmental regulation on the construction costs of coal-burning, Rand Journal of Economics, 16 (1), 127. Joskow, P. and R. Schmalensee (1983), Markets for Power. An Analysis of Electric Utility Deregulation, Cambridge, MA: MIT Press.
84
Joskow, P. and R. Schmalensee (1986), Incentive regulation for electric utilities, Yale Journal on Regulation, 4 (1), Fall, 149. Joskow, P. and R. Schmalensee (1987), The performance of coal-burning electric generating units in the United States: 19601980, Journal of Applied Econometrics, 2 (2), 85109. Littlechild, S.C. (1994), Competition, monopoly, and regulation in the electricity industry, in M.A. Einhorn (ed.), From Regulation to Competition: New Frontiers in Electricity Markets, Boston, MA: Kluwer, pp. 12550. Mac Kerron, G. (1996), Nuclear power under review, in J. Surrey (ed.), The British Electricity Experiment. Privatization: The Record, the Issues, the Lessons, London: Earthscan, pp. 13863. Mishan, E. (1968), A survey of welfare economics, in American Economic Association and Royal Economic Society (eds), Surveys of Economic Theory, Vol. I, London: Macmillan, pp. 154222. MIT, (Massachusetts Institute of Technology) (2003), The future of nuclear power. An interdisciplinary MIT study, MIT, Cambridge, MA. Newbery, D.M. (2000), Privatization, Restructuring and Regulation of Network Utilities, Cambridge, MA: MIT Press. Newbery, D.M. and R.J. Green (1996), Regulation, public ownership and privatization of the English electricity industry, in R.J. Gilbert and E.P. Kahn (eds), International Comparisons of Electricity Regulation, Cambridge, UK: Cambridge University Press, pp. 2581. Newbery, D.M., F. Roques and S. Connors (2004), Nuclear as a hedge against gas and carbon prices uncertainty. Preliminary results, CMI Research Seminar, Cambridge, UK, 6 November. NVE, (Norwegian Energy Regulatory Authority) (19992004), Statistical Data and Annual Reports, Oslo. PJM (PennsylvaniaNew JerseyMaryland) (19992004), Statistical Data and Annual Reports. Pollitt, M. (1995), Ownership and Performance in Electric Utilities: The International Evidence on Privatization and Eciency, Oxford: Oxford University Press. RAENG (Royal Academy of Engineering) (2004), The costs of generating electricity, London, March. REE (Rede Electrica de Espana) (19992004), Statistical Data and Annual Reports, Madrid. Roulet, M. (2004), Rapport au Ministre de lconomie, des Finances et de lIndustrie sur le projet industriel et nancier dEDF, Paris, November. Santaholma, J. (2003), Nuclear power investment, case Finland, IEA/NEA (International Energy Agency and Nuclear Energy Association) Workshop on Power Generation Investment in Liberalised Electricity Markets, Paris, March. Statistics Norway (19962004), Statistical Data, Oslo. Turvey, R. (1968), Public Enterprise. Selected Readings, Harmondsworth: Penguin Books. Turvey, R. and D. Anderson (1977), Electricity Economics, Baltimore, MD: Johns Hopkins University Press. Vickers, J. and G. Yarrow (1988), Privatization: An Economic Analysis, Cambridge, MA: MIT Press. Wolfram, C. (2003), How might restructuring aect the eciency of electricity generation in the US?, CSEM Working Paper, 111, Center for the Study of Energy Markets, University of California Energy Institute, Berkeley, CA.
PART II
Investment in transmission
4. Problems of transmission investment in a deregulated power market

Steven Stoft
1. INTRODUCTION
From the earliest days of commercial power production, transmission has grown steadily in importance. New wholesale power markets have sparked interest in distributed generation, but trade between these markets has only increased the need for transmission investment. As with generation, a market for the use of existing assets is not dicult to imagine, but a market to supply these assets is more problematic. In fact, the discrepancies between the properties of transmission costs and benets and the assumptions of competitive economic theory are so substantial that a market solution is probably not desirable. Even incentive regulation may prove so dicult to design and so inaccurate that a planning solution may be preferable, at least until wholesale power markets are functioning eciently and the generationinvestment problem has been solved. Transmission, an exceptionally heterogeneous product, can be both a substitute for and complement to generation, and suers from returns to scale and lumpiness,1 as well as major externalities, both positive and negative. This chapter investigates the extent to which these problems can be overcome. To simplify, this chapter ignores transmission losses because they play a relatively small role in the investment problem and one similar to the role of congestion, which is considered. After introducing some properties of congestion prices and transmission costs, three basic approaches to transmission investment are explored.* Three Approaches to Investment Three approaches to investment stand out as relatively distinct, although many mixtures of these are possible. A planning approach refers to a system that does not include any incentives specically tailored to the long-run transmission investment problem. Such an approach would be carried out
* Denitions and notations are given in the glossary at the end of the chapter.
87
88
by a group of engineers and economists under instructions to build an ecient system. In practice, it would be backed by rate-of-return regulation with a requirement that investments be used and useful. A merchant approach would allow any private company to modify the transmission system, subject to certain restrictions, and would reward (or punish) such modications by allocating transmission rights to investors. A performancebased regulation (PBR) approach would induce investment by a for-prot owner of the transmission system (a transco) by adjusting its prot level on the basis of the cost and performance of the system. Congestion: The Opportunity Cost of Using the Grid All three approaches will be assumed to exist within the framework of a wholesale power market based on publicly known nodal prices (Hogan, 1992; Harvey et al., 1996). That is, at each relevant point (node) in the network, a price is established, and these prices together clear the market. They may be purely competitive prices or they may be distorted by market power, but in any case there is one price at each node and all energy transactions at a given node take place at that price. These prices are adjusted each time there is a change in supply or demand. Such a pricing system automatically prices the use of the transmission system even though it applies directly only to energy transactions. If the price at node A is $20/MWh and at node B is $30/MWh, then the price to transmit energy from A to B is $10/MWh, while the price to transmit it from B to A is negative $10/MWh. Although nodal prices have some peculiar properties, it is important to understand that they are simply the result of the normal forces of supply and demand constrained by the physical limits of the transmission system. When these limits restrict ow, the system is said to be congested. When supply and demand are both competitive, nodal prices are simply the standard competitive market prices and have all the properties expected of such prices. Except when prices are determined somewhat arbitrarily because a vertical supply coincides with a vertical demand curve, competitive nodal prices are unique. Although they are often calculated from bids in a centralized auction, they are not the product of any special rules of calculation but are the prices at which a well-arbitraged bilateral market would arrive if the transmission constraints were enforced. Three distinct costs associated with congestion are often confused, congestion rent (CR), congestion cost (CE), and the cost of congestion to load (CL). Economists focus on the rst two, while consumers react to the third. Consider a load pocket with 1,000 MW of load and a 300 MW line into the load pocket from a large system that could supply 800 MW at $20/MWh and much more at $40/MWh. These costs are represented as a remote supply
Problems of transmission investment in a deregulated power market
89
Transmission constraint $50 Remote supply $30 $20 Congestion rent (CR) Congestion cost (CE) Net local demand 200 400 600 800 1,000 MW
Figure 4.1
Dening congestion rent and congestion cost
function for the load pocket in question. Suppose that local load is xed at 1,000 MW and that the pocket contains 600 MW of $30/MW generation and 200 MW of $50/MWh generation. The net local demand curve shows the demand for imported power net of what would be purchased locally (Figure 4.1). For example, at a local price of $40/MWh, the load pocket would consume 1,000 MW and supply 600 MW, leaving it with a net demand of 400 MW. In other words, net local demand accounts for local supply as well as local demand, and it is local supply that provides the price sensitivity of this demand curve and not actual demand responsiveness. The congestion cost is also called the, redispatch cost because it is the extra cost of dispatching more expensive generators than would be needed if the transmission system had ample capacity and did not constrain power transfers. In the present example, 100 MW of local $50 generation and 400 MW of local $30 generation must be used in place of 500 MW of remote $20 generation that would have been used had there been no transmission constraint. Congestion cost is a deadweight loss, not a transfer payment. Congestion rent is the amount collected by the owners of the rights to the transmission line. In a one-line network these rights would typically pay the owners an amount equal to the lines capacity times the dierence between the prices at the two ends of the line. In the case of a load pocket, this is the dierence between the internal price and the external price. Congestion rent is a transfer payment from line user to line owner, as using the line has no actual cost.
90
$/MW
Constraint Demand Rent Cost $/MW Constraint
Cost to consumers
Supply
Supply
MWh
MWh
Figure 4.2
Cost to consumers compared with congestion cost and rent
Table 4.1
Three views of congestion

CE CR CL $7,000/h $9,000/h $20,000/h
Congestion (redispatch) cost Congestion rent Cost of congestion to load
Finally, there is the cost of congestion to load (CL) (see Figure 4.2). With ample transmission, the load in the pocket would import 800 MW of power and use 200 MW from internal generators. The price would be set by the intersection of supply and demand at $30/MWh, so the total cost of power to load would be $30/MWh 1,000 MW, or $30,000/h. Because of the congestion, the price in the pocket is $50/MW and so load must pay $50/MWh 1,000 MW, or $50,000/h, which makes the cost of congestion to load $20,000/h. The three costs are shown in Table 4.1. As can be seen there is no particular relationship between these three concepts. Frequently consumers nd it unfair that congestion can cost them far more than the congestion cost. It does not help that it can also cost them more than congestion cost and rent combined with the excess revenue accruing to generators that seem to benet from the constraint without reason. Although the matter is beyond the scope of this chapter, it should be noted that if the transmission and generation markets satisfy the axioms of perfect competition, nodal prices will just cover the long-run costs of the ecient mix of generation and transmission. In other words there is nothing inherently unfair or inecient about the distribution of revenues under nodal pricing in a congested system. Of course this does not indicate that either transmission or generation is, or can be supplied by, a competitive market, only that any problems with these costs and prices arise from non-competitive features of the markets and not simply from the method of nodal pricing or the eects of transmission congestion on prices.
91
The Zero-congestion Fallacy Because transmission congestion imposes costs, one recurrent view holds that it should simply be eliminated. This is now the policy of the Alberta government (see Box 4.1). Although examples can be manufactured for which this is the least-cost solution, in real power markets such situations never exist. If there is one hour per year in which a remote generator is $10/MWh cheaper than the most expensive local generator in use, and if 1 MW of that generators output cannot reach local load, then the line is constrained during that hour and the cost of the constraint is $10/year. Adding 1 MW of capacity to that transmission path would cost far more than $10/year. Eliminating all congestion allowing every last megawatt of trade is simply not ecient. When transportation is expensive, it is often cheaper to consume the local product than to transport a slightly cheaper product from a distance.
BOX 4.1
ALBERTA REGULATION #174/2004, ELECTRIC UTILITIES ACT
Albertas Electric Utilities Act took effect on January 1, 1996. From the beginning, the Electric Utilities Act has imposed uniform prices (forbidden competitive prices) throughout the province. With its new requirement (shown below) to overbuild the grid, uniform prices will become competitive prices. From Section 8(1) of the Electric Utilities Act: (e) taking into consideration the characteristics and expected availability of generating units, plan a transmission system that (i) is sufciently robust to allow for transmission of 100 per cent of anticipated in-merit electric energy referred to in section 17(c) of the Act when all transmission facilities are in service, and (ii) is adequate to allow for transmission, on an annual basis, of at least 95 per cent of all anticipated in merit electric energy referred to in section 17(c) of the Act when operating under abnormal operating conditions; (f) make arrangements for the expansion or enhancement of the transmission system so that, under normal operating conditions, all anticipated in merit electric energy referred to in clause (e)(i) and (ii) can be dispatched without constraint.
92
$/MW Constraint
$/MW Constraint Net local demand
Net local demand
Remote supply MWh Loss of load
Remote supply MWh Congestion
Figure 4.3 Relationship of congestion to a transmission-cause reliability problem
The fallacy of eliminating all congestion may arise from confusion between congestion and unreliability. Unreliability is the result of having too little local generation to meet local demand net of imports. This can result from congestion, but in almost all cases, congestion is not associated with unreliability. It is simply the result of having more than enough local generation, but at a cost higher than the cost of remote generation that could be accessed with a larger transmission line (see Figure 4.3). In other cases the desire to eliminate congestion may result from a desire to increase local supply and thereby lower the local market price. For some time, this can save consumers money even if it raises the long-run cost of producing and delivering power. But if expanding transmission lowers consumer costs without lowering total cost, it is a form of monopsony power which essentially expropriates some of the sunk costs of generation in the import-constrained area. Suppose that generation xed costs in an import-constrained region are $40,000/MW-year higher than in the surrounding area, and there are 1,000 MW of such generation installed in the load pocket. A 200 MW expansion of the import line may virtually eliminate extra xed (sunk) cost recovery needed by reducing the local price to the external price. This may save consumers $200,000/year per MW of new line. This is probably much more than the cost of the line. This appears to be a saving, largely because it expropriates $40 million per year in generation xed costs. It also provides a legitimate saving by providing genuinely less-expensive external power to the expensive constrained zone. If capacity produces power in half of all hours and if the new lines capacity is half used, then each MW of line provides a real saving of $40,000/MW, one-fth as much as the initial saving through the expropri-
93
ation of generation sunk costs. In the long run, internal capacity will retire and eventually the generating capacity in the load pocket will again recover its xed costs. Perhaps the savings from the line will amount to $240,000/MW for the rst ve years and $40,000/MW after that. If this has a present value of $1,250,000/MW, then it might be thought that this is the break-even point for building such a line, but that omits the impact of regulatory risk. Investors take account of the cost of expropriations by factoring that possibility into future investment decisions. This demonstration eect will not be conned to the load pocket in question. It is impossible to predict what the cost of this will be, but the resulting investment risk premiums will aect the rate of return on all capital in the aected load pockets, not just the new investment. This is a consequence of a market-clearing electricity price. Probably the most sensible guess is that all of the money transferred to load from existing generators will be lost to load through higher risk premiums. In conclusion, it is inecient to eliminate all congestion, and it is wrong to focus on short-run consumer cost reductions when planning transmission investments. As with any type of productive investment, the goal should be to minimize the total cost of production and delivery. If the market is competitive or the regulation eective, these cost savings will be passed through to consumers. Optimal Transmission (Static) The optimal amount of transmission minimizes the total cost of producing and delivering electricity (Box 4.2). It can be determined in simple examples by using the standard rst-order condition of calculus. At the optimum, the value of an additional kW of transmission equals the cost of building it. Because realistic transmission cost functions can be quite complex, the optimal design may need to be found by evaluating many options rather than applying calculus.
BOX 4.2

OPTIMAL TRANSMISSION
Transmission investment should not eliminate congestion. Transmission investment should not minimize short-run consumer costs. Transmission investment should minimize the total cost of transmission and the production cost of power.
94
BOX 4.3
DEFINING THE RENTAL COST OF TRANSMISSION LINES IN $/HOUR
The cost of a particular transmission line might be $500,000,000, but when analysing power systems it is much more convenient to think of renting capital than buying it outright. Rental cost is naturally expressed in $/MWh, and these units are particularly convenient for computation. This is not how engineers calculate costs, but it is very convenient for economic calculations. Given the one-time cost of the line, its capacity, and a savings per MWh of energy transported, it is generally impossible to tell whether building the line saves money or not. This is because one must know how long the line will last, its maintenance costs, and the debt and equity costs associated with the project. All of these are properly taken into account by a rental cost, and so can be ignored once the rental cost is specied. The one-time cost can be converted to an amortized annual cost, of say, $50,000,000/year. Adding maintenance of say 2,000,000 year and dividing by 8,760 hours/year gives a rental cost of $5,936/h. If the line is a 1,000 MW line, then the cost is $5.94/MWh. This is the cost of renting 1 MW of the line for 1 hour and is independent of whether the line is used or not. Simple investment problems often involve a transmission cost function such as C c bQ, where Q is the MW capacity of the line. In this case it will be convenient to specify C, the lines rental cost, in $/h and c in $/h, and b in $/MWh. Sometimes c will be called the xed cost of the line meaning that it does not depend on the choice of capacity.
A simple example will illustrate the main points. Consider a city (load pocket) that can produce power at a cost of $50/MWh but can buy it for $30/MWh over a transmission line. Suppose the line can be built for a rental cost of $6,000/h plus $5/MWh. How large a line should be built? (See Box 4.3 on dening the rental cost of transmission lines.) To solve the above problem, the citys load must be specied. Suppose the peak load is 800 MW and its minimum load is 400 MW and it takes on intermediate values according to a uniform probability distribution. The savings from the line will be $20/MWh for the rst 400 MW of line capacity and will then decline linearly to zero for additional capacity up to 800 MW.
95
As long as additional capacity saves more in energy production costs than the cost of the additional capacity, the line should be larger. Since the cost of additional capacity is $5/MWh, the line should be expanded until the saving falls to $5/MWh on average. When the capacity is three-quarters of the way from 400 MW to 800 MW, load will be great enough to use the last MW of capacity only a quarter of the time. Thus the savings of the last MW will be only one-quarter of $20/MWh or $5/MWh. Hence 700 MW is the optimal capacity of the line. This conclusion is actually a bit premature. It is necessary to check that the net benet from the line is positive when the xed cost of the line is included. The total benet is $(20 400 12.5 300)/h, which is $10,750/h, while the total cost is $(6,000 5 700)/h, which is $9,500/h. The line is worth building, but if the xed costs had been $8,000/h, it would not have been. This example is the basis of the conclusion that eliminating congestion is almost never the right answer. All transmission lines are used to varying degrees at dierent times of the day and year. Their maximum potential use will occur for only a few hours. To eliminate congestion it is necessary to build enough capacity to accommodate this maximum, but that means the last megawatt of capacity is used only a tiny fraction of the time and it is almost never economical to build capacity for such infrequent use. It might be argued that the lumpiness of transmission investment will naturally cause a choice between a much-too-small line and a too-large line, and that the economic choice will turn out to be the too-large line and no congestion. Besides the fact that this is unlikely to happen for all lines, there is a deeper problem. As load grows, every line reaches a point where its capacity would be exceeded without redispatch for just a few hours per year. To avoid this, a new line, or at least an expansion will be needed, and consumers will have to pay for it. Due to the xed costs this will come to at least $6,000/h in every hour of the year in the above example. Turn next to the general static optimization problem. This asks what capacity line should be built given the way congestion changes with line capacity. A small line will be congested frequently and congestion rents will be high, while a large-enough line will never be congested. What is the condition for the optimally sized line? For simplicity assume that the line connects a local region where the total cost of energy production is CL(Q) to a remote region where it is CR(Q), and assume that power is cheaper in the remote regions so the ow is into the local region. A 1 kW expansion of the line will increase production in the remote region and decrease it in the local region by 1 kW. If the local marginal cost of power is MCL and the remote marginal cost, MCR, then the savings is the dierence. In a competitive market, prices equal marginal
96
costs so the savings is (PL PR) times 1 kW. The line is worth expanding up to the point where the marginal cost of expansion is equal to the average price dierential. Static optimal transmission result (one line) In an optimal one-line network, the marginal cost of line expansion equals the all-hour average absolute nodal price dierential between the ends of the line. In other words, the marginal cost of expansion equals the average congestion rent. In a network, things are a bit more complex. If we have two lines from A to B and power tends to ow equally on each, but A has much less capacity than B, when line A becomes congested it will limit the ow on line B. This limitation is not a physical restriction, rather the system operator will be forced to limit the total ow on both lines to protect the weaker line. If line A is expanded by 1 MW, this will increase the usefulness of line B by 1 MW. Consequently the value of expanding A is twice what would be computed from the ow on A times the congestion price of A. Static optimal transmission result (network) Let d be the power distribution factor on line AB calculated as the fraction of power owing on that line when power is injected at A and withdrawn at B. Then, in an optimal network, the marginal cost of expanding the constrained line equals the average congestion rent divided by d. Optimal Transmission (Dynamic) Power systems are dynamic. Load grows; generators are built and are retired. This dramatically increases the complexity of the optimal transmissioninvestment problem. To illustrate this, consider the simplest example of a dynamic investment problem. Suppose transmission lines come in two sizes, 600 MW at a rental cost of $5/MWh and 1,000 MW at a cost of $4.20/MWh. Suppose that the power transmitted over the path in question starts at zero and grows by 100 MW per year indenitely and that the saving from transmitting this power is always $6/MWh. Obviously it will be economical to build power lines. Building the 600-MW line would cost $3,000/h, and when load has grown to 500 MW, this would save $3,000/h. This is the break-even point, and if the small line is to be built it should be built at this point in time. The alternative is to wait another two years until load has grown to 700 MW, at which point the cost and savings of the 1,000-MW line would be $4,200/h. The smaller lines start saving money sooner, but larger lines will save $1.80/MWh in the long run, while smaller lines will save only $1.00 in the
97
$3,600/h Savings from building 1,000-MW lines
$3,000/h
$2,400/h
$1,800/h Savings from building 600-MW lines
$1,200/h
$600/h
4 Figure 4.4
12
16
20
24
28
32 Years
A positive present value is not sucient
long run. The choice between the two strategies depends on the discount rate, but as can be seen from Figure 4.4, it does not require a very low discount rate to make the larger lines the more economic choice. The conclusion to draw from this dynamic example is that, just because a line is worth building, it should not necessarily be built. It may be better to wait a while and build a larger and more economical project. This is a result of the lumpiness of the investment decision. Note that waiting has no option value in this example because the future is known with certainty. Optimal Transmission (Option Value) In the above dynamic example, two sets of projects were considered; (i) build small lines at optimal intervals, and (ii) build large lines at optimal intervals. Even if there had been uncertainty in this example, each set of projects could have been evaluated to nd its expected net present cost of transmission and generation. If all reasonable sets of projects are evaluated in this manner and the one with the lowest cost is chosen, this constitutes a complete dynamic analysis. Unfortunately the selected project may not be the best choice. This is not because expected net present cost is
98
the wrong criterion, but because the set of choices was unnecessarily restricted. Besides sets of projects, there are also investment strategies. One such strategy might be to build a small line now and then wait until load increased by 600 MW and build another small line if the wait was eight years or more and build a large line if it was eight years or less. A strategy is dierent from a specic set of projects because it waits for more information and then chooses one project or another. Quite often, some strategy will be more cost eective than any specic plan. For example, consider a system with no transmission, one city and a possible remote coal plant with a 50 per cent chance of being built. If it is built, the most ecient transmission project would start now and have a net present value of $200 million. If the coal plant is not built, this project would have a net present value of minus $100 million. Because there is a 50 per cent chance that the coal plant will be built, building the line would have an expected net present value of $50 million ((200 100)/2). If the line were started a year later it would have a (current) net present value $180 or minus $90 million depending on whether the coal plant is or is not built. Waiting a year and then building the line thus has an expected net present value of $45. Other projects could be considered that delay the line for dierent amounts of time or build lines of dierent capacity. But it is quite plausible, and this example will assume, that of all specic projects, building the line now is most valuable. In spite of this, there may be a more benecial strategy for selecting projects. If in one year, we shall know whether or not the coal plant will be built, the strategy of waiting a year and then selecting the best project is more valuable than any particular project. The expected present value of this strategy is 50 per cent of $180 million (if the coal plant is built) plus 50 per cent of $0, if the coal plant is not built. This strategy has an expected present value of $90 million which is $40 million greater than the value of building the transmission line now, the most valuable project given todays information. This $40 million is called the option value of waiting a year to decide. There is often a cost of delaying the start of projects with a positive expected net present value, but there is also generally an option value to delaying the decision to go forward with a project. It is only when this cost of delaying is greater than the option value of waiting to decide that a commitment should be made. Option value is dicult to estimate and it should be estimated for various waiting periods. In short, considering option value greatly expands the number of possibilities that must be considered.
99
2.
COST RECOVERY FOR OPTIMAL TRANSMISSION INVESTMENTS
The previous section demonstrated that the optimal grid will suer congestion which will result in the collection of congestion rents and that these rents are related to the cost of the grid. This raises the question of to what extent congestion rents on the optimal network will cover the cost of that network. To consider this question it is useful to expand the notion of congestion rent. The congestion rent collected on a one-line network is CR(LAB) Q (PB PA), where Q is the power ow from A to B on the line between A and B. This is the trading surplus collected if Q is sold at A and purchased at B. Expanding this concept to the entire grid results in dening the congestion rent for the grid to be the revenue from selling all energy injections at their nodal prices and purchasing all energy withdrawals at their nodal prices. If Wi is the net energy withdrawal at node i and Pi is the price at node i, then the congestion rent for the network G is CR(G) Wi Pi. Considering only lossless networks, it is possible to decompose the set of net energy withdrawals into a set of bilateral trades each with one injection (negative withdrawal) and one withdrawal of equal magnitude. Each bilateral trade from node A to node B, which can be any two nodes on the network, has associated with it a congestion rent, CR(BAB) Q (PB PA), where Q is now the magnitude of the injection and withdrawal of bilateral trade B. Note that one possible decomposition of the net nodal energy withdrawals into bilateral trades corresponds exactly to the power ows on the individual lines. Doing so associates with each line a congestion rent equal to the lines ow times the price at the withdrawal node minus the price at the injection node. The lossless congestion-revenue result If the set of all bilateral trades, B, sums to the total net energy withdrawals from the network, then the total congestion rent is the same whether computed by node for the entire grid, G, as the sum of congestion rents on all lines, L, or as the sum of congestion rents on all trades.
CR (G) CR (L)
All Lines
CR (B).
All Trades
Proof: Since the injections of bilateral trades are paid the nodal price and withdrawals are charged the nodal price, the net revenue collected at a node is the nodal price times the sum of withdrawals minus the sum of injections. Since these two sums add up to the net withdrawal, the net revenue collected from bilateral trades at node i is just Wi Pi, and over all nodes bilateral trading revenues sum to the congestion rent calculated
100
for the entire network. That congestion rents on lines sum to the same value depends on these power ows summing to the net withdrawals on the network. This follows from conservation of energy in a lossless network. The net power ows on lines into a node must sum to the net withdrawal from the grid at that node. This result demonstrates that the trading revenue that is collected from buying and selling power in a congested network with nodal prices (even if these prices are not the competitive prices) will exactly cover the congested rents calculated on a line-by-line basis. This assumes that there is no power loss, so that the power that ows out of one end of a line equals the power that owed into the other end. (In reality losses are typically well under 5 per cent on a high voltage transmission system.) It would be desirable if the trading surplus in an optimally built network were to cover the cost of the transmission network. The above result makes it reasonable to investigate this question by looking at the cost-recovery properties of congestion rent from a single line. Congestion Rents Recover Linear Line Costs Suppose the cost of a transmission line is strictly proportional to the megawatt capacity of the line, so that cost is given by C c Q, where c is in $/MWh and Q is in MW. In this case, according to the static optimal transmission result (one-line) given above, the marginal cost of the line, c, should equal the average congestion rent per MWh, P. As a result, the revenue from the line, PQ, will equal the cost of the line c Q. In the optimal one-line (lossless) network with linear costs, congestion rents will cover the cost of transmission capacity exactly. Not surprisingly this result extends to networks. If the power distribution factor on the Q-MW line from A to B is d, then when line AB is congested, it controls the power ow of Q/d. If the congestion price from A to B is P, then the congestion rent associated with this constraint is P Q/d and the benet of increasing the lines capacity is P Q/d. In this way the cost recovery of constraints in a network can be seen to be analogous to the cost recovery of a single line in a one-line network. An optimally constructed network with linear cost functions and no losses, will recover its xed costs through marginal-cost (competitive) pricing. Congestion Rents and Fixed-cost Recovery When investing in transmission, some costs are roughly independent of the capacity of the line and some are roughly proportional. Those that are inde-
101
pendent will be called xed costs in this chapter. The linear-cost function just considered has no xed-cost component, so to add realism, consider one of the form C f c Q. In this case, the same rst-order conditions will determine the optimal transmission system, so congestion rents will recover c Q, but not f. In other words all costs that are proportional to capacity will be recovered, but none of those that are independent of the lines capacity will be recovered from congestion rent in an optimally sized system. The cost function just considered is one example of returns to scale. Another model of returns to scale has a cost of line capacity that is proportional to the square root of capacity, C a K1/2. In this case the marginal cost of capacity is MC (a/2) K 1/2. If the line is built to the socially optimal level, MC will equal the average congestion rent, which is paid on the whole line capacity, so revenue is R (a/2) K1/2 C/2. In other words, at the socially optimal level of investment the line provides a congestion rent of exactly half of what it costs. This is true regardless of how large or small the optimal line is. Lumpy technology also exhibits xed costs and, at least, limited returns to scale. But the two concepts, returns to scale and lumpy technology, can be usefully distinguished. Figure 4.5 provides the basic intuition. Both types of technology violate the non-convexity assumption required for perfect competition. Both types have a xed-cost component. The lumpy technology shown above requires a xed cost to provide its rst megawatt of capacity, but the marginal cost of capacity is then zero up to Total cost
Returns to scale
Lumpy technology
100 MW
Quantity
Figure 4.5 Lumpy technology may not exhibit returns to scale in the long run
102
100 MW. In this sense, the cost of this technology is all xed. On a small scale, up to 100 MW, there are returns to scale; over large changes in capacity, there are no returns to scale.
3.
TRANSMISSION PLANNING WITH STRATEGIC GENERATION INVESTMENT
Under a planning approach, no performance-based incentive mechanisms are applied to the problem of deciding on long-lived transmission investment, mainly lines and transformers. The planning might be carried out by the independent system operator (ISO) or by a transco, and payments for the cost of investment will be carried out under rate-of-return regulation. To simplify discussion, it will be assumed that the planning is done by a transco that owns and operates the grid, but that there is an ISO which operates the wholesale energy market. Because PBR is not used for longrun investments does not mean it would not be used for the day-to-day operation of the grid, but the problems of ecient grid operation will not be discussed in this chapter (Joskow, 2004). As illustrated in the previous section, planning transmission investments is a complex problem, and is in fact far more complex than these illustrations suggest because of the complexity of the network. This is widely recognized. An additional complication is also recognized. Because generation is not planned, transmission planners facing a wholesale power market must forecast the location of generation and load many years in advance.2 These problems are common to all three approaches mentioned in the introduction (planning, merchant and PBR). Under each approach, the dynamics of the combined transmission/generation system and the option value of waiting must be taken into account. Under each approach the investor will not have control or direct knowledge of future generation investment, and under each, the complex cost structure and network externalities cause additional diculties. This section does not focus on the common problems but on strategic issues peculiar to the planning process when generation investment is deregulated. Strategic Manipulation of the Zero-congestion Policy The possibility of strategic manipulation is particularly acute in the case of a zero-congestion policy, such as Albertas. In the example for static transmission optimization, the optimal line size was 700 MW, and if an 800 MW line had been built there would never have been any congestion. This makes it appear that building for zero congestion would not be too expensive. In
103
this case it would require only a 14 per cent increase in transmission capacity. But consider the case of a wind farm that can be located at various distances from load. If it is close to load, the cost of transmission may be only $1/MWh while if it is in the most remote location it might be $20/MWh. Suppose that the most remote location is the windiest and it is no more expensive to build wind generators there than closer in. What is the eect of a zero congestion policy on the location of this wind farm? Clearly, it is most protable for wind generators to locate in the most remote location whether or not the benet of more wind comes close to osetting the extra cost of transmission. Moreover, since wind power has nearly zero marginal cost, it is always in merit, and consequently transmission capacity must be built to accommodate the windiest hour of the year. The last megawatt of such transmission capacity will have almost no value. (One can be sure that a zero-congestion regulation will be violated in practice simply because adding capacity to a remote location to capture one hour of supply makes so little sense.) This illustrates the fundamental point of the strategic generation investment problem. Strategic generation investment problem The policy of the transmission planners inuences the distribution of installed generation. Hence, planning policies cannot be appropriately compared on a static model of generation and transmission but must be compared using a game-theoretic approach that recognizes generation investment strategies. This problem complicates transmission planning in a market environment because the problems of predicting generation investment and planning transmission for that investment can no longer be separated. For example, one might think that planning could proceed by having two groups of planners, one of which forecasts generation investment, and the other of which takes the forecast and plans the best power lines for that predicted investment. The only communication necessary between the two groups would be the transfer of forecasts from the rst group to the second. Unfortunately, the strategic generation investment problem shows that this will not work. First, the strategic investment problem tells us that the forecaster will need to know what policy the transmission planners will follow. This is necessary because generation investors will react to this policy and the forecasters need to understand this reaction to make good forecasts. Second, the planners cannot simply build the best lines for the predicted generation. They must choose an investment policy that induces good generation investment, so they must understand how generation investment will respond to their policy.
104
The Diculty of Implementing Simultaneous Optimization There is one obvious hope for resolving this chicken-and-egg problem. If the planners choose the truly best transmission investment policy call this the ideal transmission planning policy this rule should induce generators to make the best possible investment decisions. In this case, transmission investment policy should be optimal for the implied generation investment incentives, and these should be optimal for the transmission policy. But this doubly optimal system deserves a closer look as it involves more than technical diculties. At the opposite end of the spectrum from the zero-congestion policy, is the ideal planning policy. This requires the planner to estimate future load growth and then plan both generation and transmission simultaneously to minimize the total expected present cost of delivered power.3 Ideal planning is extremely dicult, but, besides the technical diculty, another problem blocks our path. The ideal transmission planning policy Plan transmission and generation together and optimize both for expected load growth, then build that transmission and hope the market induces suppliers to invest in optimal generation. In principle, this policy works because, given optimal transmission, investors will nd the co-optimized generation to be their optimal strategy. But, both joint optimization and any real markets for generation are subject to error. Consequently, there will be times when ideal optimization directs that generation should be built in location X, and a corresponding transmission line be built to serve that generation. If such a line is built and the market decides not to build generation at location X, the planners will be severely embarrassed by their line to nowhere. The problem to focus on is not the error, but the embarrassment. Errors are taken into account by our theory of minimizing expected cost. The embarrassment causes a more fundamental problem. It prevents planners from adopting the ideal transmission planning policy. Planners will not undertake a project that can lead to such an embarrassing situation. Instead, they will simply attempt to optimize their transmission for generation that has been built, is being planned, or at the least, appears to be an obvious extension of an existing trend. They will not predict optimal generation and build transmission for that prediction. It might seem that this will make no fundamental dierence, but it does. The strategic generation problem tells us that as soon as the generators realize the actual planning policy is no longer the ideal planning policy, they will
105
no longer invest optimally for that policy. Instead they will invest optimally for the new policy, and that is an entirely dierent matter. Transmission planners will not build for theoretically determined future generation investments but will instead build for generation as determined by generation investors. This means that generation investors can manipulate the transmission planners by their selection of generation investment sites. They will learn what policy the planners are using and game that policy. This is the strategic generation investment problem. Strategic Manipulation of Optimal Transmission Planning Having given up on the ideal transmission planning policy, the question becomes, what is the best realistic planning policy? There are many choices, but only a few can be stated simply. One of these, the zero-congestion policy, has already been ruled out. Can we nd a better one? The next-best practical alternative will be considered. This species that the planner should optimize transmission taking generation and anticipated generation as given. This will be called the practical planning policy. The practical planning policy Build the transmission system that is optimal given actual and anticipated generation. If planners follow the practical planning policy, and generation decides mistakenly or perversely to locate in a remote region where fuel is cheap but transmission is so expensive that the combination is uneconomical, the planners may well have to build accommodating transmission.4 Although the result may not be optimal, it will be far more sensible than building to the point of zero congestion. To analyse the planners choice, consider the net social benet of a transmission and generation project. This project consists of remote generation which exports power over a transmission line to the central market. Net social benet consists of (i) the net benet to consumers, (ii) generation prots and (iii) transmission prots. To simplify the calculation, assume that the entire project is small compared with the central market and that long-run supply in the central market is very elastic. This implies that the remote generation project will not change the price paid by central consumers, and as a consequence it will be of no net benet to them. The project, if ecient, will displace central production with cheaper remote production and delivery, but the savings will be entirely captured by the investors. As a consequence net social benet reduces to generation prots plus transmission prots: Net Social Benet Generation Prots Transmission Prots.
106
Assume that remote generators are paid one price and central consumers are charged a higher price which creates congestion rent. From this, the cost of the transmission investment can be subtracted to nd transmission prots, which can be negative or positive: Net Social Benet Generation Prots Line Costs). (Congestion Rents
In eect, remote generation is paying the congestion rents because the central price is determined by the central market cost of supply. If remote generation is required to pay the full cost of the transmission line through congestion rents, then net social benet equals generation prots, and prot maximization by generators will maximize net social benet. In this case suppliers will invest optimally in generation and if the planners follow the practical planning policy, the combined project will be optimal. For linear (proportional) transmission costs, congestion rents do cover line costs for the optimal line. Consequently if transmission costs are linear, the practical planning policy will induce optimal generation investment and the optimal transmission investment for generation. The combined system will be optimal. The corollary of this result is that when transmission costs are not linear, investment is unlikely to be optimal. For example, consider again a wind farm. Under the zero-congestion policy, the wind farm could locate as far as it liked from the central market with complete impunity. Under the practical planning policy, the planner will build only the optimal transmission line and the further from the central market that the wind farm locates, the smaller will be that line. The marginal cost per MW of line capacity increases roughly in proportion to the length of the line for any line capacity. Since the optimal line capacity occurs at the point where the marginal cost of line capacity equals the average congestion rent, the longer the optimal line, the greater the average congestion rent. Since the wind farm will pay these rents, it will care how long the line is. This is good news. The practical planning policy should not only build more economic lines for existing generation, but also induce the existence of a more ecient spatial distribution of generation. The combined saving should be signicant and perhaps enormous. But this does not imply that the induced generation investment will be optimal under the practical planning policy. Suppose the cost of transmission is ck, which implies that for an optimal line, congestion rent is exactly half the total cost of the line. The wind-farm investors will realize that they must pay only half the cost of the line since, under the examples assumptions, the only transmission cost they pay is the
107
congestion rent. When they dierentiate prot with respect to distance, x, from the central market in order to optimize their location, their value for the dCR/dx will be half what it should be and this will cause them to locate too far from the central market. The planners will then be forced to build a line that is optimal for generation located too far away, but the result will be a combined generation-transmission project that is suboptimal. The obvious remedy for this problem is to charge generation the dierence between the congestion rent and the cost of the line. This implies a charge that varies by location and some measure of a generators size, but that is quite dierent from a congestion charge. The need for such charges has long been recognized (Brunekreeft et al., 2004; Vogelsang, 2004) and has been the focus of many economic schemes, few of which have been informed by legitimate economic analysis. Many of these go under the name of MWmile charges, though the most elaborate one was an Alberta scheme called SERP. It based locational charges on an analysis of line-by-line power ows that would be caused by a short circuit at the location in question. Without such a scheme, the fallback position is to charge every MW delivered to the system a xed (independent of location) charge per MWh. This has the advantage of preserving short-run eciency by leaving the dispatch unaected. It will, however, leave the practical planning policy sending inecient long-run signals for generation location. These insights dene a fundamental problem for transmission planning in the context of a wholesale generation market that includes marketdriven investment: The fundamental transmission-planning problem Is there a mechanism for collecting transmission xed costs which, when coupled with the practical planning policy and nodal pricing based on competitive energy prices, will improve the total eciency of the power system? Both shortrun dispatch eciency and long-run incentives for generation investment must be considered. The standard of comparison is a per-MWh charge on all supplied energy independent of location. Although it seems unlikely that any charging mechanism can be found that will be both short- and long-run optimal, it does seem likely that some improvement is possible at least for realistic networks. Because so many erroneous collection mechanisms have already wasted time and money, none should be seriously considered until some proof is given that they will lead to improved eciency on some well-specied collection of power systems. These test cases can be gross simplications of real power systems, but they must include the possibility of locational choices for generation investment.
108
4.
MERCHANT TRANSMISSION INVESTMENT
If transmission is not built by a public or regulated entity, it will be built by private investors. This happened before the power industry was regulated and it happens today. But the question of interest is how much will be built and how eciently it will be built. Certainly there are no easy assurances from theoretical economics that an unregulated market will perform well, and to date there is no empirical evidence for this proposition. There are, however, several theoretical reasons for concern. Returns to scale suggest that investors will need market power to recover their xed costs, but market power, in this industry as in all others, leads to underinvestment. Externalities generated by the use of the grid for trade suggest there will be free-rider problems, which will exacerbate the underinvestment problem. Transmission investment provides two other externalities, the value of which private investors will not easily capture. First it reduces market power in the generation market (Stoft, 1997; Borenstein et al., 2000), and second it provides reliability. Generation market power is a particularly knotty problem because energy suppliers are known to lobby government bodies in an attempt to block transmission investment that is not in their interest. Any attempt to reduce market power with transmission is likely to be the target of supplier lobbying. In spite of these diculties, transmission investment is not always as dicult to nance as many assume. Often congestion rents are viewed as the sole source of remuneration to merchant transmission. In fact, lines may be built without any assumption of congestion income. If there is a cheap but remote area for generation investment, the suppliers that locate there may build lines simply to bring their power to market. They will still have free-rider problems and the like, but they will be motivated by other than future income from congestion rents. Similarly a city may nd local supply too expensive and may build transmission out to the larger network simply to access cheaper power with no thought of future congestion rents. Alternatively all three motivations may coincide. Although congestion rent may not be the primary motivation for building lines, rights to new lines should be given to investors to encourage such investment and internalize the lines benets to the extent possible. Transmission rights can protect an investor from congestion cost that would otherwise have been paid if the investor used the line and can provide income to the extent others use it when it is congested. Another benet of transmission rights is to cause negative externalities to be internalized. Before turning to the diculties of merchant investment, it is worthwhile understanding how transmission rights should be granted in return for investment and what useful role they can play.
109
Transmission Rights Because alternating current (AC) transmission lines are so thoroughly integrated with the power grid, their owners generally cannot be given physical control or the right to charge anyone who uses the line. For example, almost any power ow from one point to another on a power system causes some power to ow on most transmission lines in the grid, although the amount that ows on remote lines is too small to matter. AC power lines are in some ways like a set of connected water pipes without valves between them. Pushing water from one point to another aects the ow in almost every pipe. Because of such physical complexities, the standard proposal is to reward the investor in a transmission line with a set of nancial rights, not physical rights. The standard nancial transmission right is a congestion revenue right (CRR) which is dened by a quantity, Q, source and sink, A and B, and a set of time intervals, T. At any point in time during T, the CRR pays (PB PA)Q, which is the congestion price from A to B times the megawatt quantity of the right. The payment has nothing to do with actual power ows associated with the owner of the right. Although CRRs can, in principle, be privately issued, they are generally issued by the ISO and that will be assumed throughout this discussion. Consequently, at any point in time, there is a well-dened set of CRRs, R, that have been issued. An important property of R is its feasibility, which is dened as follows. Corresponding to every CRR there is an imaginary power ow of Q MW from A to B, during time intervals T. This imaginary power ow has nothing to do with actual ows on the grid. Since every CRR in this set corresponds to an imaginary power ow, we dene R to be a feasible set of rights if the corresponding set of imaginary power ows could take place on the system without violating any reliability constraint. This has nothing to do with load or generation and concerns only the transmission system. The following procedure can be used, at least in principle, to reward investors in transmission upgrades. First, sell a set of CRRs, in an auction that does not withhold any feasible CRR. This should leave no valuable CRR unallocated. When a transmission upgrade is completed the system should be able to accommodate more power ow reliably, and this should expand the feasible set of CRRs. The investor is allowed to claim any set of CRRs which, combined with the existing set, forms a feasible set in the upgraded system. This allows the investor a certain amount of choice and it accounts, to some extent, for positive external aects of the upgrade. There is one more rule. If the upgrade has actually reduced the feasible set of CRRs, the investor must take counter-ow CRRs such that the new allocated set is feasible. These will have negative nancial value and will properly discourage any system downgrades, provided that the initial set of
110
CRRs matched the actual ows on the system (Bushnell and Stoft, 1996 and 1997). Unifying these two rules, the reward for a modication of the transmission system is this: Feasible CRR allocation rule for rewarding transmission investment The modier of a transmission system must take from the ISO a set of CRRs such that together with the pre-existing CRRs the new complete set is a feasible set on the modied system. This approach to rewarding investment has several advantages. First, if the pre-existing set of CRRs matches the ows on the system, it makes it unprotable to damage the system. Second, it gives the investor the maximum possible congestion rent while treating others fairly. Third, for feasible sets of rights, the cost to the ISO of paying CRRs is never more than the congestion rent collected.5 Unfortunately there are also a number of drawbacks. First, the awarded CRRs do not adequately compensate investors. Second, to live up to its potential, the set of CRRs that the ISO makes available needs to be quite complex. For example, the investor may want northsouth rights at some times and southnorth rights at others. Third the results assume investors have no market power in the energy market. Other Styles of Rights There are a number of other styles of transmission rights (Hogan, 2002; Gribik et al., 2004). PJM uses nancial transmission rights (FTRs) which in their original (obligation) form are identical to CRRs except that the revenue associated with the entire set of rights is adjusted to equal the congestion rent collected from the entire transmission system. This correction is generally small. The FTR option (as opposed to obligation), introduced into PJM in 2003, is a fundamentally dierent form of transmission right. At any point in time an FTR option from A to B pays its holder the maximum of what a similar FTR obligation pays, or zero. Unlike FTR obligations, FTR options, like other options, have only non-negative values. Using FTR options instead of obligations results in a smaller feasible set of rights because cancellation of rights in opposite directions is not allowed under the normal meaning of feasible. For example, consider a three-node triangular network with generation at A and B and with load at C, and with a 100-MW line from A to B and two 500-MW lines to C (Figure 4.6). With the standard, obligation-style, CRRs, two 500-MW FTRs, one from A to C and the other from B to C, are simultaneously feasible. With FTR options, only two 300-MW FTRs are feasible because if one option were
111
100 MW
500 MW
500 MW
C Figure 4.6 Option rights reduce the feasible set of rights
not exercised, the other would correspond to a ow that would load the AB line to its limit. Now, the power ows used to compute the feasibility of a set of FTRs are purely imaginary, so the same feasibility rule could be used with options as with FTR obligations. But in this case, two 5,000-MW options, one from A to C and the other from C to A would be feasible (because these ows cancel). The option in the congested direction, from A to C, would earn a payment of 5,000 MW times the congestion price while the ISO would collect only a congestion payment of, at most, 600 MW from A to C. Since the counter-ow option would not have a negative value and would pay the ISO nothing, the ISO would nd itself short of congestion rent by 4,400 MW. For this reason PJM computes feasibility of its options rights as follows. The FTR auctions maximize the quote-based bid value of a set of simultaneous feasible FTRs awarded in the auction. To ensure feasibility, counterow created by an FTR option bid must be ignored when the FTR bids are tested for feasibility.6 There may be no economic justication for the ISO to create option rights. The system operator is in a good position to create standard CRRs because it can back them with congestion revenue without risk and consequently does not need to charge any risk premium for them. But because adding options to the mix reduces the total amount of hedging available, it may be better to let a private derivatives market supply CRR options if there is a demand for them. The motivation for these alternative styles of rights, and for that matter the motivation for transmission rights in general, is mainly to allow the hedging of energy transactions. For example, the main FTR page of the
112
PJM website states that Financial Transmission Rights (FTRs) [are provided] to assist market participants in hedging price risk when delivering energy on the grid. . . . The FTRs provide a hedging mechanism that can be traded separately from transmission service. This gives all market participants the ability to gain price certainty when delivering energy across PJM.7 No mention is made of transmission investment. Market power is another area of concern with transmission rights (Joskow and Tirole, 2000). This also is not closely related to transmission investment. If an energy supplier has market power in a load pocket, it can enhance its power by purchasing transmission rights into the pocket. These rights will pay more when it raises the local price of energy, which makes its exercise of market power more protable. To date, the main use of nancial transmission rights has been as a substitute for prior rights held by transmission owners. This has been quite useful because of the compatibility between CRRs and nodal pricing. This substitution and the more general use of transmission rights has also provided a useful hedging mechanism for nodal price dierences, that is, congestion rents. The value of CRRs in this regard is still not well documented, but they seem to be well accepted in this role. Although their use as a partial incentive for transmission investment has long been advocated, and at least PJM and NYISO have rules in place to this eect, there does not seem to be any documented instance of CRRs playing a signicant role in any merchant transmission project. The Paradox of Transmission Rights The appeal of rewarding transmission investment with CRRs comes in part from their properties in an idealized world of perfect competition. In this economic model, marginal investments are always possible and their cost is linear. Consequently a line may be upgraded by one megawatt for 1/100 the cost of a 100 megawatt upgrade. Moreover, any investor can upgrade any line; there is no ownership of the transmission path. This brings perfect competition to each line in the system. In such a system the congestion rent that would be earned by a marginal upgrade of a transmission path would exactly equal the value of the upgrade in reducing the redispatch cost caused by congestion. For example, if a line is congested for 1,000 hours per year with a price dierential of $10/MWh, then a 1 kW (1/1,000 of a MW) upgrade of the line would save $10 1/1,000 1,000, or $10 per year in redispatch cost by allowing cheaper generation to substitute for more expensive generation. Similarly, if the investor is granted a CRR for 1/kW in the direction of the congestion, the investor will earn $10 per year, exactly what the line is worth.
113
Investment on every path will proceed as long as the marginal investment costs less than the congestion rent earned by that investment. Because this equals the value of line expansion, investment will proceed exactly to the point where it no longer pays to continue investing. Suppose a 100-MW line needs a 50-MW expansion because of a new load. Some investor may build 30 MW of the expansion, but at that point further investment may become unprotable because it lowers the congestion rent on the 30 MW of transmission rights received for the initial investment. In the real world this would most likely stop investment before the optimal transmission capacity is achieved, but in the idealized world of perfect competition, some other investor will continue the investment, perhaps for another 10 MW. Then as this investors stake in high congestion rents discourages further investment, yet another investor will take over the project. In this way every last kilowatt of economic investment will be made. Under the heroic assumptions of perfect competition, rewarding investors with all of the congestion rents provides the ideal incentive for investment (provided that the allocation of rents is also ideal). Under realistic assumptions, which include market power, paying investors more when there is more congestion on their line results in withholding of investment and too much congestion. As will be discussed in Section 5, the exact opposite payment scheme has merit when the investor is a monopoly transco. In this case, charging the investor the amount of the congestion rent instead of paying the amount of the congestion rent results in an ideal investment incentive. Returns to Scale and Iumpiness Returns to scale, as discussed above, means that optimal transmission investments will simply not generate enough congestion rent to pay for themselves. Obviously, this means that merchant investors will build less than the socially optimal level of investment (Joskow, 2003; Joskow and Tirole, 2004). The same does not hold for lumpy technology. Consider Figure 4.7, which shows transmission investment coming in lumps. If the rental cost of a lump of transmission is $8/MWh, it would not be worth building the second lump because the redispatch cost triangle that would be eliminated averages less than $8/MWh.8 Thus the social optimum is to build only one lump and the rent on this lump will be $16/MWh which pays for the line twice over. This shows that with lumpy technology, socially optimal investment may produce more than enough congestion rent to pay the cost of that investment. If we call the lumpy technology in Figure 4.5, linear-lumpy technology, indicating that it has constant returns to scale over a range of lumps, we
114
$/MW 30 Rent 14
1 Lump Demand Cost Supply
2 Lumps
MWh Figure 4.7 Optimal investment in lump technology may be preferable
can ask the question: would merchant investors underinvest in linear lumpy technology? Figure 4.7 shows that they might invest optimally and earn well above a normal rate of return. If demand were a bit greater in the gure, so that the intersection of supply and demand were to the right of 2 lumps, the social optimum would be two lumps. However, merchant investors would fail to build the second lump because it would earn almost no congestion rent. The result would be underinvestment and excess prots. Of the three possibilities, underinvestment, optimal investment and overinvestment, they will avoid the third. A static analysis would conclude that this should lead to too little investment on average but to excess prots. The disconcerting aspect of this conclusion is that it appears that on average, in fact almost always, they would earn above-market rates of return. This seems unlikely. In a dynamic market this result appears even more suspicious. Investors will anticipate the possibility of high rates of return and will invest a little early, which will increase their average investment and lower their rate of return. Considering this dynamic eect, there seems little reason to suspect that lumpy technology will lead to systematic underinvestment, or even that investment will be wrong on average. This is an optimistic result, but one more eect needs consideration. Lumpy investments pay least when they are rst made and most just before the next investment is completed. In fact, many optimal transmission investments lead to a protracted period with virtually no congestion and hence no congestion rent when they are rst completed. This eect may be dramatic. When a lumpy upgrade is rst made it is common for the line to be almost completely uncongested and this situation may last for years. The island of Nantucket o the coast of Massachusetts is served by a single direct current (DC) cable which is likely to become congested in the coming
115
few years for a few hours in August. This might cause a very partial blackout during the tourist season which is unacceptable, so a second identical cable will soon be added. This one will not be congested for perhaps another 20 years. Even when a lumpy transmission investment can expect full recovery of its costs from congestion rents over the long run, the cost recovery may well not begin for years and will be very slow when it starts. This back-end loading of the revenue stream creates grave risks for the investor. What if a new technology, such as cheaper high-voltage DC lines or aluminumzirconium wires, comes on the market before high levels of congestion kick in? What if load growth is slower than anticipated? What if gas pipelines are built to fuel new generation that competes with power imported on the transmission line (Barthold, 2003)? This investors payment stream does not mirror the stream of social benet which results from the elimination of, or reduction in, previous congestion rents. Because that benet stream starts out at the rental cost of the line, it is far less risky than the stream of congestion rent. Risk is costly, so merchant investment based on collecting congestion rents from CRRs issued in return for the investment will be much more costly than a socially sponsored investment in the same project. Low-risk investing is cheaper than high-risk investing. A simple example may help explain the relationship between the social benet stream and the congestion rent stream on a lumpy transmission investment. Suppose load in a load pocket takes on values between X and X 200 MW with a uniform probability distribution. Suppose the price dierence between supply from the load pocket and external supply is $20/MWh. Suppose additional transmission costs $5/MWh and comes in 100-MW lumps. When should transmission be built? Only when the present line is congested would a new line add value. When the line is congested half the time, as shown in Figure 4.8, the value added will range from zero when it is just barely congested to $20/MWh of new-line capacity when the load is at its peak value and the new line is fully utilized. On average, under these conditions the new line would provide $10/MWh of line capacity in social benet while congested and $0/MWh while not congested. Consequently when the existing line is congested half of the time, it will provide $5/MWh of benet on average. Since this is what the line costs to build, this is the break-even point. When the line is congested less often, investment is not warranted. If new lumps of transmission are added at the socially optimal times, congestion rent just before the new addition will be $10/MWh on average, and just after the new addition it will fall to zero. When the new line is rst put in place the line will earn nothing, but its earnings will grow until the next lump of transmission is built. If the growth in X is linear, then, on average,
116
X+ 200
Use of new line
Existing line X
Load duration Figure 4.8 Optimal investment eliminates congestion
the line will earn $5/MWh, exactly enough to cover its cost. (This example assumes a zero discount rate.) At least in this case of lumpy investment, optimal investing will be rewarded with exactly the right level of congestion rents. Note the dierence between the social benet from a transmission upgrade which starts out covering the rental cost of the line on day one and the stream of congestion rents which ow to the merchant investor. These start out at zero and only reach the break-even point half way to the point in time when the next investment will be made and rent will again drop to zero. Also note that, as shown in Figure 4.9, the social benet of building the line is much greater than its cost. It is normal to nd such consumer surpluses in a wellfunctioning market. Obviously, this is a very narrow result, but it disproves a common view which holds that lumpy investments inherently underpay investors similar to the way in which increasing returns to scale (a declining marginal cost curve) underpay optimal investing. The main problem with lumpy investments is that they pay o merchant investors very late, which makes them extremely risky for a merchant investor even though risk in social benet is low. This can greatly increase the cost of merchant lines relative to their cost if built under rate-of-return regulation. Free Riders? When large merchant transmission projects are contemplated it is often noted that many will benet from such a project in the initial years but all will attempt to avoid paying for it. As soon as the line is completed certain
117
$20/MWh Social benefit
$5 Cost of line
Congestion
Rent time
Figure 4.9
Investors should not capture full social benet
loads will nd their prices reduced and certain suppliers will nd they can sell at a higher price. Both will want the line to be built, but all will want others to pay for it. Figure 4.9 illustrates such a situation. In the rst year after the line is built, market participants will benet by $5/MWh of line capacity, yet the investors will be paid next to nothing in congestion rent. This observation usually leads directly to the conclusion that all those who benet without paying are free riding. But, in this example, that is not the case. The investor will be paid in full and the gap between social benet and congestion rent is not the result of free riding but is simply normal consumer surplus. Of course, transmission projects are likely to suer from the eects of returns to scale as well as lumpiness, so it is likely that investors will be underpaid if they depend on congestion rents. In this case free riding is a correct diagnosis, but it will be extremely dicult to assess the extent of the free riding. In particular it is wrong to believe that investors should capture the entire consumer surplus even at the beginning. An investor with market power may be able to capture some of the value that would otherwise accrue to free riders. Similarly, an investor with market power may be able to capture part of the normal consumer surplus that would be provided by optimal investing and complete xed-cost recovery. Both reasons probably help to explain the many proposals to allow the exercise of market power by merchant transmission investors. Some market power would help cover investment costs, so there is some legitimacy to the suggestion. But once started down this path, merchant investors may well ask for more market power than needed to break even. Proposals to base the fundamentals of cost recovery on market power fail to provide a rationale for a market that must rely largely on market power to cover costs. The investment eciency of such a market is unclear at best.
118
The principal argument for eciency in such a market is that suppliers with market power still minimize costs, given their output. But this argument also holds for suppliers who do not decide what transmission is needed but simply respond to an auction held by the system operator for the provision of certain transmission capacity. Until some means can be found of tailoring the exercise of market power to provide the right level of xed-cost recovery only on ecient transmission investments, the argument for the deliberate introduction of market power as a method of inducing investment is weak to non-existent. Mixing Planned and Merchant Transmission The possibility of planned transmission both discourages and threatens merchant transmission. Before a merchant line is built, potential subscribers to the project would prefer to induce the planners to build the line and spread the cost over the broader market. This discourages participation in the project by those who should buy a long-run contract for the use of the line. Essentially, this exacerbates the free-rider problem. Once a merchant line has been built, those who have not pre-paid for its use will still wish to encourage the planners to build a competing line, as actually happened in Australia (Firecone, 2003; Littlechild, 2004). As the merchant line was a DC line, it could directly charge those who used it (Brunekreeft, 2004b), but had it been an AC line those wishing to use the path would still have reason to lobby the authorities to overinvest by building a second line, thus driving down congestion rents and providing a cheaper alternative than use of the merchant line. This possibility is a threat to merchant investment. Because merchant transmission has so far proven itself entirely inadequate, some have suggested a mixture of merchant and planned transmission (Hogan, 1998; Rotger and Felder, 2001; Chandley and Hogan, 2002). For this to succeed, the level of discouragement and threat must be reduced. This can be accomplished if the role of planning can be dened with enough clarity. If merchant investors know which lines will and will not be built by the planners, then planning should not discourage investment in lines that will be needed for merchant purposes, but which the planners will not build. One suggestion for the bright line between planned and merchant lines is that planners will build only lines which would require the cooperation of many parties because in this case the free-rider problem is thought to be particularly severe. While this would give merchants clear guidance at the two extremes, there would inevitably be an important middle ground of ambiguity, perhaps encompassing most transmission expenditures.
119
Another proposal for separation was made several years ago in the context of the Alberta market (Stoft, 2002). That market, to a greater extent than most, fails to send adequate locational signals for generation investment and, like all power markets, lacks proper real-time demand elasticity. Consequently, transmission investment is occasionally required for reliability purposes. It was proposed that the planners build only for reliability, and that when such a project is undertaken, merchant investment be allowed to expand the project for the incremental cost of the expansion, thus avoiding signicant xed costs. To further facilitate merchant investment, it was proposed that the transmission administrator (planner) also facilitates joint investments by joining merchant projects under certain circumstances when lumpiness is a problem. The transmission administrator would buy a part of the line and keep it out of use until a new party decided to purchase it. This proposal was not viewed as ideal, but only as a better alternative than the rule Alberta eventually did implement, requiring that congestion be completely eliminated. It also has the advantage of not depending on or blessing the exercise of market power.
5.
PERFORMANCE-BASED REGULATION FOR TRANSMISSION MONOPOLIES (TRANSCOS)
The planning process provides non-directive and generally weak incentives. Planners know that if they do a demonstrably poor job, they may nd themselves out of work. This provides an incentive and most engineers are actually quite motivated by this and by professional pride and a desire for professional recognition. Consequently, it is a mistake to believe that the planning approach lacks good incentives. However, these incentives may dier from a pure incentive to minimize the total cost of delivered power and may put too much weight or not enough weight on reducing complaints about the occasional outage or about congestion that inhibits trade. Consequently, it may be better to design explicit formulas that determine monetary rewards for minimizing total cost. These rewards cannot easily be applied to individuals, so the standard approach is to apply them to the prots of a regulated monopoly, a transco. Any regulation of a monopoly provides nancial incentives, but often these have not been explicitly designed or even considered. The incentives of cost of service regulation are usually poorly thought out and derive mainly from unintentional lags in rate setting and the subjective application of rules such as the requirement that investments must be used and useful. When nancial incentives are explicitly designed, the regulation is called performance-based regulation (PBR), or incentive regulation.
120
A Direct Approach to PBR for Transmission Investment If it is assumed that power is available to all customers at the competitive price, then the objective of transmission investment is the minimization of the total cost of delivered power. Without the assumption of availability, cost could be minimized by maximizing blackouts, and in the limit, delivering no power at all. Because of occasional blackouts (load-shedding events), the cost-minimization framework must be maintained by assigning a cost to unserved load. Assuming that this assignment of cost can be accomplished, the goal of transmission investment is total cost minimization. This goal is easily translated into a theoretical scheme for incentive regulation (Gans and King, 2000; Lautier, 2000). A monopoly transco should be paid a xed but generous sum, R, per megawatt of delivered power less the cost of congestion (CE, the redispatch cost) and less the cost of unserved (lost) load, CLL. In most power systems, $10/MW hour of delivered power would be more than sucient for R.9 The transcos prots would then be: Prot R CE CLL CT,
where CT the rental cost of the transmission system. As with a single transmission line, the rental cost of the system includes the cost of capital as well as maintenance. Note that the transco does not keep the congestion rent. If the transco can reduce the sum of CE and CLL by more than $1 by investing and thereby raising CT by $1, it will nd it protable to do so and this will be benecial to society. Hence this incentive mechanism aligns the transcos incentives with social welfare. Any reduction in CE CLL CT increases the social surplus by the same amount, and this amount goes into the pocket of the transco. Note that this incentive scheme properly rewards eort which has a non-monetary cost to the transco and is consequently not observable by the regulator. If the transco can increase its monetary prot, as dened above, by $10, but only by expending $9 worth of unobservable eort, it will pay it to do so, as it should, since this is socially benecial. As will be seen shortly, this property is shared only by what are called high-powered incentive mechanisms. This scheme presents three diculties, measuring CE and CLL, and setting R. Congestion costs, CE, are the dierence between the actual production cost of energy and the lower cost that could be achieved without any transmission limits. This dierence can be fairly well approximated in any system with centralized bidding and nodal pricing. Occasionally there may be some diculty with knowing how much certain generators could
121
produce were they not constrained by the transmission system, but most systems have on record a realistic estimate of each generators output capacity and this should serve as an adequate proxy for the true value. (Losses are quite easily estimated.) The cost of unserved load is far more problematic. The standard error for such an estimate is probably a factor of three. In other words, if it is estimated to be $15,000/MWh, there is probably only a 68 per cent chance (roughly) that the true value is between $5,000/MWh and $45,000/MWh. (This is a purely subjective conjecture.) More problematic is the fact that this cost occurs erratically. Major cascading blackouts may happen about once every 20 years and result in half the load being lost for six hours. The cost of such a blackout would be 6 0.5 $15,000 per MW of load which is equivalent to about $7/MWh for an entire year. When such a blackout does occur it could cost the transco a years revenue and result in bankruptcy. Perhaps a solution to this is to undervalue lost load by a factor of 10 or 20 and use a nominal value such as $1,000/MWh. This would still result in a rather erratic cost stream, but it might be tolerable. Certainly, consumers would pay a signicant risk premium to a transco under such an incentive. Sometimes it will also be dicult to tell whether loss of load is due to generation or transmission problems, and this will lead to litigation costs and other ineciencies. The problem of setting R is the most fundamental problem of regulation. The regulator always has less knowledge (information) of the costminimizing solution than does the regulated rm. Because of this, the regulated rm can extract some information rent from the regulator. In general the stronger the incentive, the better the performance of the regulated rm, but the more rent it can extract. The present scheme provides the strongest incentive; the transco keeps every dollar of cost that it saves. As a consequence it will be able to extract considerable rent. Specically, the regulator knows that if it sets R below total cost, CE CLL CT, the transco will go out of business, a result that must be avoided, but it does not have a good estimate of total cost. Its only reasonable choice is to set R three or four standard deviations above the cost-minimizing investment level of CE CLL CT. Given the level of uncertainty in estimating this total, this is likely to result in an extremely high rate of return for the transco. Two Approaches to Reducing Information Rent The advantage of the PBR scheme just described is that it strongly motivates the supplier to expend cost-saving eort that is dicult if not impossible for the regulator to observe. The disadvantage, as we have just seen, is that it requires the payment of high information rents. In the case of transmission
122
investment, these rents could be extremely high and well beyond any acceptable level. Because of such high information rents, it is desirable to reduce the power of the incentive mechanism. Many types of PBR are forms of pricecap regulation including the mechanism just described, though it is a rather non-standard price-cap incentive. In that scheme, R can be thought of as the xed part of a two-part transmission price, the other part being the congestion price. There are two standard ways of reducing the power of a price-cap incentive. First, the cap, R in this case, can be reset periodically. The more frequently it is reset, the lower the power of the incentive it provides. Second, prots under the mechanism can be shared between the monopolist and the consumers. The smaller the share kept by the monopolist, the lower the power of the incentive. In both cases, lower power will correspond to lower information rents paid to the transco. Consider the periodic resetting of the price cap. When the price cap is reset, the objective is to provide the monopolist with a certain allowed rate of return during the next period. To this end, the values of CE , CLL and CT will be estimated for that period. If the period is short, most of the transcos costs (CT) will be correctly anticipated and covered by the allowed rate of return. The longer the period, the more cost will be saved or incurred unexpectedly. This will lead to unexpected changes in CE , CLL which will change revenues. These intra-period changes in expenditure (CT) and revenue (R CE CLL) result in prot deviations from the targeted allowed rate of return and this provides some incentives for both cost minimization and benecial investment. The longer the period between rate cases, the greater the proportion of expenditures for which the price-cap mechanism can provide an incentive. At one extreme lies pure price-cap regulation and at the other pure rate-of-return regulation. In between we nd actual rate-of-return regulation in which price caps are reset roughly every three years. Timing is the major problem with using periodic price-cap setting to achieve a lower-powered incentive and lower information rents. Investment costs are often incurred over a much shorter period of time than the benets from the investment. Hence the reset period may be long relative to costs but short relative to benets. In this case a lower proportion of costs than revenues will be captured in the resetting process. This will tend to discourage ecient investment. This is related to the well-known incentive problems that occur shortly before a rate case. At this time, it becomes advantageous to make costs appear high and revenues appear low. Also the incentive for investment is diminished shortly before a rate case, because the regulator may view such expenditures as already paid for. Prot sharing avoids these timing issue because it does not make periodic adjustments, but instead shares prot in some xed proportion on a
123
continuous basis. Economic prot, for example R CE CLL CT, accounts for a normal rate of return on investment (in CT), so prot sharing can never prevent a supplier from achieving a normal rate of return, it will only bring it closer to that level.10 If unshared prot, , is maximized by a certain investment strategy, then half of the unshared prot, /2, will also be maximized by exactly the same investment strategy. Consequently, if there were no information problem with computing , prot sharing would leave the transcos behavior unchanged. This would be ideal. The excess wealth transferred by the high prots of pure price-cap regulation could be reduced by any amount simply by setting the sharing parameter appropriately and this would cause no loss of eciency. But there are information problems, and the fundamental trade-o of regulation ensures that, if the power of the incentive is greatly reduced, eciency will suer. That is true whether the incentives power is reduced through periodic resetting of the cap, as previously described, or through continuous prot sharing, but it is easier to explain the eect in the continuous prot-sharing context. To glimpse the contradiction inherent in ignoring the information problem, consider, the case in which the dollar-valued economic prot, , divided by the invested capital needs to equal 50 per cent for the initial determination of R in order to avoid any signicant probability of bankruptcy. If the normal rate of return on equity is 15 per cent, then the monopolist will start out making a 65 per cent rate of return on equity.11 To reduce this, consider a prot-sharing ratio of 1 per cent for the transco and 99 per cent for load. This reduces to /100 and brings the initial return on equity to 15.5 per cent. If the transco raises to 100 per cent with a superb eort, it will receive 16 on equity and if it performs terribly, letting fall to 0 per cent, it will still receive 15 per cent on its equity. Even though it prefers 16 to 15 per cent, this limited reward is not likely to induce the eort level required to raise divided by invested capital from 0 per cent by 100 per cent. The important point about eort is that it is a real cost that is not included in CT because it is not monetized. It is a cost that does not appear on the books. It will, of course, aect costs that do appear on the books. Eort will reduce these costs for the same level of transmission performance or increase performance for the same level of monetary cost. Lack of eort raises monetary cost relative to performance. Lack of eort, like eort, can take many forms. It can take the form of gold-plating oces and equipment or shirking by management and workers. Inecient expenditures can purchase tickets to the San Francisco Giants baseball games as was done by a major California utility. Graft is another type of lack of eort; for example, the transco can subcontract a job to someone
124
who pays a kickback to someone at the transco. As the rewards for eort (good behavior) fall, all types of unwanted behavior will increase. Prot sharing on a 50/50 basis takes away half of , which means it takes away half of revenues minus costs. If eort had a monetary value, then a $10 eort that produced an $11 revenue would increase by $1 before prot sharing and by only $0.50 after prot sharing. In either case the eort is worthwhile and will be induced. But because the eort goes unobserved, its cost is not shared by the prot-sharing mechanism, and so the result of the $10 eort is an after-sharing revenue of $5.50 which fails to compensate for the $10 eort. Consequently, with 50/50 prot sharing such eorts will not be undertaken. A $10 eort that produced a $30 increase in revenue would still pay o even after prot sharing allowed the transco to keep only $15 of the revenue. So prot sharing does not eliminate the incentive to provide unmeasured but costly eort, it only reduces it. As the share of prots kept by the transco decreases towards zero, the incentive to expend eort decreases towards non-existent. This explains why prot sharing is limited as a means to control information rents. Diculties with PBR for Transcos The price-cap mechanism just described is most likely impractical because it suers from at least two severe diculties. First, just as with merchant transmission the transcos investments will pay o with very long lag times and are consequently very risky (Brunekreeft and McDaniel, 2005). Typically, a large cost must be incurred over a period of several years, then for several more there will be little or no return on the investment and nally, 10 or 15 years after the start of the project, signicant payback will begin. This is just one possibility, but a very plausible one. The risk of this delayed payback contrasts sharply with societys risk, which is much less because the societal payback starts immediately upon completion of the line at a rate equal to the rental cost of the line. Second, transmission investments are tightly linked to reliability (Crew et al., 2004; Sun et al., 2004), and the reliability part of this mechanism, CLL, introduces severe risks and promises to be extremely controversial if attempted. Most of the major blackouts in the United States over at least the last 35 years have been linked more to transmission problems than to generation problems. This has tended to involve operations and investments other than major line and transformer investment, such as tree trimming, computer systems for state estimation, and the setting of line-trip relays. It should be possible to design a separate mechanism for performance in these areas, but that will not completely disentangle reliability from line investment.12
125
The motivation for upgrading a transmission path, which typically involves several lines, is congestion on that path and the congestion cost, CE, that it causes. But this cost can be reduced in two ways, rst by a physical upgrade, and second by re-rating the path to a higher capacity. The second approach is far cheaper, in fact is almost free, but it decreases reliability. Unfortunately path ratings are not easily audited as they are somewhat controversial even among engineers. This is because they are not based primarily on hard data, such as the temperature at which a wire melts. Instead, ratings must couple hard data with somewhat subjective data, including probabilities of contingencies such as line and generator outages.13 Any powerful incentive to upgrade lines will also be a powerful incentive to cut corners on contingency ratings. Consequently if there are strong incentives to upgrade lines, there must also be heavy penalties for cutting corners on path ratings. Since path ratings are too dicult for regulators to monitor, these penalties must instead be applied directly to blackouts which are very costly but occur very rarely. Imposing the cost of lost load, CLL, is such a penalty, but as explained, it introduces severe risks and would be extremely controversial. The danger of degrading reliability appears to create severe diculties for designing a useful PBR incentive for upgrading lines. A third diculty, less severe than the rst two, faced by any transco proposal is the fundamental planning problem described above. Any realistic transco incentive will induce the transco to invest in the lines that are optimal for existing generation, not for optimal generation. The consequence will be that generation investors will build generation based on the transcos response. For this circularity to produce the ecient outcome, the charges used to supplement the congestion rents and provide the transcos revenues must be allocated in a way that induces the correct location of generation. A number of PBR alternatives for transcos have been suggested, including some by Vogelsang based on the price-cap tradition, and which are reviewed by Roselln (2003) and Vogelsang (2004). Vogelsangs most recent proposal is related to these and the social surplus mechanism described above. Unfortunately this proposal is still not able to solve the investment problem taking into account reliability and load growth. His current conclusion is Long periods mark the limits of regulatory commitment and are still short relative to network investments. As a result, incentives should be further weakened by adjustments based on rate-of-return regulation with a used and useful criterion. (p.141). PBR for transcos will be useful for shorter-term incentives (Joskow, 2004), but it cannot yet be relied on to solve the long-term investment problems.
126
6.
CONCLUSION
Currently, wholesale power markets are undergoing a slow and inecient development process marked by such events as the California meltdown, the overbuilding of gas-red generation in large parts of the eastern United States, the largest blackout in US history, and the complete redesign of the British market. In particular the generation investment problem seems to remain far from solved, though some reasonable incentive mechanisms seem to be on the drawing board. Moreover, it should be recognized that transmission investment is crucial to the functioning of the new energy markets. The less congestion, the less market power in wholesale energy markets (Stoft, 1997; Borenstein et al., 2000; Gilbert, et al., 2002). For example, San Francisco and New York both suer from serious market power because both have limited transmission and must rely for signicant portions of their energy on local suppliers. The more transmission into these cities, the more competition in the wholesale energy market. Moreover, in every market in the United States, there are numerous examples of generation units under regulatory must run contracts. These give the market administrator the right to require the plant to run and in return provide regulatory payments which are often substantial. Such contracts exist largely where these generators have extreme market power during some hours of the year because of transmission limitations. Such situations have been numerous and problematic from the beginning and show no signs of disappearing. Fortunately it is extremely cheap to overbuild the transmission system a small amount and thereby reduce market power below the level that would be found under an optimized network. This is because, at optimal investment, the derivative of total system cost with respect to increased capacity is zero. That is the rst-order condition for optimality. Neither a merchant approach nor a PBR approach is conducive to alleviating wholesale market power problems by overbuilding the network. Both have biases towards underinvestment, and both are likely to be erratic in their behavior during the decades it will take to tame the likely aws in their designs. Rate-of-return regulation is more adaptable. Provided the regulator declares in advance that a line will be considered used-anduseful, it should not be dicult to get the transco to build it. Neither a merchant approach nor a transco/PBR approach has yet been developed to the point where it could be considered useable in practice. Both appear to be in rather early stages of theoretical development. Moreover, it appears that any application of these approaches will require a level of understanding and subtlety that is not yet apparent among regulators, at least in the United States. Given these diculties and the poor
127
record of the energy market deregulation process, it appears to be too early to begin any policy initiatives based on either of these approaches. With transmission investment costs amounting to only 38 per cent of the retail costs (Joskow and Tirole, 2002), it is better to rely on a relatively safe approach to transmission investment, even though it is a bit less ecient than results promised by some poorly understood theoretical approaches. This is not to say that merchant transmission investment should be discouraged. It should be allowed and regulated only lightly (Brunekreeft, 2004a), but it cannot be depended on and should not be the focus of those concerned with ensuring sucient transmission capacity. Rate-of-return regulation applied to a transco or to wire companies under the direction of an ISO is not without its drawbacks. Besides the fundamental planning problem discussed above, there is the additional problem in the context of a deregulated generation market that both generators and load will constantly lobby regulators for more or less transmission. If the fundamental planning problem is not solved, some generators will lobby for more transmission to get a free ride to market, while others will argue for less to keep their local price up and allow them to exercise more market power. Load will argue for more transmission from cheap regions to expensive ones, not just to save production costs and dampen market power (legitimate reasons), but also to exercise monopsony power against generation in the high-cost regions. For years to come, transmission investment appears to be the knottiest problem in the deregulation process.
NOTES
1. Returns to scale imply that transmission investment costs are non-convex. Lumpiness refers to having to buy an integer number of transmission lines selected from a small set of available capacities, but it is better understood as simply referring to a cost function with a uctuating slope. This is not as bad as it sounds, because with growth, almost any line that saves more than its rental cost at the time it goes into service will continue to be economic in the long run. In fact it requires even more. Load is to a small extent determined endogenously by the price of power, so it should be determined simultaneously, but this complication will be ignored. This section concerns strategic manipulation by generation, but it assumes that the wholesale generation market is in every respect competitive. Manipulation is not the result of market power, but of generations rst-mover advantage in the optimization game. This assumes that the set of feasible transmission ows is convex, which is not quite true, and that the feasible set of rights is based on average feasible power ows and not the power ows that are possible only under ideal conditions. From PJMs document FTR market frequently asked questions, updated February 1, 2005, question 49. Found on www.pjm.com/markets/ftr/ftr.html, the website of PJM Interconnection, a regional transmission organization (RTO) that coordinates wholesale power trading in
2. 3. 4.
5. 6. 7.
128
a region of the US from Michigan to Washington DC, to Kentucky. It includes 44 million customers and 135,000 MW of generating capacity. On the left is the cost of $16/MWh, but it falls linearly to zero before reaching the right end of the lump. For completeness the cost of replacing losses should be included in CE. For example if prot 1, then net operating revenue, R CE CLL, equals CT 1, which means that transmission investments are earning more than a normal rate of return. If 1 prot is cut in half, operating revenues still cover CT , and investments still earn more 2 than a normal rate of return. Economic prot is prot above the normal rate of return on invested capital. Wilson (1997) addresses reliability in a franchise transco context. Stability ratings, though more rmly based in physics, also involve subjective judgments as to how close is too close to the point of instability.
8. 9. 10.
11. 12. 13.
REFERENCES
Barthold, L.O. (2003), Whither EHV? Distributed generation reverses the trend, IEEE Power and Energy Magazine, 88, 85. Borenstein, S., J. Bushnell and S. Stoft (2000), The competitive eects of transmission capacity in a deregulated electricity industry, Journal of Economics, 31 (2), 294325. Brunekreeft, G. (2004a), Regulatory issues in merchant transmission investment, Cambridge Working Papers in Economics, CWPE 0422, and CambridgeMIT Working Paper, 38. Brunekreeft, G. (2004b), Market-based investment in electricity transmission networks: controllable ow, Utilities Policy, 12 (4), 26981. Brunekreeft, G. and T. McDaniel (2005), Policy uncertainty and supply adequacy in electric power markets, TILEC Discussion Paper, DP 2005-006. Tilburg Law and Economics Center. Forthcoming in revised and shortened version: Oxford Review of Economic Policy. Brunekreeft, G., K. Neuho and D. Newbery (2004), Electricity transmission: an overview of the current debate, Cambridge Working Papers in Economics, CWPE 0463, and CambridgeMIT Working Paper, 60. Bushnell, J. and S. Stoft (1996), Electric grid investment under a contract network regime, Journal of Regulatory Economics, 10, 6179. Bushnell, J. and S. Stoft (1997), Improving private incentives for electric grid investment, Resource and Energy Economics, 19, 85108. Chandley, J.D. and W.W. Hogan (2002), Independent Transmission Companies in a Regional Transmission Organization, Cambridge, MA: Center for Business and Government, John F. Kennedy School of Government, Harvard University. Crew, M., P.R. Kleindorfer and M. Spiegel (2004), Reliability, regulation and transmission investment, Risk Management and Design Processes Center, Working Paper 04-20-PK, Philadelphia. Firecone Ventures Pty Ltd (2003), Regulatory and institutional framework for transmission (Australia): Final Report, Victoria, November. Gans, J. and S. King (2000), Options for electricity transmission regulation in Australia, Australian Economic Review, 33 (2), June, 14561. Gilbert, R., K. Neuho and D. Newbery (2004), Allocating transmission to mitigate market power in electricity networks, Journal of Economics, 35 (4), 691709.
129
Gribik, Paul R., D. Shirmohammadi, J.S. Graves and J.G. Kritikson (2004), Transmission rights and transmission expansions, draft to IEEE Transactions on Power Systems, 2004. Harvey, S., W. Hogan and S. Pope (1996), Transmission capacity reservations implemented through a spot market with transmission congestion contracts, mimeo, Harvard University. Hogan, W.W. (1992), Contract networks for electric power transmission, Journal of Regulatory Economics, 4, 21142. Hogan, W.W. (1998), Transmission investment and competitive electricity markets, mimeo, Harvard University. Hogan, W.W. (2002), Financial transmission rights formulations, mimeo, Harvard University. Joskow, P.L. (2003), Remedying undue discrimination through open access transmission service and standard electricity market design, AEI-Brookings Joint Center for Regulatory Studies, Regulatory Analysis, 03-1, Washington, DC, February. Joskow, P.L. (2004), Performance based regulation, May (slides), Washington, DC. Joskow, P.L. and J.J. Tirole (2000), Transmission rights and market power on electric power networks, Rand Journal of Economics, 31 (3), 45087. Joskow, P.L. and J.J. Tirole (2002), Transmission investment: alternative institutional frameworks, mimeo, Toulouse. Joskow, P.L. and J.J. Tirole (2004), Merchant transmission investment, Journal of Industrial Economics, 53 (2), 23364. Lautier, T.-O. (2000), Regulation of an electric power transmission company, Energy Journal, 21 (4), 6192. Littlechild, S. (2004), Regulated and merchant interconnectors in a Australia: SNI and Murraylink revisited, Cambridge Working Papers in Economics, CWPE 0410, and CambridgeMIT Working Paper, 37, May. Roselln, J. (2003), Dierent approaches towards electricity transmission expansion, Review of Network Economics, 2 (3), 23869. Rotger, J. and F. Felder (2001), Promoting ecient transmission investment: the role of the market in expanding transmission infrastructure, sponsored by Transnergie US Ltd, November, Westborough, MA. Stoft, S. (1997), The eect of the transmission grid on market power, Lawrence Berkeley National Laboratory Working Paper, LBNL40479, Berkeley, CA. Stoft, S. (2002), A proposal for long-run and short-run congestion management in Alberta, Before the Alberta Energy and Utilities Board, File No. 1803-4, March 4. Sun, H., M. Sanford and L. Powell (2004), Justifying transmission investment in the markets, Electricity Transmission in Deregulated Markets, December 1516, Carnegie-Mellon University, Pittsburgh, PA. Vogelsang, I. (2004), Transmission pricing and performance-based regulation, Electricity Transmission in Deregulated Markets, December 1516, CarnegieMellon University, Pittsburgh, PA. Wilson, R. (1997), Implementation of priority insurance in power exchange markets, Energy Journal, 18 (1), 11123.
130
GLOSSARY AND SYMBOLS

CE Congestion cost. Excess cost of energy from dispatching out of merit order because of transmission constraints. CR Congestion rent. Revenue from energy injections at nodal prices less revenue from energy withdrawals at nodal prices. CL Congestion cost to load. Excess cost of energy to load due to transmission constraints. CLL Cost of unserved (lost) load. CT The rental cost of the transmission system paid by a regulated transco. Includes the annualized costs of capital and maintenance. CRR Congestion revenue right which pays (PB PA) Q, during time periods T. FTR PJMs nancial transmission right. The obligation variety is the same as a CRR except that the revenues are adjusted for any revenue surplus or insuciency in total congestion rents. The option variety omits the negative payments possible with an obligation. G The transmission grid. ISO One of several independent system operators that run markets in the US. They have now switched status to become RTOs, (regional transmission organizations) in keeping with the Federal Energy Regulatory Commissions changing terminology. K The capacity in MW of a transmission line. NYISO The New York Independent System Operator. PA , PB Nodal energy prices at nodes A and B. PL Price in the local region PJM The ISO now covers Michigan, Pennsylvania, Washington DC, Tennessee and more, with 44 million consumers and 135,000 MW of generating capacity. PR Price in the remote region Q The megawatt ow named in a CRR, or on a line. R The regulated price paid to a transco per MWh of delivered power. RTO Regional transmission organization. The new name for an ISO in the US. T The time period covered by a CRR, for example, peak hours during 2010. Transco A regulated monopoly transmission company. Wi The net power withdrawal at node i.
5.
1.
Patterns of transmission investments

Paul Joskow
INTRODUCTION
A transmission network with good performance attributes is essential to support well-functioning competitive wholesale and retail markets for electricity. The transmission network allows decentralized generators, marketers, distributors and large consumers to trade power in competitive markets. It can expand the geographical expanse of competition among power suppliers, giving consumers access to lower-cost energy and operating reserves. By expanding the geographic expanse of competition the transmission network can increase the eective number of competitors and reduce market power and thus prices. A well-functioning transmission network facilitates the entry of new generators to match demand and supply eciently at dierent network locations to achieve economic and reliability goals and supports the development of demand response options for wholesale and retail market participants. Electricity sector liberalization has not changed the physical constraints or physical laws that govern reliable transmission network operation or its role in supporting economical supplies of electricity. The network must still satisfy the same physical parameters and constraints (frequency, voltage, stability, coordination with interconnected networks) and provide for operating reserves to respond to uncertain realizations of demand and unplanned outages of equipment to maintain reliability and avoid major losses of load or a widespread network collapse. However, electricity sector liberalization has necessitated changes in the organization of the electric power sector and the tools available to operate the network economically and reliably and to stimulate investment in the network to reduce congestion and maintain the physical integrity of the network. Implementing eective transmission investment policies has proven to be especially challenging as countries liberalize their electricity markets. In the US, transmission congestion has increased and barriers to needed transmission investment are perceived to be a growing problem. Transmission line relief orders (TLRs) in the Eastern Interconnection have grown by a factor of ve since 1998. Congestion charges in the traditional PJM
131
132
(PennsylvaniaNew JerseyMaryland) area grew by a factor of 10 between 1998 and 2003. Congestion charges in the New York independent system operator (ISO) have more than doubled since 2001 (Joskow, 2005). Congestion has grown rapidly in Texas, Southern California, New York City, and New England as well. At the same time, investment in new transmission capacity has lagged the growth in electricity demand and the growth in new generating capacity (Hirst, 2004). In Europe, as wholesale power trading has grown, transmission congestion limits the geographic expanse of competition, limits opportunities fully to exploit generating capacity with the lowest operating costs, has led to concerns about generator market power within several countries (Newbery, 2004) and has created reliability challenges. As market liberalization proceeds, there has been very little investment in inter-transmission system operator (TSO) transmission capacity in Europe or the US. Intra-TSO congestion is a growing problem in some European countries as well. Policy makers in many countries with competitive power markets are increasingly concerned about reliability problems and reliability considerations, and associated engineering operating and planning criteria are playing an increasingly important role at the interface of wholesale market design, transmission pricing and transmission investment policies. In this chapter I discuss a number of issues associated with the creation of an institutional environment that supports the identication of and ecient investment in transmission infrastructure. I illustrate how the wholesale market and transmission investment frameworks have addressed these issues in England and Wales (E&W) since 1990 and in the PJM regional transmission organization (RTO) in the US since 2000. I am led to the following conclusions: 1. The simple models of transmission network congestion and investment that are used by economists have little to do with the way transmission investment is actually planned and developed, and the associated transmission services priced within the boundaries of individual TSOs today. Economic models and analysis need to be expanded to better capture the factors that TSOs and regulators consider when they identify transmission investment needs, especially as they relate to the implementation of reliability criteria used for network investment planning and system operations. The application of a set of complex electric power network models, engineering reliability criteria, and simulation studies using these models guide almost all intra-TSO transmission investment that is taking place around the world today. Commonly used economic models of transmission networks and transmission investment
2.
133
opportunities do not capture these reliability criteria or their application adequately, if they do so at all. 3. Policy makers in a number of countries have sought to distinguish between reliability and economic transmission investments. The former are conceptualized as being needed to meet engineering reliability criteria while the latter are conceptualized as being developed to reduce congestion costs (and losses). These two categories of investment are often treated as if they are distinct and independent. This is nonsense. Reliability-driven transmission investments are not independent of the variables thought to create the need for economic transmission investments. Reliability investments can have signicant eects on current and forecast locational marginal prices (LMPs) for energy and operating reserves, can have signicant eects on intraTSO congestion and losses, and can aect inter-TSO transmission capacity, congestion and losses as well. 4. Changes in network operating practices, TSO discretion in the procedures used to evaluate whether and when reliability criteria will be violated, and TSO discretion in the implementation of reliability criteria in actual operating practices can have signicant eects on locational prices for energy and operating reserves, congestion costs and rents, the cost of losses, and incentives to invest in transmission capacity to reduce congestion. System operators need discretion to operate transmission networks reliably. However, discretionary decisions aect the level and locational distribution of wholesale market prices and the associated incentives to invest in both generating and transmission capacity. 5. There are major asymmetries between the way intra- and inter-TSO transmission investment planning, evaluation and pricing are implemented. Dierences in inter- and intra-TSO transmission investment frameworks reect organizational and political boundaries, as well as the attributes of the legacy networks controlled by incumbent TSOs, rather than the physical attributes of the larger synchronized network, portions of which are controlled by individual TSOs. InterTSO investment opportunities can best be addressed through wider area planning using a common set of reliability criteria and evaluation principles and by integrating wholesale power markets and harmonizing the principles for setting transmission service prices across control areas to support either regulated or merchant transmission investment. 6. Horizontal integration of previously independent TSOs can have signicant eects on network operations, generator dispatch and LMPs for energy and operating reserves, congestion costs and
134
7.
8.
9.
10.
incentives to invest in transmission facilities to meet reliability and economic goals by internalizing inter-TSO ineciencies under a larger geographic TSO umbrella. Merchant transmission investment has and is likely to make a very small contribution in the overall portfolio of transmission investment projects that will be made in the future. The merchant model that seems to be evolving is one in which either regulated entities (and ultimately their customers) take on the risk of entering into long-term performance contracts with a developer of a high-voltage direct current (HVDC) transmission link to expand interconnection capacity between TSOs or in situations where there are very limited interconnections between TSOs with large sustained dierences in prices, where market participants are willing to enter into long-term transportation contracts in return for rm rights and where the interconnector is sized so as to have a small impact on the dierence in locational prices. In addition to the problems with relying primarily on a merchant transmission investment model discussed in Joskow and Tirole (2005b, forthcoming), the sensitivity of locational prices for energy and operating reserves and associated congestion rents and costs to regulated investments in reliability-driven transmission projects combined with discretionary changes in TSO implementation of operating reliability rules create signicant additional barriers to intra-TSO merchant transmission investment. The interconnection rules and associated cost responsibilities governing the interconnection of new generators and interconnections of new inter-TSO transmission links to a TSOs internal network have signicant eects on locational incentives faced by new generators and on both the economic attractiveness and the economic eciency of merchant transmission investment projects. Deep interconnection rules that require market participants to pay the relevant interconnections costs at dierent locations, and the associated allocation of cost responsibilities or interconnection service prices, provide superior locational incentives to shallow interconnection rules and interconnection prices that do not vary by location. Most transmission investment projects are being developed today and will be developed in the future by regulated entities. Accordingly, the creation of a sound, stable and credible regulatory framework to govern regulated transmission investments is very important. The absence of such a framework for the identication of transmission needs, for transmission cost recovery, for mechanisms to align investor incentives with public interest goals, and for ecient pricing
135
11.
12.
13.
14.
of the associated transmission service is a major barrier to the ecient mobilization of transmission investment. An attractive regulatory framework will accommodate but not rely on merchant transmission investment. There exists no single mechanical silver bullet incentive regulation mechanism that can be developed to govern transmission investment. A practical regulatory framework will inevitably include a mix of cost-of-service regulation with an overlay of performance-based regulation mechanisms based on benchmarking, prot sharing (sliding scale) and ratchets (see Chapter 4 by S. Stoft). The development and application of performance norms, formal investment criteria, as well as considerable regulatory judgment is an inevitable component of a sound regulatory process. One component of such a regulatory framework is a transparent regional transmission investment planning process with clear rules for achieving dened reliability and economic goals. The bifurcation of regulatory responsibilities in the US between the states and the federal government (Federal Energy Regulatory Commission: FERC) and in Europe between Brussels and the members countries creates signicant potential disincentives to transmission investment in what is only a partially liberalized sector. Full unbundling of transmission service and the transfer or harmonization of regulatory responsibility for all transmission service to federal authorities would be very desirable. In order to implement an eective regulatory process, regulators will need more information about the performance of the transmission network, will have to establish performance norms and criteria, and apply performance-based regulatory (PBR) systems that align TSO incentives with public interest performance goals. These incentive mechanisms must satisfy rm viability/participation constraints and reect rent extraction goals in the context of information asymmetries between the regulator and the rms it regulates. Chapter 4 discusses incentive regulation in more detail. TSOs that are also vertically integrated into generation and marketing activities create additional regulatory challenges because of the conicts of interest between operating and investment decisions made by the TSO and their impacts on the protability of generation and marketing businesses that make use of the same transmission network. Regulatory rules requiring functional separation eliminate any benets of vertical integration if they are followed while providing imperfect protection against abusive self-dealing behavior by the TSO. The creation of truly independent TSOs reduces the regulatory
136
15.
burdens and creates entities whose management is focused on the transmission business. Separating SO and TO functions may be a second-best response to vertical integration between transmission, generation and power marketing, but it also is likely to lead to some ineciencies.
2.
ATTRIBUTES OF TRANSMISSION INVESTMENTS
Transmission investment policies must respond to a number of interdependent questions. What are the societal goals that a transmission investment framework should seek to achieve? What are the respective roles of economic goals, reliability goals and other potential public policy goals? What are the physical and economic attributes of dierent types of transmission investments? How are transmission investment needs identied? What entities are expected to develop the new facilities? How are the associated costs expected to be recovered through transmission charges or price arbitrage prots resulting from transmitting power from a relatively high wholesale price location to a lower wholesale price location? Which entities that make use of the network should pay for its various components? Where does transmission end and distribution begin? While policy makers talk about transmission investment in general, in reality those responsible for identifying investment needs and opportunities typically divide transmission investment into a number of dierent categories. If we are going to make progress in understanding the transmission investment problem from a theoretical and empirical perspective, we need to better coordinate economic analysis with the conceptual framework that governs the consideration of transmission investment by system operators, transmission owners and policy makers. Let me note as well that there is no uniform denition of the facilities that make up the high-voltage transmission network that is subject to the control of the system operator. In E&W, the transmission network license includes only facilities with voltages of 275 kV and 400 kV. In the US, transmission facilities typically, though not always, include lines that operate at 66 kV and above with various exceptions based on dierences in network topology, legacy ownership and regulatory arrangements. In France, Rseau de Transport de lElectricit (RTE) transmission network includes facilities with voltages similar to those in the US.1 Dierent rules may be applied for dening which side of the fence interconnection facilities lie. I shall ignore these dierences in the denition of the facilities that comprise the transmission network in dierent countries in what follows.
137
However, we should keep in mind that these dierences complicate comparisons of transmission network performance, since performance indicia like congestion costs, losses, network component availability, unserved energy (loss of load), operation and maintenance costs and so on, will depend on which network components are included in the denition of transmission and which are not. Categorization of Transmission Investments Dierent TSOs also categorize (and characterize) transmission investments in a variety of dierent ways, use a wide range of very dierent methods to assess charges to cover the capital and operating costs of transmission facilities, to cover the costs of congestion and losses, and to assign responsibility for payments to cover the cost of investments in new transmission network facilities. In the discussion that follows, I shall make use of the following categorizations which are broadly consistent with those used in the transmission planning and investment frameworks in the US and the UK. Generator interconnection investments When new generators are constructed they must have interconnections to the transmission network in order to sell energy and ancillary network support services in the wholesale market. Some minimal level of investment is required merely to connect the generator to the closest point of interconnection to the network and to allow the generator to deliver its maximum generating capacity to the network at this point of interconnection. At a minimum, these investments will include new (or reinforced) transmission lines between the generating plants switchyard and the rst point of interconnection to the high-voltage network and investments in transformer capacity at the point of interconnection to the network to accommodate the reliable injection of additional power into the network at the proper voltage. The investments required will vary directly with the generators maximum capacity, the maximum capacity of proximate generating facilities that share an interconnection point on the network, the voltage at which the power is delivered to the network, and the reliability of the interconnection facilities as measured by their planned (for maintenance) and unplanned outage rates under dierent system conditions. Interconnection investments alone do not assure the associated generator that there will be adequate transmission capacity to transmit the power from the point of interconnection to the network on to other locations on the network without curtailments or additional charges due to congestion. As a result, as I shall discuss in more detail presently, if the
138
generator expects to be able to utilize fully its generating capacity to deliver power to serve demand nodes dispersed around the network without experiencing curtailments or incurring congestion charges, investments deeper into the network are likely to be required. Alternatively, the generator or its customers will have to secure transmission rights to utilize the scarce transmission capacity that already exists from others. Distribution network and large retail customer interconnection investments Distribution networks and large customers who take power directly from the transmission network must also have transmission facilities that interconnect them to the high-voltage transmission network. These interconnection investments are the ipside of generator interconnection investments except that distribution networks typically have multiple points of interconnection with the transmission network and individual loads locational decisions will, in most cases, be insensitive to interconnection costs. At a minimum, these investments will include new (or reinforced) transmission lines between the distribution networks facilities and the rst point of interconnection to the high-voltage network and investments in transformer capacity at these points of interconnection. The investments required will vary directly with the distribution networks maximum coincident demand, the number and attributes of interconnection points, the voltage at which the power is delivered to the distribution network before being further stepped down by the distributor, and the reliability of the interconnection facilities as measured by their planned (for maintenance) and unplanned outage rates under dierent system conditions. Interconnection investments per se do not assure the distributor that there will be adequate upstream transmission capacity to transmit all of the power it needs to meet its end-use customers demand because there may be congestion between the distributors point of interconnection and generation nodes on the network under some operating conditions. However, a distribution company will not add interconnection capacity unless it can ll that capacity with energy drawn from the transmission network by securing, in one way or another, the network capacity deeper into the network needed to gain access to enough energy to meet the demand of its distribution service customers. Of course, the distributor may also be able to balance supply and demand with generation embedded in the distribution network (distributed generation) and with load reduction programs, including the impacts on consumer demand of real-time pricing or priority rationing contracts (Chao and Wilson, 1987). As already noted, there is no wellaccepted rm line between distribution and transmission.
139
Intra-TSO economic transmission network upgrade investments By intra-TSO, I mean investments made within the footprint of a specic TSO. The TSO may cover only a portion of a larger synchronized AC network as in the US and Europe. Economic models of transmission network operations and investment focus on the eects of transmission capacity (whether in the context of a simple two-node network or a multinode network with loop ow) on congestion costs and congestion rents. Congestion costs, congestion rents, dierences in locational prices caused by congestion and the prices of congestion are discussed in Chapter 4. Economic models of transmission expansion should, in principle, also include the cost of losses (Joskow and Schmalensee, 1983, pp. 367) in both locational prices and investment planning. And loss cost considerations play a signicant role in traditional engineeringeconomic system planning models.2 However, perhaps for convenience, many contemporary economic models have ignored losses, although in the wholesale markets operation in New York and New England, LMPs reect both the marginal costs of congestion and the marginal cost of losses. Indeed, marginal losses lead to signicant dierences in LMPs in these markets even when there is no congestion. In what follows, when I refer to congestion costs I am using the term to encompass the cost of losses as well. So-called economic transmission investments (whether intra- or interTSO) are motivated by the opportunity for such investments to reduce the social costs of congestion. Optimal economic investment involves a tradeo between investing in additional transmission and the associated reduction in congestion (and loss) costs. That is, the incremental cost of transmission investment should be compared to the incremental reduction in the cost of congestion (sometimes referred to as the redispatch cost) on the network (Joskow and Tirole, 2005b). In the absence of lumpy investments,3 and assuming that all nodal prices reect the relevant marginal social opportunity cost at each node, it is optimal to make expenditures on economic transmission capacity up to the point where the marginal cost of transmission investment is equal to the (expected) reduction in transmission congestion and loss costs. Since transmission investment is an expenditure today that creates a long-lived asset and congestion is a ow that depends on future supply and demand conditions in both the electricity and input markets (for example, fuel prices), the benets of an economic transmission investment are necessarily uncertain at the time they are made and are realized over a period of many future years. Inter-TSO economic investments (interconnectors between TSOs) By inter-TSO, I am referring to investments that are designed to increase transfer capacity between two (or more) TSOs and to reduce congestion
140
between them. When TSOs operate portions (control areas) of the same synchronized AC network, the dierences between intra- and inter-TSO economic transmission upgrades are primarily institutional, reecting historical ownership structures, political boundaries and dierences in wholesale market design and regulatory mechanisms. The underlying physical attributes of investments at dierent locations on the larger AC network controlled by multiple TSOs are basically the same as would be the case if there were a single TSO for the entire network. That is, with a single TSO, inter-TSO investments would by denition become intra-TSO transmission investments governed by the same market, regulatory and transmission investment frameworks. However, dierences in the market designs and transmission investment frameworks of the multiple TSOs controlling portions of the same synchronized network, incompatibilities between the institutions governing interconnected TSOs, and various transaction costs resulting from horizontal separation that aect wholesale market prices and congestion on both networks, are likely to aect transmission investment decisions. It is frequently the case that intra- and inter-TSO transmission investments are treated even conceptualized very dierently due to these institutional dierences rather than to basic physical and economic realities. Dierences in market design and coordination between interconnected TSOs on the same synchronized AC network can aect the economic attributes and evaluation of opportunities to expand transmission capacity to reduce congestion both between the TSOs networks and even within their individual networks. This is the case, in part, because dierences in market design and network operating practices can aect locational prices and dispatch decisions within both of the individual TSOs control areas. These eects are exacerbated when multiple TSOs adopt operating protocols that are based on ctional physical characterizations of the interconnected free-owing AC network for example, that a large synchronized AC network is really several separate networks connected by radial lines with no loop ow and no congestion within each TSO. Individual TSOs rst tend to resolve congestion inside their networks and then facilitate residual economic trades between networks. These policies tend to push congestion out to the borders between TSOs and reduce economic eciency. While this is not a necessary result of having multiple TSOs on the same free-owing network, it appears to be a practical reality of the decentralized operating protocols adopted by individual TSOs. A good example of this is the experience following the integration of PJM West (Allegheny Power Systems) with the incumbent PJM system to the East in April 2002 when signicant changes in congestion patterns and locational prices occurred.4
141
This suggests that the horizontal consolidation of TSOs into a single TSO covering the larger geographic footprint of the real physical network could lead to very dierent evaluations of and incentives for economic transmission investments. By internalizing wholesale market and transmission institutions under a single TSO, both the locational price and congestion patterns that drive economic transmission investments are likely to change. Transmission upgrade evaluation policies as they relate to interTSO transmission investments are likely to change as well. As we shall see, inter-TSO economic network upgrade opportunities and intra-TSO transmission network opportunities may be evaluated very dierently by TSOs on the same AC network. The internalization of transmission investment decisions and the integration of wholesale market institutions are two of the primary motivations in the US for FERCs eorts to create large RTOs that consolidate the multiple control areas that now exist. Consolidating previously separate control areas is expected to transform inter-TSO economic transmission investment opportunities into intra-TSO transmission investment opportunities governed by a single transmission investment framework, a common wholesale market design, and wider market area with a set of fully coordinated locational prices. There are, of course, situations where inter-TSO economic transmission investments involve the creation or expansion of interconnections between truly separate AC networks. For example, by building HVDC interconnectors between two separate networks, opportunities to increase trades of power from high- to low-price areas can be exploited. The HVDC link between England and France, the HVDC links between Quebec and New England, and the HVDC link being constructed between Tasmania and Victoria, Australia are examples.5 Interconnection investments to support inter-TSO transmission links Building or expanding an inter-TSO transmission facility (an interconnector in European parlance) is only the rst step in increasing trade between two TSOs whether they are on the same synchronized network or govern independent networks. The new interconnector will withdraw power under the control of one TSO and deliver it to the network controlled by the other. Facilities need to be constructed to aect the interconnection with each network, just as would be the case for a generator with equivalent capacity located at the point of interconnection at the delivery end or a large load located at the point of interconnection at the withdrawal end of the new inter-TSO link. Moreover, just as in the case of new generators, whether or not the interconnector capacity can be fully utilized to deliver power to serve load depends on network congestion beyond the point of interconnection to each network and how scarce
142
transmission capacity on the rest of each network is allocated. Interconnectors may also have reliability implications, especially when they are relatively large and become binding contingencies that aect the evaluation of whether the network is meeting established reliability criteria. However, unlike a generator seeking to locate on a single network, a proper evaluation of the value of and incentives to invest in an interconnector justied by the cost reductions realized by expanding use of low-cost power to displace the use of higher-cost power (plus the change in total net surplus resulting from lower prices and increased demand on the importing network) will depend as well on the compatibility of the interconnection investment policies and the deeper network upgrade policies on both networks. Some TSOs socialize the costs of these deeper network upgrades into a general postage stamp transmission service tari rather than requiring generators or interconnectors, causing the need for additional deep network investments to pay for them. This is called a shallow interconnection pricing policy. In other TSOs, the costs of deeper network investments required to restore reliability parameters and/or relieve congestion are charged to generators and interconnectors at the locations where power ows cause the need for these deeper network investments. This is called a deep interconnection policy. As discussed further below, PJM has a de facto deep interconnection policy while most other TSOs in the US have shallow interconnection policies. In E&W, the use of system charges vary by location and are, eectively, a deep interconnection pricing policy. Reliability transmission network investments These are transmission investments that must be made to restore exogenously specied TSO planning reliability criteria that may be violated as a consequence of changes in demand patterns, generation investment and generation retirements. (Planning and operating reliability criteria are not typically the same.) TSO reliability criteria have generally been carried over from the old regime of regulated vertically integrated monopolies. As I shall illustrate with several examples below, virtually all of the transmission investment underway today in the US and, eectively, in E&W are either direct interconnection investments as discussed above or some type of reliability investment. I am informed that this is the case in many other countries as well. Ones rst reaction might be that this is a terrible situation. It suggests that current transmission investment frameworks consider only reliability and ignore the economic costs of congestion! However, while reliability and economic transmission investments are often treated as if they were distinct and independent types of transmission investments, this is a complete ction. Investments made to restore engineering reliability criteria can have very signicant impacts on congestion and locational prices
143
and, accordingly, on the valuation of and incentives for economic transmission investments. Similarly, economic investments can have impacts on reliability parameters. Neither reliability transmission investments nor the interrelationship between reliability criteria and economic parameters are given much attention in the literature on competitive electricity markets or transmission investment. Yet so-called reliability investments are playing an increasing role in the overall intra-TSO investment prole and exacerbate incompatibilities between inter- and intra-TSO transmission investment. The engineers and the economists interested in transmission investment issues clearly need to be introduced to each other. These issues will be discussed in more detail after the case studies of E&W and PJM are presented in Section 7, below. Physical Attributes of Transmission Network Components The standard metaphor for transmission investment is the construction of a major new transmission line on new rights of way. While major new transmission lines can cost hundreds of millions of dollars, many socially desirable projects are relatively inexpensive and do not require expanding the geographic footprint of the network. These latter investment opportunities are especially important in a world where the construction of major new lines is constrained by Nimby (not in my backyard) constraints. The distribution of project costs for transmission investment projects identied in a recent New England ISO transmission plan is indicative of the patterns of transmission investment opportunities. The 2004 transmission plan includes 245 projects with a total expected cost of $2.1 billion. The ve most expensive projects are projected to cost $1.4 billion and the remaining 240 projects a total of about $700 million (ISO New England, 2004). The full distribution of reliability project costs in the New England transmission expansion plan is displayed in Table 5.1. Of the roughly 50 transmission Table 5.1 Reliability upgrade projects: New England regional expansion plan 2004 ($ million)
Projects Top 5 Next 5 Next 15 Remaining 220
Source: ISO New England, (2004).
Total cost 1,388 322 296 132
Average cost 277.6 64.4 19.7 0.6
144
projects listed in PJMs economic transmission investment market window in November 1994 (discussed in more detail below) estimated investment costs vary from $20,000 to $39 million. These investments all seem to be lumpy in the sense that they mitigate the congestion identied completely and could not be nanced out of the residual congestion rents. Projects to enhance transmission networks include a wide range of physical components that are to be added to the network or to replace components that are already in the network. They include:

new relays and switches; new remote monitoring and control equipment; transformer upgrades; substation facilities; capacitor additions; reconductoring of existing links;6 increasing the voltage of specic sets of transmission links; new transmission lines on existing corridors; and new transmission lines on new corridors (above or underground).
In addition, the eective capacity of the network may be increased at little or no cost with the adoption of better remedial action schemes or special control schemes that increase the speed with which other transmission links or generating plants can respond to unplanned equipment outages. Changes in operating practices and the way contingencies are evaluated and handled when they occur can also magically increase (or decrease) eective transmission capacity. The diversity of network components that can be added to or substituted for existing network components reects in part the factors that limit transmission capacity. On most networks, transmission limitations are driven by reliability criteria and associated assessments of the ability of the network to physically balance supply and demand without shedding load involuntarily or violating network voltage, frequency or stability criteria that would increase the probability of a network collapse. These reliability criteria typically reect the objective of keeping the probability of involuntary load shedding to a very low level and the probability of a widespread network collapse to zero. The limitations on utilization of the network are frequently one or more sets of contingency constraints evaluated under a variety of system conditions (study conditions) that maintain the probabilities of load shedding or network collapse to acceptable levels rather than binding pre-contingency thermal limits on particular lines. The binding constraint limiting transmission capacity could be the reliability of a breaker, the speed with which a switch can be pulled, or the ability accu-
145
rately to monitor line sag in real time. Better or faster communications between system operators controlling portions of the same synchronized AC network can also relax contingency constraints and increase the eective capacity of the network. Accordingly, when we think about expanding transmission capacity we should have in mind the full range of physical and behavioral options as well as the importance of engineering reliability criteria and associated contingency studies and constraints. Note that the discussion in this subsection also implies that measuring transmission capacity, or changes in transmission capacity, using measures of the length of transmission lines for example, MW miles is not appropriate. Especially in light of the diculties of siting major new transmission lines, increases in transmission capacity are likely to focus on deepening the existing transmission infrastructure and minimizing the expansion of its geographic footprint. When new lines are necessary, siting diculties will also lead to more underground links and the use of more costly routes to avoid environmentally and politically sensitive areas. Legacy Infrastructure Considerations It is important to recognize that electricity sector liberalization reforms take place with an existing infrastructure composed of long-lived assets with particular attributes. The attributes of the legacy infrastructure reect historical institutional arrangements, corporate boundaries, political boundaries, historical patterns of urban and industrial development, and historical economic and technological opportunities. The attributes of this legacy infrastructure will aect the behavior and performance of the system for many years into the future. We can change the institutions but we cannot erase the existing infrastructure in place at the time sector liberalization reforms are implemented but only change it gradually over time. For example, in the US the electric power sector evolved with a large number of vertically integrated utilities serving geographic areas that varied widely in size. This structure was signicantly inuenced by federal and state laws passed during the 1930s that sharply restricted mergers of proximate utilities, especially when they served more than one state. Infrastructure development focused most intensively on the geographic areas served by individual utilities with transmission networks developed to link generators owned by the utility with the load centers within the utilitys geographic franchise area. The strengthening of the transmission infrastructure connecting vertically integrated utility control areas proceeded later and more slowly. In many cases it was motivated primarily by reliability considerations rather than with the goal of importing large amounts of power from neighboring vertically integrated utilities (Joskow
146
and Schmalensee, 1983). So, for example, New England has only limited transmission interconnections with New York State (about 1,500 MW connecting two networks with peak loads of about 29,000 MW and 35,000 MW, respectively). This reects much more the ownership structure of utilities in this area of the US during the last half of the twentieth century (there was no common ownership between utilities serving areas in both New York and New England while some utilities in New England had operating companies and generating facilities in two or more New England states) and historical political boundaries (the New England states joined together to form the New England Power Pool in 1969 while New York created its own power pool at about the same time) than it does any natural economic and technological attributes. Similarly, large integrated utility holding companies like AEP and Southern developed strong transmission networks covering several states in which they had operating companies while small independent vertically integrated utilities in other areas of the country have weak interconnections with their neighboring utilities and, as a result, enter the liberalization with weak regional networks.7 In Europe, where several countries relied on one or a small number of vertically integrated utilities, or as in Spain, consolidated responsibility for a shared high-voltage transmission network, there tend to be much stronger intra-country than inter-country transmission networks. This has led European transmission policy to focus on expanding interconnectors between countries rather than on intra-country wholesale market design, locational pricing and transmission policies, sometimes using the argument (almost certainly wrong) that the national networks are so strong that there is no internal congestion. In Italy, for example, there are several congested interfaces, in addition to the congested transmission interfaces with France, Switzerland, Austria and Slovenia. Clearly the attributes of the legacy infrastructure are likely to have signicant implications for the need for additional transmission investment to support competitive wholesale power markets. Dimensions of Transmission Network Performance While I am focusing here on transmission investment, transmission networks have multiple and interrelated performance dimensions. The design of supporting organizational, regulatory and market institutions and judgments about the overall performance of the transmission network should take them all into account. These performance attributes include:

Costs of congestion, losses and ancillary network support services. Network operating and maintenance costs.

147
Availability of network components and eciency of outage restoration in response to congestion and loss costs. Reliability of the network involuntary losses of load and network collapse. Costs of market power and other market ineciencies aected by the operation of and investment in the network. Eciency with which the investment framework mobilizes investment to expand the intra-SO network to meet reliability and economic goals. Eciency with which the investment framework mobilizes capital to expand inter-TSO transmission capacity to meet reliability and economic goals. Eciency with which innovations in software and hardware technologies are adopted for improving all aspects of network performance.
3.
TRANSMISSION NETWORK ORGANIZATION
Transmission network organizations have both vertical and horizontal dimensions and we see a variety of dierent vertical and horizontal structures across countries. These organizational dierences are likely to aect the incentives to make transmission investments as well as how transmission opportunities are evaluated. These include:
Full vertical integration This model is characterized by vertical integration between system operation, transmission ownership and maintenance, generation, retail and wholesale marketing. In these situations, regulations governing non-discriminatory access to the network, non-discriminatory transmission pricing, and non-discriminatory evaluation of and investment in transmission facilities are extremely important but very dicult to implement satisfactorily. The fully integrated TSO has an inherent conict of interest because its transmission network operating, maintenance and investment decisions aect the value of its generation portfolios and marketing businesses. Moreover, in such companies, the transmission business is likely to represent a small fraction of the income of the enterprise as a whole, and, as a result, transmission is less likely to be the primary focus of management attention. Independent transco This model is characterized by the separation of transmission network functions (SO and TO functions) from generation and power marketing functions. This is the independent transco
148
model that has been adopted in England and Wales, Spain, New Zealand, Italy (soon) and France (if we ignore the Electricit de France (EdF) holding company aliation). System operation, network maintenance and network investment are vertically integrated and can be managed in a coordinated manner by the transco. The conict of interest inherent in an organization where the TSO is not independent of market participants no longer exists and the rms management is now focused on the provision of transmission services. ISO This model is characterized by the separation of system operations from transmission facility ownership, investment and maintenance, as well as from ownership of generation and marketing businesses. The independant system operator (ISO) does not own or maintain transmission assets, but is responsible for scheduling and dispatching generation and load in coordination with operating reliability criteria and market rules, managing and enforcing procedures and rules for allocating scarce transmission capacity, interconnection arrangements, administering taris governing transmission service prices, and working with TOs and other stakeholders on the coordination of maintenance schedules and planning for new transmission investments to support changes in the demand for and supply of generation services. This is the model that has been or is being adopted in large portions of the US, Alberta, Argentina, Norway and other countries.
There are several rationales for creating a separate independent system operator rather than an independent transco. It may not be politically feasible to force the separation of transmission ownership from generation ownership and marketing activities. An ISO is created to sit on top of the vertically integrated utilities to provide an independent network manager and tari administrator to govern relationships between market participants and the vertically integrated owners of the transmission networks facilities. There may be geographically balkanized ownership of transmission assets (either regulated or unregulated) and the horizontal integration of transmission assets is deemed to be politically infeasible or undesirable, especially if merchant investment is expected to play an important role in the system. The ISO can then manage a larger physical network with multiple transmission owners more eciently than would be the case if each TSO operated its own control area. Finally, it is sometimes argued that generation and transmission compete (that is, they are horizontally as well as vertically related) with each other, that even a transmission owner with no generating assets cannot be truly independent and will have incentives to discriminate against generators on the network. In this case, an ISO that has no direct interest in the nancial performance of the owners of any of the
149
assets that comprise or utilize the transmission network will be unbiased. This naturally leads to the question of what the ISOs objectives are and what incentives inuence the monopoly ISOs behavior and performance. Other things equal I would expect dierent organizational arrangements to have dierent performance attributes and to create dierent regulatory challenges. I oer the following hypotheses:
Vertical integration among transmission, generation and marketing creates signicant regulatory challenges to mitigate incentives to disadvantage generation and marketing rivals. Moreover, since the regulatory response to vertical integration is typically to require functional separation of the SO/TO functions from generation and marketing and to apply regulations that are designed to force the rm to operate as if its SO/TO functions are not aliated with generation and marketing businesses, there are no remaining social benets to vertical integration between SO/TO functions and generation, marketing and other unregulated lines of business that make use of the aliated transmission network. What is the point of continuing common ownership of entities that regulators are trying to ensure behave completely independently? Vertical separation of system operations from ownership and maintenance of transmission facilities is likely to make coordination between system operations, network maintenance and outage restoration, and investment more costly than if the TO/SO functions were combined. Moreover, to the extent that transmission owners also own generation and are engaged in power-marketing activities, it will be dicult for the SO and the regulator to ensure that TO behavior, especially related to maintenance, interconnection investment and investment to reduce congestion, will not be aected by their impacts on aliated generation and marketing companies. Both the SO and the regulator will have imperfect information about the TOs cost opportunities, eorts and incentives. For example, one of the easiest things to accomplish is to fail to get a permit to build a new transmission link that will reduce congestion into an area where an aliate owns generating capacity. Indeed, I suspect that PJMs hostility toward regulated economic transmission investment (more below) is not unrelated to the fact that all of the transmission owners under the PJM umbrella are also vertically integrated into generation and marketing. Limited horizontal expanse of SO functions is likely to create ineciencies. The more control area operators there are on the network, the more conservative will reliability criteria be, reducing the availability of inter-TSO transmission capacity, and the more
150
dicult it will be for separate market areas eciently to coordinate wholesale trading of power and the allocation of scarce transmission capacity. As I shall discuss presently, there are signicant asymmetries between the framework governing intra-TSO transmission investment and inter-TSO investment. Internalizing inter-TSO links through horizontal integration is likely to lead to less congestion and more transmission investment. These are, of course, only hypotheses that should be veried through empirical analysis.
4.
PRINCIPLES TO GUIDE TRANSMISSION INVESTMENT REGULATORY FRAMEWORKS
A sound transmission investment regulatory framework must address several interrelated issues. The following discussion reects my view that the bulk of intra-TSO transmission investment will be mediated through a regulatory process of some type and that so-called merchant investment will play a limited role.8 Merchant investment of one type or another may play a larger role in mobilizing investment for expansion of inter-TSO transmission facilities (interconnectors) as a result of various institutional and political constraints. Merchant opportunities may emerge as well if incumbent TOs are permitted to develop unregulated merchant projects on their own networks, exploiting the market power that they possess. I shall also assume that all of the TSOs revenues come from entities that use the network; there are no government subsidies, and a viable TSO, SO or TO must balance its budget. 1. Objectives and performance norms The regulatory framework must specify clearly what the regulators objectives are for the TSO (or SO and TOs if they are separated) that is, what the TSO is expected to accomplish how the TSOs performance will be measured, what norms and benchmarks will be applied to evaluate its performance, and what instruments the TSO may use to achieve these performance objectives. In the case of an organizational structure that separates SO and TO functions, the division of responsibilities and mechanisms for coordinating relationships between the SO and the TOs under it must be clearly dened. As I shall illustrate presently, integrating so-called reliability goals and criteria with economic goals and performance norms is especially important. TSO participation or viability constraints The regulatory framework must recognize that there is a rm viability or participation constraint
2.
151
3.
4.
that any regulatory mechanism must adhere to (Tirole and Laont, 1993). This budget balance constraint can be dened simply as the requirement that any acceptable regulatory mechanism must have the property that expected revenues from the provision of transmission services must at least cover the costs that the regulated rm incurs to provide these services. Private rms cannot be expected to oer to supply services if they do not expected to be compensated for the associated costs. State-owned rms cannot satisfy hard budget constraints (no government subsidies) unless they can recover the cost of providing transmission services from transmission service revenues. If transmission service costs have non-convexities (for example, scale economies), actual prices for transmission service must depart from ecient prices. We are in the world of second best. Rent extraction goals The ip side of the rm viability or participation constraint is the impact of higher prices on consumers. The higher are the prices charged by the regulated rm the lower is the surplus left to consumers and, where prices exceed their ecient levels, the lower is aggregate welfare. In a world with asymmetric information, where the regulator has less information than does the regulated rm about its costs, it is well known that there is a trade-o between providing the rm with incentives to supply eciently (cost and quality dimensions) and rents left to the regulated rm from charges to consumers that exceed the rms costs of production (Laont and Tirole, 1993). Over time, we would like to see the benets of lower costs owing through to consumers as lower prices. Incentive alignment Regulators have imperfect information about a regulated rms cost opportunities, service quality, managerial eort, consumer demand and other factors that inuence the cost and quality of services provided by the regulated rm. Regulatory mechanisms should be designed to reect the asymmetry of information available to the regulated rm and the regulator while making ecient use of the information that is available to the regulator. The goal of eective regulatory mechanisms is to align the incentives faced by the regulated rm with the performance goals established by the regulator. This can be accomplished by (partially) tying the regulated rms prots to its ability to meet or beat performance goals established by the regulator. The power of such incentive schemes is necessarily limited by the information that the regulator has about the basic cost and demand conditions faced by the regulated rm as well as by rm viability constraints and rent extraction goals. Under all realistic situations, the second-best regulatory mechanism will partially tie a regulated rms revenues to the actual costs that it incurs and partially
152
5.
6.
place the regulated rms prots at risk for variations in performance. This can be accomplished with a sliding-scale mechanism (protsharing formula), and/or with periodic ratchet mechanisms that realign the rms revenues with its costs from time to time (see Chapter 2 in this book). Incomplete contract considerations Regulatory frameworks can be viewed from a contractual perspective in which regulatory rules dene the terms and conditions of an incomplete contract between the regulator and the regulated rm. The regulatory contract also denes a renegotiation framework that allows the terms and conditions of this contract to be adjusted over time as supply and demand conditions change (Joskow and Schmalensee, 1986). Investments in transmission facilities are long-lived assets that provide services for many years into the future. While the costs of investments are incurred up front, the revenues that the rm will receive from these assets will be realized from transmission service revenues extending over the life of the asset. On the one hand, once the investment is made, the regulated rm must be concerned that it may be subject to a regulatory hold-up aimed at conscating the ex post quasi-rents created by the investments. Investors in regulated assets will seek a credible commitment that such hold-ups will not occur. A credible full-contingent claim contract negotiated ex ante would be ideal from this perspective. On the other hand, the regulator is not in a position to dene an ecient fullcontingent claim contract ex ante that also satises a budget balance constraint. Over the life of regulated transmission investments supply and demand conditions are likely to change considerably, aecting both the protability of the regulated rms investment and the rents extracted from consumers. Moreover, the regulator will learn more about the attributes of the regulated rm, its costs, revenues and the quality of service over time as well. An eective regulatory process is like a good incomplete contract (Joskow, 1988). It denes initial terms and conditions, performance norms, formula adjustments to reect changing economic conditions, and an adjustment process that provides an ecient framework for adjusting these terms and conditions when they fall outside of a self-enforcing range. Transmission service price structures It is convenient to think about the components of the regulatory framework above as establishing the aggregate revenues (or prots) that the regulated rm can earn under various contingencies. These allowed revenues reect rm viability, rent extraction and incentive alignment considerations or, to oversimplify, the regulated TSOs current budget constraint is determined rst. Prices must then be established for the various services
153
7.
8.
that the TSO provides. These prices should provide ecient signals to transmission system users so that their behavior can adjust to reect the (marginal) costs of the services provided to them in the short and long runs. They must also be set at levels that produce the aggregate revenues (or prots) that the regulated rm is allowed to earn based on the terms and conditions of the regulatory arrangements discussed above. Other terms and conditions of network access In addition to the specication of the prices for using the transmission network, other terms and conditions of service must also be dened. This is especially important when the TSO or the TO is not independent of market participants. These terms and conditions include the rules governing the process through which interconnection requests by generators or merchant transmission projects will be processed, specication of cost responsibility for interconnection and network reinforcements, the application of reliability criteria to evaluate the availability and cost of providing transmission service, the specication and allocation of physical or nancial transmission rights, and the mechanisms for allocating scarce transmission capacity in the short and long runs. Some of these terms and conditions are ultimately linked to the attributes of the wholesale markets that are supported by the transmission network. Relationships between transmission and wholesale market institutions In the early years of electricity sector liberalization in the United States, Europe, Japan, Australia and other countries it was often argued by policy makers that there was a natural and fairly simple separation between competitive power markets and the transmission network that is necessary to support these markets. It is quite clear today that no such simple separation exists. Organizing power markets and transmission institutions as if a clear separation exists inevitably leads to serious problems. Ecient power markets, ecient transmission operation and investment behavior, and the satisfaction of reliability goals at the lowest reasonable cost are all fundamentally interdependent. Competitive market prices for power (spot and forward) are signals of the value of both energy and transmission capacity at dierent locations. These price signals can be used to allocate scarce (congested) transmission capacity to highest-valued (lowest-cost) users, can enable consumers to express their willingness to pay for reliability and express their risk preferences regarding price volatility, can allow generators to factor locational and time series dierences in power prices into operation and investment decisions, can allow transmission networks to incorporate the costs of
154
9.
congestion, the value of reliability and other factors into maintenance and investment decisions and so on. However, the social value of these price signals and the costs and benets of agents responding to them are only as good as the eciency of the markets that produce them. Moreover, some of the attributes of electric power networks for example, the possibility of network collapses can make investments in reliability a public good (Joskow and Tirole, 2006). Other market imperfections (for example, generator market power, lumpiness in investments, imperfectly dened property rights) and regulatory interventions (for example, price caps, SO procurement behavior, non-price rationing ibid.) aect the prices for generation and the value of scarce transmission capacity in both the short and the long runs and can distort rather than improve transmission operating and investments decisions. Accordingly, a well-functioning transmission network depends on the design and implementation of sound wholesale market institutions as well as a sound regulatory framework (economic, reliability, network planning) for transmission network owners. Transmission planning Transmission networks do not and will not evolve through the workings of the invisible hand of competitive markets. Even if one were to believe that all transmission investments should be market driven and developed by merchant investors, the impacts of proposals for new transmission links on the network must, at the very least, be evaluated by the SO to dene the attributes of the incremental network capacity that a merchant project creates and the combinations of any incremental transmission rights that are consistent with the changes in the feasible set of power ows anticipated to be created by the investment, whether the operation of the new facilities would lead to conicts with existing transmission rights, and the specic allocation of transmission rights that will be conveyed to the transmission developer (Joskow and Tirole, 2005b). As I shall discuss presently, in the real world, entry (and exit) of generating plants and changes in demand patterns aect both network congestion as reected in simple economic models of transmission networks (Joskow and Tirole, 2000) as well as reliability constraints as dened by system planners and operators. Investment opportunities driven by economic criteria and investment needs driven by reliability criteria are highly interdependent. At least in the current state of play, a transmission planning process is required to evaluate some aspects of regulated reliability-driven transmission investments, regulated congestion cost-driven transmission investments, merchant transmission investments and generator interconnection investments.
155
10.
Transmission planning processes should be transparent, provide for stakeholder input, and reect the objectives and norms dened by regulators for the transmission network. Merchant transmission investment The regulatory framework, including the transmission planning process, should accommodate proposals for merchant transmission investments. Merchant transmission investment was initially conceived as unregulated transmission investment projects that would be developed on an entrepreneurial basis in response to congestion (dierences in locational prices) between points on the same network or to dierences in electricity prices on dierent networks that the merchant project connects. Basically, merchant investors would recover their costs by buying power at one end of a link where it is cheap and reselling it at the other end where it is expensive; or selling the rights to use the merchant link to third parties to engage in this type of trading behavior. That is, the merchant investor makes money by arbitraging price dierences between the locations to which the merchant investment creates new transmission rights to buy and sell wholesale power. The volume of talk about merchant investment far exceeds the investment activity of merchant investors, despite the fact that the transmission frameworks in Australia, New England, New York, and PJM were designed to accommodate market-driven investments for economic transmission investment opportunities. Two small merchant links have been developed in Australia which intended to earn revenues and prots by arbitraging spot-price dierences between the networks in adjacent states. Both projects have now applied for regulated transmission status. A merchant underwater HVDC transmission link is being constructed between Tasmania and Victoria. However, the project is being developed in response to a request for proposals (RFP) from the state-owned electric power company in Tasmania and will be supported by a long-term contract (whose costs can be recovered from the sponsors customers) between the sponsor and the developer. A competitive RFP process initiated by the municipally-owned Long Island Power Authority (LIPA) supported by a 20-year long-term contract governs the completed Long Island Sound HVDC link between Connecticut and Long Island, New York. A similar 20-year contractual arrangement is supporting a proposed HVDC project between PJM and Long Island. HVDC projects linking PJM with New York City and between Upstate New York and New York City (recently cancelled due to the failure to obtain nancing) have been discussed for several years. Thats about it.
156
The merchant model that seems to be evolving is one in which regulated entities (and ultimately their customers) take on the risk of entering into a long-term performance contract with an HVDC transmission link developer to expand interconnection capacity between networks with no or limited interconnections and large sustained dierences in prices that are not aected signicantly by the addition of the link.9 Perhaps a better term for this model is private initiative transmission investments. It should be recognized as well, that the nancing costs for a merchant project are signicantly higher than those for an equivalent regulated project. A recent analysis of the nancing costs for a $100 million merchant transmission project10 indicated that the cash ow required to nance a regulated project developed by a utility and subject to traditional cost of service regulation would be $9.4 million per year. The annual cash ow for the same merchant project with a long-term contract (taking on construction cost and performance risk but not market price risk) using project nancing was estimated to be $13.9 million per year. The annual cash ow for the same merchant project without a long-term contract (taking on, in addition, market price risk) using project nancing was estimated to be $16.5 million per year. Thus, the nancing costs for a traditional merchant project that relies on variations in spot market prices would be about 70 per cent higher than a regulated utility nanced project. This capital cost variation suggests that the eciency benets of merchant versus regulated projects would have to be quite large to justify relying on merchant investment. Joskow and Tirole (2005b) identify lumpiness as one barrier to ecient investment under a merchant transmission investment model. Lumpiness is a relative not an absolute size concept. That is, whether an investment project is lumpy or not must be measured relative to the impact of the most eciently sized project on the congestion rents that it would reduce. The post-investment congestion rents are the source of the revenue that a merchant investor would count on to support the investment. Regardless of the absolute cost of the project, if an ecient (benets greater than costs) project of optimal scale were to eliminate congestion completely, for example, there would be no way for it to be nanced under a merchant investment framework. Similarly, a large project of optimal size (for example, a 1,000-MW HVDC link to New York City) may not have such a large eect on price dierences as to make the investment uneconomical. Some commentators have suggested that the lumpiness problem can be addressed by treating very large projects dierently from small projects. This policy prescription reects a
157
misunderstanding of what lumpiness means in this context. Indeed, this policy advice is likely to get it backwards. As we shall see in the discussion of PJMs transmission investment policies below, there are many small projects that completely mitigate congestion and, accordingly, would not be nanced on a merchant basis. At the same time, the merchant projects that are attracting the most attention are large projects that link market areas with demands that are much larger than the scale of the projects and have signicant sustained congestion and the associated locational price dierences. These large projects are small relative to the size of the markets that are being linked and, as a result, their completion is not expected to have a large eect on dierences in locational prices. While I view the opportunities for merchant transmission projects as being limited primarily to inter-TSO investments that fall outside of TSO regional planning procedures, where there exist large sustained price dierences and where a regulated entity is willing to provide long-term contract support for the project, there is no reason why such projects should not be accommodated in the regulatory and planning process. A practical model for doing so has emerged in PJM and I shall discuss it further below.
5.
TRANSMISSION REGULATION AND INVESTMENT FRAMEWORK IN ENGLAND AND WALES
In 1990, the electricity sector in E&W was privatized and restructured to create competitive wholesale and retail markets for power. The state-owned generation and transmission company (Central Electricity Generating Board: CEGB) that historically had provided wholesale power to distribution entities (area boards) and large industrial customers in E&W was broken into three generating companies and a single regulated transmission company (National Grid Company (NGC)). NGC owns the E&W highvoltage transmission network (400 kV and 275 kV facilities), maintains the network and is responsible for making investments in it to meet its obligations specied by various license conditions. It also is a joint owner with RTE (the French transmission operator) of a 2,000-MW HVDC transmission link between France and E&W.11 There has been much written about the design and performance of the wholesale power markets in England and Wales (for example, Henney, 1994; Wolfram, 1999; Sweeting, 2000). Accordingly, I shall provide only a brief description of these wholesale market arrangements. From 1990 until
158
March 2001, the wholesale market for power was built upon a mandatory bid-based pool which determined the economic dispatch and associated uniform market-clearing price for energy (and where applicable capacity payments) for each of 48 thirty-minute periods each day. Generators were eectively provided with rm transmission service in the sense that if NGC had to dispatch generators out of bid merit order to deal with congestion and other network operating constraints it had to pay generators either to reduce their scheduled generation or to increase it. In March 2001, the New Electricity Trading Arrangements (NETA) was introduced. NETA replaced the mandatory pool with a new wholesale market design that was structured to encourage generators and load to enter into bilateral contracts and to minimize the amount of trade going through a centralized pool. NETA requires generators and loads to submit generation and demand schedules up to a short period before actual dispatch. These schedules became nancial commitments on the part of generators and loads. NGC is then responsible for balancing the system using oers to buy and sell increases and decreases in real-time generation supplies mediated through a pay-as-bid balancing market. NGCs balancing responsibility includes real-time balancing of demand and supply for energy and management of network congestion and other network operating constraints. Generators or load that voluntarily deviates from their schedules must (eectively) buy or sell energy in the balancing market. As before, generators paying interconnection and use of system charges (below) are eectively buying rm transmission service and must be compensated if NGC needs to increase or decrease their output from the pre-scheduled levels to manage congestion and other network constraints. Among other things, NGCs license conditions and associated codes and standards specify the operating procedures and principles governing NGCs relationships with all users of the transmission system (generators, distributors and retail electricity suppliers). Under its transmission license, NGC must operate the network in an ecient, economical and coordinated manner and oer its services based on non-discriminatory terms and conditions. Transmission System Security and Quality of Service Standards have been developed to govern NGCs responsibilities. These codes and standards dene reliability criteria that are to be used by NGC to plan needed enhancements to the transmission system and to identify transmission investment requirements. NGC evaluates transmission investment needs and alternatives to meet these obligations on an ongoing basis. It publishes an annual Seven-Year Forward Statement12 which provides forecasts of demand, supply, approved transmission enhancements and expected transmission enhancements that would be needed to accommodate additional generation at various locations on the E&W grid. The
159
Seven-Year Statement is made available to provide information to new generators regarding the capabilities of the network to accommodate new generating capacity at various future dates and the network enhancements that NGC has identied as being required to accommodate new generating capacity of various amounts at dierent locations on the network. The Connection and Use of System Code (CUSC) species a contractual framework for interconnection to and use of the network. NGC is also the system operator for E&W and thus is vertically integrated into all aspects of transmission operation, maintenance and investment.13 NGC is subject to regulation by the Oce of Gas and Electricity Markets (Ofgem). Separate but compatible incentive regulation mechanisms are applied to the transmission owner (TO) and system operating functions (SO). These regulatory mechanisms eectively yield values for the revenues NGC is permitted to earn from charges for transmission service and system operations. Transmission customers (generators and retail suppliers) pay NGC for transmission service pursuant to a regulated tari. The tari has two basic components. The rst is a shallow connection charge that allows NGC to recover the capital (depreciation, return on investment, taxes and so on) and operating costs associated with the facilities that support each specic interconnection (now using the plugs methodology). The second component of the transmission tari is composed of the transmission network use of system (TNUoS) charges. The general level of charges is set to allow NGC to recover its cost-ofservice-based revenue requirement or budget constraint as adjusted through the incentive regulation mechanism that I shall discuss presently. The structure of the TNUoS charges provides for price variation by location on the network based upon (scaled) dierences in the incremental costs of injecting or receiving electricity at dierent locations as specied in the investment cost-related pricing methodology. So, for example, generators pay signicantly higher transmission service costs in the North of England than in the South (where the prices may be negative) because there is congestion from North to South and deep transmission network reinforcements are more likely to be required to accommodate new generation added at various locations in the North but not in the South. Similarly, load in the South pays more than load in the North because transmission enhancements to increase capacity from constrained generation export areas benet customers in the South more than those in the North. The locational TNUoS charges for 2004/2005 for generation and demand are displayed in Tables 5.2 and 5.3. The stated objective of this pricing mechanism is as follows:
[E]cient economic signals are provided to Users when services are priced to reect the incremental costs of supplying them. Therefore charges should reect
160
Table 5.2 Schedule of transmission network use of system generation charges (/kW), 2004/2005
Generation zone Zone area Generation tari (/kW) Short-term generation tari (/kw) STTEC period 28 days 1.891940 1.211112 1.306676 0.865602 2.250223 1.472388 0.606847 0.426739 0.000000 0.000909 0.364065 0.000000 0.000000 0.000000 0.000000 STTECC period 35 days 2.364925 1.513890 1.633345 1.082002 2.812779 1.840485 0.758559 0.533423 0.000000 0.001137 0.455081 0.000000 0.000000 0.000000 0.000000 STTEC period 42 days 2.837910 1.816668 1.960014 1.298402 3.375334 2.208582 0.910271 0.640108 0.000000 0.001364 0.546097 0.00000 0.000000 0.000000 0.000000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Note: STTEC
Northern Humberside North West Pennies & North Wales Dinorwig Anglesey East Anglia West Midlands South Wales & Gloucs Oxon & Bucks Estuary Central & SW London South Coast Wessex Peninsula
9.009237 5.767201 6.222266 4.121912 10.715347 7.011370 2.889748 2.032089 2.150590 0.004330 1.733641 6.604821 1.507146 3.829097 6.836065
short term tari of electricity.
Source: NGC (2004b).
the impact that Users of the transmission system at dierent locations would have on National Grids costs, if they are to increase or decrease their use of the system. These costs are primarily dened as the investment costs in the transmission system, maintenance of the transmission system and maintaining a system capable of providing a secure bulk supply of energy. (NGC, 2004a, p.12)
Finally, in its role as system operator, NGC has an obligation to balance the supply and demand for energy in the system in real time (energy balancing) and to meet operating reliability criteria (system balancing). These costs include the net costs NGC incurs to buy and sell power in the balancing market (or through short-term bilateral forward contracts) to balance supply and demand at each location, including to manage congestion, provide ancillary services, and other actions it must take to meet the networks operating reliability standards. These costs are recovered from
161
Table 5.3 Schedule of transmission network use of system demand charges (/kW) and energy consumption charges (p/kWh), 2004/2005
Demand zone 1 2 3 4 5 6 7 8 9 10 11 12 Zone area Northern North West Yorkshire North Wales and Mersey East Midlands Midlands Eastern South Wales South East London Southern South Western Demand tari (/kW) 4.940866 8.325173 8.455923 8.709914 10.771600 12.600874 11.007104 16.130442 14.321101 16.761568 15.679987 17.798154 Energy consumption tari (p/kWh) 0.656585 1.100254 1.171611 1.107068 1.479424 1.733413 1.394934 2.228075 1.773924 2.430277 2.076489 2.198679
Source: NGC (2004b).
system users through an uplift charge based (mediated through an incentive regulatory mechanism discussed further below) on the quantities of energy supplied to or taken from the network. The regulatory framework for determining the revenues that NGC can recover through the use of system charges and the energy and system balancing charges is based on a set of incentive regulation mechanisms. These mechanisms have a cost-of-service base, a performance-based incentive, and a ratchet that resets prices from time to time to reect NGCs realized or forecast costs. A base annual aggregate revenue requirement for use of system charges is established at the beginning of each ve-year price review period (though the latest period is being extended to seven years). The starting revenue requirement is determined based on a fairly standard cost of service principles. A rate base (regulatory assets value: RAV) is dened that is composed of the carrying value for the existing assets that make up the transmission system plus the forecast cost of incremental capital expenditures budgeted for the next ve years to meet NGCs interconnection and system security criteria described above. The nal investment budget is determined by Ofgem through a public consultation process. Depreciation rates and a cost of capital (allowed rate of return) are dened and applied to the RAV to yield allowed capital charges for the starting year. Current allowable operating and maintenance (O&M) expenditures are dened and
162
added to the year one capital charges. A target rate of productivity improvement in O&M charges the X factor is then dened. The value of X is determined through a regulatory consultation process based on NGCs forecasts of O&M requirements, wage escalation, and various benchmarking studies performed for Ofgem by independent consultants. The starting value for allowed capital charges is then adjusted each year for budgeted incremental capital additions and changes in an ination index while allowances for O&M costs are escalated based on a general price index minus X. Unbudgeted capital expenditures during the price review period can be considered in the next price review, though NGC may be at risk for amortization charges during the period between reviews. Underspending on capital may also be considered in the next price review and adjustments made going forward. After a ve-year (or longer) period another price review is commenced, the starting price is reset to reect then-prevailing costs, and new adjustment parameters dened for the next review period.14 In its role as the E&W system operator, NGC has also been subject to a set of incentive regulation mechanisms. Each year forward targets are established for the costs of energy and system balancing services. A sharing or sliding scale formula is specied which places NGC at risk for a fraction (for example, 30 per cent) of deviations from this benchmark (up or down) with caps on prots and losses. Table 5.4 displays the attributes of the SO incentive mechanism in eect since NETA went into operation. However, Ofgem and NGC were unable to agree on a new SO incentive mechanism for 2006/2007 and a traditional cost of service recovery mechanism was adopted as a default. Ofgem has also applied a new incentive regulation mechanism that would apply to network outages that lead to variations in the fraction of lost energy resulting from transmission network outages (Ofgem, 2004b). This brings us nally to the transmission investment framework. NGC has the obligation to identify transmission investments required to meet its obligations under the Grid Code, the Transmission System Security and Table 5.4
Parameter Target expense (m) Upside sharing (%) Downside sharing (%) Cap (m) Floor (m)
Source: Ofgem (2003b).
E&W system operator incentive mechanism under NETA

First year 484.6514.4 40 12 46.3 15.4 Second year 460 60 50 60 45 Third year 416 50 50 40 40
163
Quality of Service Standards, and the Connection and Use of System Code. The Transmission System Security and Quality of Service Standards are engineering reliability criteria used for planning purposes that have largely been carried over from the pre-restructuring era. They are of fundamental importance for transmission investment planning purposes. The transmission planning process is built around a set of reliability criteria designed to meet these security and quality of service standards. The Standards specify criteria (to oversimplify) for dening a set of boundary circuits and associated power ows over which the generating capacity on one side of the boundary must be able to ow reliably (thermal, voltage and stability) over the boundary to serve demand there if any two circuits are out of service. NGC performs power ow studies based on forecasts of demand and generating capacity at various locations to identify boundaries (individually or collectively) where reliability criteria may be violated during the forecast period. Transmission investment projects are then identied which will restore the relevant reliability criteria when and if they are expected to be violated. Depending on the nature and magnitude of the transmission investments identied, various siting approvals must be obtained for proceeding with actual investments. NGC will also seek to include these projects in the investment case for the subsequent price review. These planning criteria do not take the economic cost of congestion directly into account. However, the reliability criteria eectively provide rm transmission service to system users under the study conditions used for transmission planning purposes and necessarily mitigate congestion under the study conditions in the process of meeting reliability criteria. However, variations in supply and demand conditions, as well as outages of transmission facilities, can lead to congestion in real-time operations. Through the balancing incentive mechanisms, NGC must pay for a share of the costs of balancing the system in the face of congestion that may arise in real-time operations. This provides additional incentives to NGC to make transmission investments with short paybacks that were not included in the plan upon which the price control was based or to advance investments in the base plan to reduce congestion and other system balancing costs. It also provides incentives for NGC to maintain the network and spend resources on restoration of outages when these expenditures are valuable because they reduce system balancing costs. As with all incentive regulatory mechanisms, these mechanisms reect a balancing of the incentives to reduce costs and meet quality standards and capturing the rents from cost reductions for consumers (Tirole and Laont, 1993, Chapter 1). Investments in interconnectors with other networks are not covered directly by NGCs license. The existing interconnector with France is now organized as a separate merchant business and the associated capacity is
164
allocated by auctioning physical rights of various durations. In principle, both NGC and third parties are free to propose adding interconnectors between NGCs network and, for example, France or Belgium. The regulatory treatment of such facilities can be negotiated with Ofgem, though the assumption has been that these facilities would be built on a merchant basis. No interconnectors have been added since the CEGBs restructuring in 1990, so how this would play out in practice is unclear. Moreover, the UKs interconnector policies are in the process of being harmonized with the regulated third party access regulations specied by recent EU directives (Ofgem, 2003b, 2004a).15 The liberalization program in E&W is, in my view, the most successful in the world. An important component of this successful restructuring was the creation of a single independent transmission company that combines transmission ownership, maintenance and system operating responsibilities in an organization that spans the entire transmission network covering E&W. The utilization of and continuous improvement in benchmarking and incentive regulation have also contributed to the excellent performance exhibited in the transmission segment since the mid-1990s when these mechanisms were rst introduced. While the fact that England and Wales is (almost) an electrical island, eased the successful reliance on a single national network compared to the loop ow and related issues that national networks must face in continental Europe, the E&W experience contains many lessons for successful operation of and investment in highvoltage transmission networks in liberalized electricity sectors.
6.
MARKET, REGULATORY AND TRANSMISSION POLICIES IN PJM
It is dicult to describe or evaluate transmission investment policies in the US in a simple way. This is the case for several reasons. First, transmission policy in the US has been in a constant state of change for the last decade. Second, the regulatory responsibility for important aspects of transmission policy is split between the federal government and the states and reects the legacy of vertically integrated utilities regulated primarily by the states. Third, dierent states have taken very dierent approaches to liberalization of the electricity sector (Joskow, 2005). No federal laws have been enacted clearly to promote wholesale and retail competition or the changes in supporting institutions required to help to make these competitive initiatives achieve their goal of providing long-term benets to consumers. Fourth, the availability of consistent data on transmission prices, investment, and network performance is extremely limited (US EIA, 2004). Accordingly, I
165
shall focus here on transmission pricing and investment policies in PJM where FERCs vision for the ideal model for wholesale market design and transmission institutions (the so-called standard market design: SMD) has been implemented and for which we now have several years of experience. A more detailed discussion of US transmission pricing and investment policies can be found in Joskow (2005). Industrial Organization of PJM and Wholesale Market Design PJM entered the electricity liberalization era as a multi-state power pool (tight pool) which centrally dispatched the generating facilities for vertically integrated utilities in Pennsylvania, New Jersey, Maryland, Delaware and Washington DC based on the marginal costs of the generating units owned by PJMs member utilities. PJMs origins and experience in economic generator dispatch, management of network reliability, and system planning can be traced back to the 1920s when it began to be created by the private vertically integrated electric utilities in this area. In 1998, the PJM agreement was restructured to turn the cost-based power pool into a set of bid-based wholesale spot power markets and supporting institutions, including transmission pricing and investment protocols. PJM is now an ISO and has been qualied as an RTO by FERC pursuant to Order 2000. It is structured as a for-prot limited liability company with an independent board of directors, though it presently operates de facto as a non-prot organization. PJM is not a market participant, does not own generation, transmission and distribution assets and is not engaged in wholesale or retail marketing.16 It is responsible for system operating reliability and for applying reliability rules and criteria developed by regional reliability councils (Mid-Atlantic Area Council (MAAC) in the case of the original PJM footprint). PJMs geographic footprint has expanded in the last couple of years to include transmission owners in portions of Pennsylvania that were not previously in PJM, and utilities covering portions of West Virginia, Ohio, Kentucky, Indiana, Virginia and Illinois.17 (The Midwest ISO (MISO) includes the transmission owners covering the rest of these Midwestern states.) The transmission owners in PJM are all vertically integrated utilities that also own generating capacity and distribution companies, and have unregulated wholesale and retail marketing aliates. They continue to have transmission operating functions, including transmission maintenance, outage restoration and investment responsibilities, subject to various agreements between the transmission owners and PJM and supporting FERC regulations. The prices for unbundled transmission service made available by these transmission owners to third parties (generators, retail and wholesale
166
marketers, and unaliated distribution companies) is regulated by FERC. The prices for bundled transmission service that the vertically integrated transmission owners make available to their own retail customers (those who have not agreed to be supplied by competitive retail suppliers in those states with retail competition) are eectively regulated by each state as part of the overall regulation of the prices for bundled retail service. Thus, the same transmission facilities are compensated through two regulated revenue streams, one (unbundled) governed by FERC regulation and one (bundled) governed by state regulation. The prices for transmission services are set based on traditional cost-of-service or rate of return principles (as discussed in more detail in Joskow (2005) and below) applied to each transmission owners facilities. Although FERC Order 2000 encourages it, there are no formal incentive regulation mechanisms applicable to costs or quality of service that is applied by FERC or the state regulators to either the TOs in the PJM area or to PJM itself. PJM operates (voluntary) day-ahead and real-time (adjustment or balancing) bid-based markets for energy and ancillary services. Market participants submit bids and oers to the day-ahead and real-time markets. LMPs that balance supply and demand at each location on the network and the allocation of scarce transmission capacity are performed together using a least-cost bid-based security-constrained dispatch (state-estimator) model that incorporates the physical topology of the network and reliability constraints. The LMPs reect equilibrium marginal energy costs and the marginal cost of congestion at each location (marginal losses will be included soon, as in the LMP systems in New York and New England). Participation in day-ahead and real-time markets is voluntary in the sense that generators, loads and marketing intermediaries may submit their own day-ahead schedules for energy and ancillary services to the RTO and can (try to) use bilateral arrangements to stay in balance in real time. However, bilateral schedules are still liable for congestion and loss charges and any residual imbalances are settled at the real-time prices. Congestion is priced based on the dierence in LMPs between the designated delivery and receipt points of generation supplies chosen by a transmission service customer. Load serving entities (LSEs distribution companies or competitive retail suppliers which have responsibility for supplying retail consumers) in PJM have forward generation capacity obligations based on their expected peak loads in each month and must contract forward for capacity or pay penalties. PJM operates capacity markets, but it appears that bilateral arrangements govern the allocation of most qualifying generating capacity. These capacity markets are in the process of being restructured as this is written. Generators must meet certain transmission deliverability requirements to qualify as capacity resources. As discussed further below, these
167
deliverability requirements play an important role in the transmission investment process and in providing locational incentives to generators. Transmission Pricing and Related Policies PJM administers an open access transmission tari that requires the transmission owners in PJM to oer transmission services at non-discriminatory cost-based prices. This tari (along with the PJM Operating Agreement and the PJM Reliability Assurance Agreement which are interdependent) establishes prices for various categories of transmission service available to third party transmission users;18 denes how the associated revenues are distributed to TOs; species interconnection rules and obligations for generators, merchant transmission owners (none yet) and regulated TOs; species the denition, allocation mechanisms, accounting and settlements for nancial transmission rights (FTRs); and establishes a process for identifying and approving regulated transmission expansion projects and the allocation of the associated costs and FTRs.19 These transmission pricing arrangements are being revised as this is written. Transmission Investment Framework Transmission investments in PJM are grouped into several categories. Direct interconnection investments When a new generating unit or merchant transmission projects seeks to connect to the PJM network, the TO in whose transmission zone the project will be located performs a study of the direct capital and operating costs associated with the new transmission facilities required to make the direct connection to the network. The proposed generating project is responsible for 100 per cent of these direct interconnection costs. About $304 million of investments that appear in PJMs July 2004 Regional Transmission Expansion Plan (RTEP) (PJM Interconnection, 2004f) update fall in this category, out of a total approved projects of about $785 million.20 Direct interconnection costs are therefore treated similarly in PJM and E&W. Interconnection network reliability investments PJM and the TO in whose transmission zone the facility is located also evaluate the impact of the proposed project on network reliability. A series of engineering studies are performed to assess whether the proposed project, as an increment to the existing facilities on the network, will lead to any violations of PJMs reliability criteria. These criteria are much more complex than the simple N 1 operating reliability criterion that is often
168
discussed in the literature. The reliability assessments involve a set of assumed study conditions under various contingencies: when all facilities are operating; N 1; N 2; multiple contingencies; and delivery to load criteria. These criteria and their application have not changed signicantly since before the new PJM markets were created and take no account of the LMP mechanisms or of the associated market mechanisms for allocating scarce transmission capacity. If the engineering studies indicate that reliability criteria are violated, the expected costs of network investments required to restore the reliability parameters are identied. The proposed generator will be required to pay for these costs, though they may be shared with other generators in the construction pipeline that benet from these network enhancements (the cost allocation mechanism is fairly complicated). The generator will receive its proportionate share of any new FTRs (or the revenues produced when these rights must be auctioned) created as a consequence of the network facility enhancements it is required to pay for. It is important to note that these reliability assessments are based on a set of engineering assumptions and study conditions that may bear little relationship to the way the network would actually operate if the network enhancements were not made and increased congestion was realized. That is, if the generators were built and these deep network enhancements were not made, the network would not necessarily suer a violation of its operating reliability criteria. Instead, redispatch would have to be used to balance the system. Generator deliverability investments If a generator or HVDC merchant transmission project wants to qualify as a capacity resource under PJMs Reliability Assurance Agreement and wholesale market Operating Agreement, as is typically the case since there is signicant capacity value in the PJM market, they must meet a nal reliability criterion called generator deliverability. Engineering studies are performed to determine whether (oversimplifying a complex process) the full power that the proposed generator can produce can be reliably delivered outside of its transmission zone under a set of engineering study conditions that assume all existing generators are dispatched rst to meet load.21 If the generator deliverability condition is not satised the generator must either pay for any necessary network enhancements (and receive any incremental FTRs) or purchase rm transmission service from a third party that supports deliverability. Interconnection network enhancements and deliverability network reliability enhancements together account for about $207 million of investments in PJMs 2004 RTEP update (as of July 2004). These obligations are conceptually most similar to the generator component of the locational TNUoS charges in E&W. Thus, generators are
169
obligated to pay for about $511 ($304 million direct interconnection + $207 million deep network upgrade investments) of the roughly $785 million of transmission investments approved through the 2004 PJM planning process or about 75 per cent. Thus, PJM eectively has a deep interconnection pricing policy. It should be noted that interconnection network investments and deliverability network investments provide potentially powerful locational incentives to new generating projects. The network upgrade costs at some locations may be zero (or even negative) and at other locations these costs may be substantial (as are the generator TNUoS charges in E&W). New generators can reduce their investment costs by selecting a location where these network upgrade obligations are low rather than high. It is likely that these interconnection network upgrade cost obligations play a more important role in generator location decisions than do variations in LMPs. Other network reliability investments The PJM RTEP process may indicate that one or more PJM/MAAC reliability criterion is expected to be violated for other reasons, for example, load growth or generator retirements at specic locations. PJM can direct TOs to make the necessary investments required to restore the reliability parameters. The associated costs are then recovered from charges to the load that benets from the investments. These costs amount to about $274 million in the 2004 RTEP. This appears to be the fastest-growing category in the RTEP planning process and would include network upgrades required as a consequence of retirements of existing generating facilities. Merchant transmission investments The original design of the PJM system was predicated on the assumption that any economic transmission investments that were not required for reliability would be made on a merchant basis. The costs of merchant transmission projects would be borne by the developer and the developer in turn would receive the nancial transmission rights created by the investment. The incentive for merchant investment would then be the market value of the transmission rights created by the project. The associated expected value of the transmission rights created is then the expected dierence between the LMPs between the aected delivery and receipt points times the incremental transmission capacity between these points created by the investment (Joskow and Tirole, 2005b). In the case of AC facilities, a merchant investor would receive any incremental FTRs resulting from the investment. HVDC merchant transmission facilities are treated like generators and eectively create physical import or export rights to the AC network.
170
Table 5.5
PJM interconnection charges: proposed Erie West HVDC

Estimated cost ($m)
Direct connection facilities Deep network upgrades Single contingency Second contingency Multiple facility contingency Generator Deliverability Other Total cost interconnection cost 3.5 year construction time
9.5 91.5
102
Note: Transmission Interconnection Quene #G00_MTX3 is a TransEnergie U.S. Ltd. request to connect the southern terminal, of the Erie West to Nanticoke HVDC intertie, to the Erie West 345 kV substation. TransEnergie proposes to construct an HVDC converter station in the vicinity of Erie West, and a double circuit 345 kV line to connect Erie West to the converter station. The northern terminal of the intertie will be connected to Nanticoke substation in the Ontario Hydro system. The interconnection request is nominally rated at 1,000 MW net of losses on the HVDC system. The developer has requested Firm (Capacity) Transmission Injection Rights in the amount of 1,000 MW and Firm Transmission Withdrawal Rights in the amount of 1,060 MW at the HVDC terminal in PJM. Project #G00_MTX3 is scheduled for commercial operation in 2004. Source: PJM Interconnection (2003).
Merchant transmission projects must also pay for direct interconnection and deep network upgrade costs in essentially the same way as do new generators. Table 5.5 22 illustrates the results of the PJM interconnection study process and the estimated costs of direct interconnection and reliability network upgrade costs for a proposed merchant HVDC project under Lake Erie connecting Ontario with Pennsylvania (now cancelled). The total interconnection costs for this project were estimated to be $102 million of which about 10 per cent were direct interconnection charges and 90 per cent deep network upgrades to restore a long list of reliability problems expected to be created by the project. PJMs deep interconnection pricing policies for new generators and merchant investment projects are not typical of the pricing of interconnection and transmission use of system services elsewhere in the US. A shallow interconnection policy is more typical. Generators pay direct interconnection charges as in PJM. The costs of network upgrades deeper in the network are then typically rolled in with the legacy network costs to create use of system charges that are identical at all interconnection points on an individual TOs network. FERCs most recent interconnection rule provides for shallow rather than deep interconnection charges.23 As RTOs have grown, FERC has
171
endeavored to (eectively) reallocate these costs to eliminate pancaking and to shift network use charges to load from generators (Joskow, 2005). These reallocations of transmission costs have been quite controversial. Several merchant transmission projects have been proposed through the PJM interconnection and regional transmission planning process, primarily DC interconnectors with neighboring control areas. Two transformer upgrades have been made by a TO in PJM as merchant projects in return for FTRs. None of the proposed DC interconnectors has yet gone into operation and several have been cancelled. The most active projects are HVDC interconnections between PJM and New York City and Long Island. The farthest along is a project that has been awarded a long-term contract for transmission between PJM and Long Island by LIPA, a municipal utility which can pass on the associated costs to its regulated customers without approval of a state or federal regulatory agency. LIPA already has a long-term contract for all of the 330 MW capacity of the Cross Sound Cable connecting New England with Long Island, the only merchant project completed so far in the US. HVDC links to New York City and Long Island are especially attractive for a number of reasons. The LMPs in New York City and Long Island are consistently signicantly higher than those in neighboring areas about $20/MWh on an annualized basis. In addition, these are both very dicult places to nd sites for new power plants and have extremely high construction costs. In addition, HVDC links from PJM and New England can be brought in under water where Nimby issues should be less of a problem (though this did not mute the controversy over the Cross Sound Cable process). Finally, on Long Island there is a municipal distribution utility that is willing and able to sign long-term contracts for the transmission capacity developed in this way. This means that the developer does not have to rely on dierences in spot market LMPs to produce the revenues for the project, reducing nancing costs and opportunism problems. Economic planned transmission facilities PJM resisted doing any analysis of economic transmission investment opportunities or including such potential investments in its regional transmission plan and requiring TOs to proceed with them if merchant investors did not show any interest in them. As before, by economic transmission investment opportunities I refer to transmission investments whose expected economic benets arise from reducing congestion (and losses). When the expected incremental reduction in congestion and loss costs exceeds the incremental cost of the network enhancement then the investment is economical.
172
PJMs dream that the invisible hand would lead merchant investors to come forward to make intra-TSO investments in response to congestion rents has not been matched by reality. After a contentious regulatory proceeding, in 2003 FERC issued an order that required PJM to include potential economic transmission investments in its planning process. PJM has now developed a process to identify transmission constraints that create unhedgeable congestion and to assess the benets and costs of potential network enhancement projects that would mitigate this congestion. When projects that mitigate unhedgeable congestion are identied and pass certain cost/benet thresholds they are included on a market window list. The projects on this list are then open for one year to proposals from merchant investors. If satisfactory proposals are not forthcoming, PJM may direct incumbent TOs to build the projects as regulated projects and include them in the PJM tari for cost recovery. The process is complex, still evolving, and the phrase unhedgeable congestion somewhat misleading. The process for identifying so-called unhedgeable congestion actually yields an estimate of the costs of congestion after netting out congestion rents. To oversimplify,24 PJM denes unhedgeable congestion as congestion which cannot be hedged with the existing portfolio of FTRs. The best way to think of PJMs unhedgeable congestion concept is as an approximation to the social cost of congestion. And this appears to be the number that one actually would want to use in order properly to evaluate potential economic transmission investment opportunities. For the 14-month period from August 2003 to September 2004 there was $1.6 billion of gross congestion in PJM (including congestion rents), of which $336 million was dened as unhedgeable.25 Where unhedgeable congestion is identied, a set of simple costbenet assessments of transmission upgrades are then performed by PJM. The actual unhedgeable congestion values attributed to each constraint over the previous 12-month period is divided by the estimated cost of a transmission upgrade that would mitigate the congestion costs identied.26 This is dened as the benet/cost ratio, though it is actually a measure of the simple payback period for each identied investment opportunity assuming that congestion rents do not change in the future. When these assessments yield benet/cost ratios that exceed certain specied thresholds a project is put on a list of potential regulated economic transmission projects. Market participants are then given a year to propose alternative market solutions to the identied projects. If market solutions are not forthcoming, the projects are added to the PJM RTEP and the incumbent TOs in whose transmission zones the projects are located are directed to make the investments. The resulting costs, net of revenues from the auctioning of FTRs created by the
173
investments, are then recoverable through the PJM Open Access Tari from the customers of the LSEs who are expected to benet from the investments. The responses to the rst market window open for proposals to resolve this economic congestion were set in 2005. Roughly 50 potential economic transmission investment projects have been identied since this evaluation process was implemented in March 2004 and market windows are now open for merchant projects to ll these needs before regulated transmission projects are added to the RTEP.27 The costbenet analyses indicates that seven of the identied projects have simple paybacks of three months or less (again, assuming that unhedgeable congestion does not change in the future). Another 12 have simple paybacks of less than four years (see Table 5.6). If FERC had not forced PJM to examine economic transmission investment projects, all of these would have been left on the table in the hope that merchant investment would eventually come forward. It should also be noted that in several cases, fairly small investments completely eradicate the congestion so that they are not conducive to being supported by merchant investments. Apparently three merchant transmission network upgrade proposals have been made so far in connection with the rst market window, though studies have not been completed and they may be aected by other transmission investment projects that have been proposed.28 Inter-TSO (interconnector) investments The expansion of interconnections with neighboring control areas is not included in the PJM planning process, though procedures are in place that govern the rules governing pricing of the costs of interconnecting interTSO facilities to the PJM network. Accordingly, by default, inter-TSO transmission investments are left to merchant developers. As already discussed, a few merchant HVDC links with New York City and Long Island have been proposed and at least one is likely to move forward, supported by a 20-year contract with LIPA. There is little if any additional merchant investment activity on the horizon. However, by incorporating neighboring TSOs into PJM by expanding its boundaries, it is eectively internalizing inter-TSO transmission investment opportunities (as well as integrating generator scheduling, wholesale market and congestion management mechanisms) into the intra-TSO transmission investment planning process. As these additional TSOs are integrated into PJM, the PJM generator interconnection, reliability and economic investment protocols will apply to what were previously inter-TSO opportunities that have largely been ignored due to the balkanization of transmission ownership and system operations. Just as a fairly large number of economic transmission investment opportunities popped up once PJM actually looked for them, I expect
174
Table 5.6 Market window economic transmission projects in PJM as of November 2004
Monitored facility LINE 230 KV ADA-BRUX LINE 500 KV BED-BLA BED-BLA LINE 230 KV ADA-BENX LINE 138 KV BRU-EDI LINE 69 KV SHI-VIN LINE 500 KV FTM-PRU PJMW500 LINE 230 KV NWA-WHI EAST JACK ME 230 KV 4 BA-P YORKANA 230 KV 1A BANK LINE 230 KV CED-CLIK LINE 230 KV BER-HOB LINE 138 KV EDI-MEAR LINE 500 KV ELR-HOS LINE 69 KV EDG-NSA *Unhedgeable congestion ($) 1,091,588 1,607,237 83,999,705 4,146,221 1,134,130 3,397,773 307,337 3,284,457 2,739,456 2,264,606 2,454,986 1,647,801 709,851 654,222 499,774 112,364 47,120 Limit Circuit switcher Wavetrap Voltage Circuit switcher Circuit switcher Conductor Disconnect switch Voltage Conductor Voltage Transformer Transformer Disconnect switch Cable Circuit switcher Wave trap Disconnect switch Cost to relieve limit ($) 200,000 75,000 525 million 200,000 200,000 500,000 45,000 525 million 1,000,000 525 million 2,500,000 2,500,000 50,000 2 million 200,000 300,000 20,000 Cost/ benet 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.254 0.254 0.254 0.254 0.254 0.254 0.254 0.254 0.254 0.254
175
Table 5.6
(continued)
Unhedgeable congestion ($) 200,355 9,272,381 Limit Wave trap Transformer Cost to relieve limit ($) 200,000 2,500,000 Cost/ benet 0.254 0.254
Monitored facility LINE 230 KV BRA-FLA JACK ME 115 KV 5 BA-S
Source: PJM Interconnection (2004f).
that many more reliability and economic projects will emerge as PJMs transmission planning footprint grows to incorporate what were previously separate TSOs. Two major new regulated transmission projects between what were previously separate TSOs were proposed in 2006. Despite the investment in new intra-TSO facilities in PJM, congestion charges in PJM continue to grow (see Table 5.1). Moreover, the prospect of a growing number of generation retirements is also leading to a need for network reliability investments. Since there are no exit fees, these charges are likely to be paid for by the TOs in the areas where the retiring generators are located (PJM, 2004c).
7.
RELIABILITY VERSUS ECONOMIC TRANSMISSION INVESTMENT
All economic models of transmission investment that I am aware of focus on transmission investment as a mechanism to reduce the costs of congestion (Joskow and Tirole, 2000, 2005b). Some (properly) include the cost of losses as well. When transmission capacity is congested, high-cost generation must be substituted for low-cost generation to balance supply and demand. The incremental cost of the high-cost generation that must be dispatched due to transmission capacity constraints plus any deadweight loss associated with reduced demand resulting from higher locational prices is the cost of congestion. Transmission investment should then optimally be made (ignoring lumpiness, market power and other market imperfections) up to the point where the incremental cost of transmission capacity is equal to the incremental reduction in the expected present discounted value of congestion and loss costs. These models bear little if any relationship to the way intra-TSO transmission investments are actually evaluated by TSOs in the US and E&W.
176
As we have seen, in E&W and PJM, virtually all of the transmission investments that have been approved have been justied either by direct interconnection costs or by reliability considerations.29 The E&W system does not even appear to have a transmission investment concept akin to economic transmission investments that are justied by savings in congestion costs aside from the incentives to reduce congestion costs embodied in the SO incentive mechanism. In New England, with a similar market design to PJMs, the New England ISO manages a very detailed regional transmission expansion planning process that examines needs and opportunities for both reliability transmission investments and economic transmission investments. This process includes models that forecast congestion. The 2004 update to the New England ISOs regional transmission expansion plan identied $2 billion ($1.5 to $3.0 billion) of transmission investment projects and essentially all of them are justied as reliability investments (ISONE, 2004). Not a single project was identied which could be supported by congestion cost savings alone. In fact, many network upgrade investments that are justied on reliability grounds could just as well be categorized as economic transmission investment opportunities. In many cases, if the investments were not made, the network could still be operated reliably, but there would be more congestion, more controlled load shedding, and much higher prices in some areas. Moreover, many reliability investments aect the future trajectory of LMPs and incentives for generation and transmission investments. On the other hand, economic transmission investments can also often confer reliability benets as well. Thus, in my view, at the very least, reliability and economic transmission investments are interdependent. At worst, the distinction between them is analytically awed. Moreover, the distinctions between reliability-driven and congestion cost-driven transmission investments creates a very signicant asymmetry between the treatment of intraTSO network investments and inter-TSO network investments. The former are evaluated and priced as reliability investments while the latter must be justied and paid for based on congestion cost savings alone, by default on a merchant basis. It is fairly clear that transmission investments driven by reliability criteria have signicant eects on LMPs and network congestion. In addition, discretionary changes in system-operating practices, including changes in the ways that operating reliability criteria are applied or evaluated, can have dramatic eects on the capacity of portions of the network and on the resulting congestion rents and congestion costs. In the studies underlying the New England ISOs 2004 regional expansion plan it is quite evident that reliability investments get triggered well before locational prices or congestion are allowed to rise anywhere close to the
177
value of lost load (ISONE, 2004). In PJM, the data that have been made public regarding economic transmission opportunities also make it clear that reliability investments can have a very signicant impact on transmission congestion and the incentives for transmission investment to reduce congestion costs. As already mentioned, roughly 50 projects have been initially listed in the market window for potential regulated economic transmission investment. Some 32 per cent of these projects subsequently were tagged with the notation reliability upgrade expected to mitigate congestion. One of these projects had 12-month unhedgeable congestion costs of $192 million. The full list is contained in Table 5.7. Two additional projects were designated as beneting from changes in operating practices. One of these projects had 12-month unhedgeable congestion costs of $90 million. These example are, of course, only indicative of the more general observation that so-called reliability transmission investments, as well as discretionary changes in operating practices and study assumptions, can mitigate a lot of congestion that would otherwise emerge on the network well before it is actually revealed. This in turn has implications for the consideration of economic transmission investment models that are driven by the trade-o between transmission investment and the costs of congestion. In particular, for a potential merchant investor, the possibility that reliability-driven transmission upgrades and discretionary changes in operating practices and the implementation of operating reliability criteria will signicantly reduce or eliminate congestion, is likely to be a signicant deterrent to investment that must be supported from congestion rents. This discussion should not be read as implying either that reliability criteria are unnecessary (in Joskow and Tirole, 2006, we explain why operating reliability criteria are necessary due to the threat of network collapses that make reliability a public good) or that they have been set incorrectly. It does imply two things: (a) we need to better understand the economic justication (costs and benets) for these reliability criteria and (b) economic models of transmission investment need to take into account the factors that create a need for administrative reliability criteria and the impacts of reliability criteria that are applied in practice.
8.
CONCLUDING THOUGHTS
Major questions have been raised about whether and how ecient levels of transmission investment can be mobilized in liberalized electricity sectors. Signicant barriers to ecient transmission investment continue to exist in many countries with liberalized electricity sectors. These barriers are primarily institutional rather than fundamental. The experience in England
178
Table 5.7 Examples of transmission congestion mitigated by reliability investments in PJM

Monitored facility LINE 230 KV GRE-POR WYLIERID500 KV TRAN 5 CEDAR BRANCHBU500 KV 5001 BRANCHBU500 KV 5002 NORTH PE LINE 138 KV LAN-MIN LINE 69 KV LEW-MOT2 LINE 230 KV MAR-MRP LINE 138 KV GLA-MTP LINE 69 KV BEC-PAU HUDSON 230 KV HUDSON2 WYEMILLS138 KV AT-2 SICKLER 230 KV SICK #1 LINE 69 KV CED-SAN LINE 69 KV TAL-TRA *Unhedgeable congestion ($) 268,024 6,797,499 5,480,787 192,863,356 3,556,256 1,841,999 383,541 180,726 61,392 1,738,983 536,976 138,865 316,952 592,446 209,335 30,141 Conductor Limit Line trap Transformer Voltage Transformer Transformer Voltage Line trap Stranded bus Wavetrap Conductor Conductor Transformer Transformer Transformer RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade prior to spring of 2004 RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion RTEP reliability upgrade expected to mitigate congestion
Note: *Previous 12 months. Source: PJM Interconnection (2004f).
179
and Wales demonstrates, however, that liberalization does not necessarily lead to depressed levels of transmission investment. The experience in PJM illustrates that regional planning mechanisms and transmission investment criteria can be used eectively to identify transmission investment needs and to price transmission services to provide good locational incentives. The PJM experience also illustrates some of the problems of separating SO and TO functions, vertical integration of TOs with generation, and the bifurcation of regulatory responsibilities between incompatible state and federal regulatory processes. Let me supplement the summary of my conclusions contained in the Introduction to this chapter with the following observations.
Industrial structure Many countries have failed to fully restructure their electricity sectors to support competition. The creation of independent regulated TSOs with system operations, transmission network ownership, maintenance and investment responsibilities with adequate geographic scope is the foundation of ecient operations and investment programs. The full unbundling of transmission service prices subject to a single regulatory regime is a natural complement to the creation of such TSOs. The structure adopted in England and Wales is superior to the RTO structure being promoted in the US. However, both are superior to structures with no ISO at all. Geographic scope TSOs typically span only portions of larger synchronized AC networks. The mobilization of investment for intraTSO transmission enhancements is much better developed than is the mobilization of inter-TSO transmission investments. This was a problem (perhaps not perceived) before liberalization and it is a continuing problem today. In the US, the eort to consolidate control areas under larger RTOs provides one path to reducing the seams problems at the boundaries between TSOs. The creation of a single TSO for Great Britain that covers Scotland, as well as England and Wales, reects a similar motivation. However, there are practical and political limits on the consolidation of TSOs in many countries. This implies that new cooperative mechanisms need to be developed to harmonize reliability criteria, economic criteria, transmission pricing and investment policies, and wholesale market mechanisms to better integrate inter-TSO behavior so as to smooth out the seams as much as is feasible. Regulatory framework Most of the transmission infrastructure that is in place and future investments in it are likely to be governed by some regulatory framework. A clear, credible and transparent regulatory framework that species the TSOs responsibilities, performance
180
norms and regulatory mechanisms consistent with these objectives and performance norms is essential. All regulatory frameworks are imperfect. However, there is no choice but to draw on available experience and regulatory tools to develop and to apply the best feasible regulatory frameworks. A practical regulatory framework will inevitably include a mix of cost-of-service regulation with an overlay of PBR mechanisms based on benchmarking, prot sharing (sliding scale) and ratchets. The development and application of performance norms, formal investment criteria, as well as considerable regulatory judgment is an inevitable component of a sound regulatory process. One component of such a regulatory framework is a transparent regional transmission investment planning process with clear rules for achieving dened reliability and economic goals. The regulatory framework in E&W has many attractive properties. The bifurcation of regulatory responsibilities in the US and the failure to fully unbundled transmission service prices create signicant disincentives to ecient transmission investment. Reliability versus economic investments The liberalization programs in most countries carried along with them the transmission network planning and reliability rules and evaluation criteria from the era of regulated vertically integrated monopolies. Transmission investment activity today is driven almost entirely by reliability criteria. Where did these criteria come from? Why are they the right criteria? Little eort has been made to review these rules and criteria in light of the development of markets that both provide information that can be used to evaluate the costs and benets of these reliability standards and provide market mechanisms that can be used to achieve reliability criteria more eectively. Intra-TSO reliability-driven transmission investments and intra-TSO congestion cost-driven investments are, at the very least, interdependent. At worst the distinctions between them are not particularly useful. Clearly, economic and reliability criteria need to be better integrated into the transmission investment planning and regulatory arenas. Modest steps to do so are now taking place in the RTOs in the Northeastern and Midwestern US. Investment characteristics Transmission investment opportunities involve much more than the construction of major new transmission links. Because many transmission limitations reect contingency limits and associated reliability rules (which should be re-evaluated as noted above), there are often investment opportunities of modest cost that can increase signicantly transmission capacity. The institutions and regulatory mechanisms to identify and undertake these opportunities need more attention. This is especially important in an
181
era when it is dicult to obtain permission to build new transmission corridors. Merchant transmission investment Market-driven transmission investment may be a complement to regulated transmission investment but it is not a substitute. Merchant transmission investment has and is likely to make a very small contribution in the overall portfolio of transmission investment projects that will be made in the future. The eorts to debate its role have been a distraction from more productive initiatives. Wholesale market design Ecient transmission network operation and investment decisions are necessarily interdependent with the design, operation, incentives and price signals generated by the wholesale markets for power and ancillary services. Good wholesale market design, the ecient allocation of scarce transmission capacity, and ecient investment programs go hand in hand and cannot easily be separated. Economic models of transmission investment The simple models of transmission network congestion and investment that are used by economists have little to do with the way transmission investment is actually planned and developed, and the associated transmission services priced within the boundaries of individual TSOs today. Economic models and analysis need to be expanded to better capture the factors that TSOs and regulators consider when they identify transmission investment needs, especially as they relate to the implementation of reliability criteria used for planning and system operations. Economists and network engineers need to develop better ways to work together.
We have made a lot of progress in understanding the challenges associated with stimulating ecient levels of transmission investment in liberalized electricity markets but there is still a lot of work to do.
NOTES
1. 2. 3. 4. 5. Some 63 kV and above. See www.rte-france.com/htm/an/qui/qui_reseau_lignes.htm. I am grateful to Ignacio Prez-Arriaga for reminding me not to forget the costs associated with losses. See Chapter 4 for a discussion of lumpy investments. PJM (2004b, Ch. 6). These links can also support bi-directional economic power trading opportunities. For example, New England typically imports from Quebec over a DC link during the day and exports power to Quebec at night so that Quebecs hydroelectric dominated system can store water at night when prices are relatively low in New England and sell it back during the day when prices in New England are relatively high.
182
6.
Reconductoring with new conductor technology can also increase eective transmission capacity without adding new transmission corridors or towers. Reconductoring with gap-type conductors allowed the company [NGC] to increase the capacity of a critical transmission line by 24 per cent without requiring changes to the transmission towers, Electric Transmission Week, March 14, 2005, p. 14, SNL Energy. There were a few major interregional transmission facilities developed in the US to allow high generation cost areas to access less costly power in remote areas. The Pacic intertie projects (AC and DC) linking the Pacic Northwest and British Columbia with California began to be developed in the 1960s with support from the federal government, federal and municipal power entities (Bonneville and Los Angeles Department of Water and Power) and cooperative agreements with the three vertically integrated investorowned utilities in California. Two HVDC interties were developed in the 1980s to link Quebec with New England. These projects were promoted by Hydro-Quebec (the lowcost power supplier) and were supported by a cooperative agreement involving all of the major vertically integrated utilities in New England (the high-cost power buyers). When vertically integrated utilities took ownership interests in generating facilities outside of their traditional service areas they developed transmission facilities to allow them to gain access to the power generated by these facilities. Most of the transmission infrastructure linking Southern California with Arizona, Nevada and New Mexico was developed in this way, as was the very high voltage network in PJM. For a similar view, see Chapter 4. Merchant opportunities may emerge as well for incumbent TOs seeking to exploit their market power if the regulatory framework permits it. Presentation of Gary Krellenstein, JP Morgan, December 16, 2004, CMU Transmission Conference, Carnegie Mellon University. There is also a six-circuit AC interconnector between Scotland and England. The costs of this interconnector and associated facilities are included in the TOs use of system charges (except that there is a separate charge for use of non-rm capacity above the 850 MW of rm capacity that existed in 1990). The interconnectors capacity is presently allocated using an administrative procedure that involves pro-rata allocations when requests for capacity reservations exceed capacity. After the British Trading and Transmission Arrangements (BETTA) went into eect the assets forming the ScotlandEngland interconnector have been subsumed into the Great Britain transmission system. The regulator is developing new mechanisms to allocate scarce capacity across this interface (Ofgem 2003d). See www.nationalgrid.com/uk/library/documents/sys_04/default.asp?action=&sNode =SYS&Exp=Y. Under the recently enacted reforms, NGCs system operating functions have been expanded to cover Scotland as well (BETTA). However, in Scotland the incumbent vertically integrated companies will remain the transmission owners. There is also an incentive regulation mechanism that governs network losses that involves annual adjustments in the benchmark. The Dutch government recently granted permission to TenneT, the manager of the highvoltage grid in the Netherlands, to nance a regulated transmission interconnection between Norway and the Netherlands, after taking direct economic, reliability and competition considerations into account. Decision on the Application of TenneT for permission to Finance the NordNed Cable in Accordance with section 31(6) of the Electricity Act of 1998, 23 December 2004 (English translation). In theory an independent transco could qualify as an independent system operator or RTO as well, but this would require substantial ownership restructuring in the US context. The APS network (PJMWest) was integrated into PJM in April 2002; the Commonwealth Edison network in May 2004; and the AEP network in October 2004. Virginia Electric Power (Dominion) became part of PJM in 2005. The incumbent regulated transmission owners, all of whom were previously (and most of whom still are) vertically integrated utilities providing generation, distribution and
7.
8. 9. 10. 11.
12. 13. 14. 15.
16. 17. 18.
183
19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29.
transmission services to retail customers (native load) do not actually purchase transmission service under the PJM open access transmission tari to use their own transmission networks to serve their retail customers. Instead they provide the transmission service internally and the associated costs are included (recovered) in the regulated bundled prices they charge to their retail customers. However, they are subject to all of the other terms and conditions of the PJM Tari, PJM Operating Agreement and the PJM Reliability Assurance Agreement. A more detailed discussion of the structure of this transmission tari can be found in Joskow (2005). The $785 million gure covers projects completed since 2000 as well as future projects that are scheduled for completion over the next few years. The rate of investment is signicantly lower than in E&W from 1990 to 2001, though the systems had similar peak loads. New generator deliverability criteria were recently proposed. All interconnection studies performed through PJMs RTEP Process can be found on the PJM website: www.pjm.com/planning/rtep-baseline-reports/baseline-report.html. FERC Order 2003, Standardized Generator Interconnection Procedures, July 23, 2003. For a detailed discussion of the procedures that were recently adopted by PJM, see PJM FERC Filing in Docket Number RT-01-2-01, dated April 21, 2004, www.pjm.com, accessed June 15, 2004. PJM congestion spreadsheet downloaded from www.pjm.com on December 4, 2004. Unlike the New England ISO, PJM has refused to include congestion forecasts in its planning process. PJM FERC Filing in Docket Number RT-01-2-01, Appendix 1, dated April 21, 2004 and PJM market window spreadsheet downloaded December 4, 2004. Available on the PJM website www.pjm.com. See www.pjm.com/planning/project-queues/merch-queue-o.jsp accessed on May 27, 2005 and Developer Argues Exelon-PSEG projects could disrupt role of merchant transmission, Electric Transmission Week, May 30, 2005, pp. 910. In E&W an unknown portion of additional transmission investments or planned reliability investments that were moved forward to an earlier date were driven by the annual SO incentive scheme. As previously discussed, PJM has adopted a new framework for regulated economic investments.
REFERENCES
Chao, H.-P. and R. Wilson (1987), Priority service: pricing, investment and market organization, American Economic Review, 77, 899916. Henney, A. (1994), A Study of the Privatisation of the Electricity Supply Industry in England and Wales, London: EEE Limited. Hirst, E. (2004), US transmission capacity: present status and future prospects, Edison Electric Institute, Washington, DC, June. ISO New England (2004), Regional Transmission Expansion Plan (RTEP04) Technical Report, October 21. Joskow, P.L. (1988), Price adjustment in long-term contracts, Journal of Law and Economics, 31, 4783. Joskow, P.L. (2005), Transmission policy in the United States, Utilities Policy, 13, 95115. Joskow, P.L. and R. Schmalensee (1983), Markets for Power: An Analysis of Electric Utility Deregulation, Cambridge, MA: MIT Press. Joskow, P.L. and R. Schmalensee (1986), Incentive regulation for electric utilities, Yale Journal on Regulation, December, 149.
184
Joskow, P.L. and J. Tirole (2000), Transmission rights and market power on electric power networks, Rand Journal of Economics, 31 (3), pp. 45087. Joskow, P.L. and J. Tirole (2005a), Retail electricity competition, Rand Journal of Economics, (forthcoming), http://econ-www.mit.edu/faculty/download_pdf.php? id=918. Joskow, P.L. and J. Tirole (2005b), Merchant transmission investment, Journal of Industrial Economics, 53 (2), 23364. Joskow, P.L. and J. Tirole (2006), Reliability and competitive electricity markets, Rand Journal of Economics, (forthcoming), http://econ-www.mit.edu/faculty/ download_ pdf.php?id=917. Laont, J.-J. and J. Tirole (1993), The Theory of Incentives and Procurement in Regulation, Cambridge, MA: MIT Press. Newbery, D. (2004), The benets of electricity deregulation: Europe, presentation, MIT Center for Energy and Environmental Policy Research, Cambridge, MA, December. NGC (2004a), The Statement of the Use of System Charging Methodology, Eective from 1 November. NGC (2004b), Interim Great Britain Seven Year Statement, November. Ofgem (2003a), Annual Report 200203, 14 July. Ofgem (2003b), NGC System Operator Incentive Scheme from April 2004, Initial Consultation Document, December. Ofgem (2003c), ScotlandEngland Interconnector: Access Criteria from 1 April 2004, Consultation Paper, December. Ofgem (2004a), Proposals for the Amendment of the Licensing Application Regulations, (Interconnectors) Consultation Document, June. Ofgem (2004b), Electricity Transmission Network Reliability Incentive Schemes, Final Proposals, December. PJM Interconnection (2003), PJM Economic Planning Implementation Stakeholder Process Meeting, August 29. PJM Interconnection (2004a), PJM Economic Planning Implementation Stakeholder Process Meeting, February 10. PJM Interconnection (2004b), State of the Market Report 2003, March 4. PJM Interconnection (2004c), Open Access Transmission Tari, updated to May 6. PJM Interconnection (2004d), Amended and Restated Operating Agreement, updated to June 15. PJM Interconnect (2004e), Transmission Expansion Advisory Committee Meeting, June 21. PJM Interconnection (2004f), Regional Transmission Expansion Plan, July, www.pjm.com, December 14. PJM Interconnection (2004g), Generator retirement analysis, presentation, Transmission Expansion Advisory Committee, undated. Sweeting, A. (2000), The wholesale market for electricity in England and Wales: recent developments and future reforms, MIT Center for Energy and Environmental Policy Research, WP-2000-007, November. US Energy Information Administration (EIA) (2004), Electricity Transmission in a Restructured Industry: Data Needs for Public Policy Analysis, December. Wolfram, C.D. (1999), Measuring duopoly power in the British electricity spot market, American Economic Review, 89 (4), 80526.
PART III
Coordination between investments in generation and transmission
6.
Long-term locational prices and investment incentives in the transmission of electricity

Yves Smeers*
INTRODUCTION
1.
The economic principles that rule capacity expansion under constant return to scale are well known (for example, Crew et al., 1995). One invests so as to equalise short- and long-run marginal costs. If these convenient assumptions prevailed in the transmission of electricity, one would invest so as to equalise the marginal transmission capacity cost with marginal congestion and loss costs. This is not possible. The transmission of electricity suers from many undesirable economic properties that make the direct application of these principles impossible. It combines both economies of scale and lumpy investments which render the denition of longrun marginal cost illusory. It is also plagued by considerable externalities. Because the external benets of transmission investments are not appropriable by those who are at the origin of these investments (Bushnell and Stoft, 1996), there is no hope of fully internalising these externalities. The objective of this chapter is to present a model of network investments in a competitive electricity system based on the idea of long-term locational signals introduced in the European Regulation 1228/2003 on Cross-border Exchanges in Electricity, hereafter referred to as the Regulation (European Parliament and Council, 2003). We conduct this analysis under the following drastic simplifying assumptions:
The architecture of the market essentially consists of spot energy and transmission markets. These are coordinated by the transmission system operators (TSOs). Such architectures have been extensively explored in the literature and are now commonly implemented. Current proposals of the association of European Transmission System Operators (ETSO) and power exchanges (Europex) also go in that direction (see Section 2).
187
188
Coordination between investments
We ignore the distinction between day-ahead and real-time markets and assume only one spot market that integrates both. This simplication is acceptable if the day-ahead and real-time markets are well arbitraged, a currently heroic assumption in the European context. We justify this assumption by our objective to concentrate on the mix of long- and short-term locational signals covered by the Regulation. This requires that the detail of short-term operations be somewhat disregarded. We assume no market power and disregard contracts. The reader will thus nd no reference to physical and nancial contracts, or bilateral and centralised markets. This simplication is important because these contracts inuence the incentive of agents in imperfect competitive markets. This inuence disappears if we assume that there is no market power, as we do in this chapter. The contract issue should be taken up as soon as one departs from the assumption of pricetaking agents. We assume perfect information. This implies, in contrast with the assumption underlying the new theory of regulation, that the regulator perfectly knows the demand functions of the consumers as well as the production sets and costs of both the transmission and generation activities. If reliability considerations need to be invoked, then all agents foresee the same contingencies with the same probabilities of occurrence.
Except for taking network indivisibilities explicitly on board, we essentially make all the assumptions of the perfect competition model that we particularise, as described above, to accommodate some idiosyncrasies of the electricity system. We then concentrate on network indivisibilities by adapting and slightly generalising ONeill et al. (2004), which deals with indivisibilities in the unit commitment problem. The chapter is organised as follows. Section 2 introduces the notion of locational prices in the context of the Regulation. Section 3 casts the chapter in the economic literature on electrical networks. Section 4 presents a reinterpretation of ONeill et al. (2004) in terms of network pricing. Section 5 relates the interpretation of these results to notions appearing in the Regulation, namely cost reectiveness and non-discrimination. Section 5 also presents a transposition to the network problem of comments made by Hogan and Ring (2003) on ONeill et al. Section 6 extends the model to address cost causality. This requires an extension of ONeill et al.s Theorem 2: it is presented in the Appendix, using the theory of conic duality. Section 7 takes up the question of non-discriminatory prices. It indicates that contrary to the common wisdom underlying the European
Long-term locational prices and investment incentives
189
debates preceding and following the introduction of the Regulation, nondiscriminatory prices may entail welfare losses in a decentralised market with indivisibilities. Section 8 briey discusses institutional matters. Section 9 lists remaining questions that should be addressed in future work. The conclusions in Section 10 can be summarised as follows: long-term locational prices are fraught with diculties that are largely unexplored. Their introduction in the Regulation may have helped to achieve a political compromise. The task remains to make it workable.
2.
INVESTMENT, LOCATIONAL SIGNALS AND REGULATION
The concept of long-term locational prices as a signal for guiding investment and location of generation plants in restructured electricity markets appears in the Regulation on Conditions for Access to the Network for Cross-border Exchanges in Electricity that came into force in July 2004. This law is essentially an outgrowth of the work of the Florence Regulatory Forum,1 set up by the European Commission to nd means to facilitate electricity exchanges between Member States in the so-called internal electricity market. The Regulation contains ve parts: 1. 2. Articles 1 and 2 set the scope of the law and introduce important denitions. Articles 3 and 4 state the conditions for accessing the network. Article 3 presents a system of cross-border compensations between TSOs. Article 4 introduces the notion of locational prices, which is the main focus of this chapter. Articles 5 and 6 discuss congestion management at the interconnections. Article 5 makes general informational recommendations. Article 6 requires that congestion management be market based. Article 7 presents conditions for exemption of new interconnectors from the general rules. Finally, Articles 8 to 15 elaborate on procedural but crucial matters and in particular the role of Comitology (Article 13).
3.
4. 5.
This chapter concentrates on long-term locational prices that it considers together with congestion management. The Regulation requires that locational prices be ecient, cost reective and non-discriminatory (Article 4, paragraph 1). At the time of the adoption of the Regulation (June 2003), none of the documents elaborated by the Florence Regulatory Forum had proposed any method for computing locational prices that would satisfy
190
these three criteria. The Forum and related studies (for example, PrezArriaga et al., 2002) certainly looked at long-term locational signals. But they did so in purely accounting terms, that is, by allocating network costs among network users. Cost allocation methods reect costs only in a very weak sense that does not imply any causality: agents are not charged for the cost that they cause; they only collectively pay for the total costs incurred. Cost-allocation methods are also discriminatory in an economic sense because they do not rely on marginal costs. The fact that the marginal cost of the electricity network is dicult to dene, makes the combination of these two criteria dicult if not impossible to implement in a non-ambiguous way. Finally, cost-allocation methods are not economically ecient in the sense of inducing investments in the right location (see Curien, 2003; Green, 2003; and Lvque, 2003). The lack of economic eciency is particularly dangerous in the restructuring context: it distorts the incentive to invest and hence endangers the security of electricity supply. Lack of ecient locational signals and economic incentives to invest did not matter much in the old regulatory days. The regulatory obligation to satisfy demand implied some so-called optimal mix of capacities and locational decisions (taking constraints such as site adequacy and environment into account). This is no longer true in a competitive system where long- and short-term locational prices are the sole market instruments that can induce both the right mix of equipment and their proper location. It is thus of the essence that the prices be right and produce the good incentive to invest. Unfortunately, economic theory tells us that allocated costs oer no guarantee in that respect. The absence of any precise reference to ecient long-term locational signals in the work of the Florence Regulatory Forum should not come as a surprise. We do not know at this stage how to construct ecient long-term locational signals, let alone ecient, cost-reective (in the strong sense of cost caused) and non-discriminatory long-term price signals. The reason is simple: a generator locates on some site or it does not. Location and the choice of a plant type are discrete decisions, and we know very little about how to induce the right discrete decisions through prices. It is indeed a basic principle of economic theory that discrete decisions are dicult if not impossible to drive through price mechanisms because of non-convexity phenomena (see Scarf, 1994, for a clear statement of the problem and Prez-Arriaga and Smeers, 2003, for a discussion of the question in electricity networks). The problem becomes particularly acute if one notes that the impact of these decisions generally covers a time period of several years. While long-term locational signals are a rather unexplored area, shortterm locational signals for dealing with congestion are a well-known but highly controversial subject in Europe. These are treated in Articles 5 and 6 of the Regulation. Congestion management involves continuous decision
191
variables so that one understands much better how to decentralise through a market process. The consequence is that dierent market-based congestion methods exist and others are proposed. It is certainly surprising that the Regulation recommends introducing long-term locational prices that we do not know how to construct and implement but avoids suggesting well-understood short-term locational prices that have proved eective in several systems.
3.
RELATED LITERATURE
Dierent approaches to grid investments exist in the literature. They are discussed in Roselln (2003), which puts our work in perspective. We refer the reader to that survey for further information and a guide to the literature. A global and in-depth analysis of investment in transmission can be found in Woolf (2003). Hogan (2002 and 2003) and Pope and Harvey (2002) extended the theory of nodal prices and nancial transmission rights (FTRs) to the problem of investments in the grid. Their objective is to provide a market-driven mechanism for expanding the grid. This analysis underlies the notion of merchant lines found in the Federal Energy Regulatory Commission (FERC) standard market design (SMD) proposal, which eventually also found its way into the Regulation (Article 7). We discuss neither merchant lines nor their compatibility with the rest of the Regulation but instead focus on how our model relates to Hogan, Pope and Harveys theory. The basic idea of these authors is that it is possible to decentralise (at least some) investments in the grid to economic agents (generation companies, consumers, investors and so on) provided that these receive long-term FTRs that guarantee the payment of congestion rents over the life of the project. The system operator (SO) grants the long-term rights in an auction. Because investments in new lines can destroy existing transmission rights, some restrictions on the allocation of the long-term rights are necessary to keep the set of granted rights physically feasible for the network. This latter process is rather complex (Hogan, 2002) but the bottom line is that congestion rents remunerate the investors who invest in the network. Joskow and Tirole (2005) argue that the theory suers from several shortcomings and Hogan replied to some of these comments (Hogan, 2003). We shall not elaborate on these discussions, which can be found in summarised form in Rosellns paper, but concentrate on one of the points unveiled in Joskow and Tirole (2005) and clearly recognised in Hogan (2002 and 2003), namely investment lumpiness. FTRs were primarily introduced as hedges against random congestion charges (Hogan, 1992). FTRs are forward contracts on spot transmission
192
prices. They are special forward contracts though, in the sense that they need to be traded through the TSO in order to guarantee that they are compatible with the physical possibilities of the network. Under the usual noarbitrage assumption, FTRs satisfy the standard property derived from nance theory that the forward price is the expectation, in some riskneutral probability, of the congestion price. In other words, the economic signal embedded in long-term FTRs contains only congestion charges. In order to see this more clearly, consider a deterministic world. There is no need to hedge and an FTR is then exactly equivalent to a payment of the congestion charge in real time. Resorting to long-term FTRs for inducing investments in the grid is then equivalent to using congestion charges as incentives to invest in the grid. Investment lumpiness limits this potential incentive. We elaborate in the following on the consequences of this point and argue that, notwithstanding investment lumpiness, capacity expansion of the grid can still be decentralised provided that one introduces more prices than the sole congestion charges. In Section 8 we shall return to the interpretation of that result in terms of Hogans theory. The transco model is a second approach to investments in the grid.2 A transco is a company that builds and operates the network for prot. In the terms of this chapter and taking the concepts of the Regulation on board, the transco is remunerated with both long- and short-term locational charges. It then develops the grid by solving a network capacity expansion problem with a view to maximising its prot. The network is a natural monopoly that gives market power to the transco. If not regulated, the company will set long- and short-term charges in a way that maximises its prot but degrades welfare. The question is thus to nd charges that provide the incentive to be ecient and permit the rm to recover its costs. This has dierent aspects. Perhaps more than any other regulatory question, investments in the grid have a strong avour of asymmetry of information. The electric network is indeed a complex technology that may be dicult to grasp from outside the industry. The problem of incentivising the transco will then require the regulator to oer a menu of contracts. Models of this type usually assume an explicit demand function for the services of the regulated company (for example, Vogelsang, 2001). For the sake of analytical convenience, these analyses also resort to simplications that neglect the multi-product nature of transmission services. We depart from these considerations and see the transco as an ordinary prot-maximising rm. We also consider that the demand for transmission services is not explicitly given by a demand function but derived from the interactions between prot-maximising generators and consumers. Last, we also assume an electricity network that oers dierent point-to-point services. Our objective, in following that path, is
193
not to overlook the important problem of the market power of the transco, but to attempt to construct a partial equilibrium model with price-taking agents that we intend to expand in future developments. Our objective can be stated as follows: consider a world of perfect information with price-taking agents (including the transco). The question is to nd the price signals that induce the transco to invest in an ecient network and allows one to recover the cost of building this network. The need for adequate price signals for a transco derives from the now well-recognised fact that a (price-taking) transmission company that receives only congestion charges and an additional xed revenue has an incentive to underinvest in order to increase congestion. It is also admitted that the sole recovery of congestion charges allows the transco to recover only a relatively small fraction of the total costs of the grid (Prez-Arriaga et al., 1995). One can thus expect to see multi-part taris emerge as necessary instruments for achieving both the eciency and cost-recovery objectives. Non-linear taris are commonly encountered in the regulatory literature for dealing with asymmetry of information. They also appear in non-convex economies (see Bjorndal and Jrsten, 2004, for a recent treatment through dual price functions). This latter context is the one adopted in this chapter. In short, and possibly in contrast with the existent literature on transcos, we disregard the question of market power, but concentrate on the existence of price signals that induce a price-taking transco to manage congestion and develop the network in an ecient way even though investments are lumpy. We do this by constructing a model that fully accounts for the multi-product nature of the transmission infrastructure, and develop an equilibrium framework where the demand for transmission services results from other agents actions. This was the approach formerly adopted in studies of operations and market design. We hope to pave the way to similar developments in two directions: on the engineering side, we introduce the capability to accommodate realistic grid representations; on the economic side, we derive price-taking equilibrium conditions susceptible of being extended to accommodate market power.
4.
A MODEL OF LOCATIONAL AND CONGESTION PRICING
The mix of discrete decisions (that one does not known how to decentralise through prices) and continuous decisions (that one knows very well how to decentralise through prices) found in the electricity grid is not unique. It also appears in another area of electricity restructuring, namely unit commitment and plant dispatching. As mentioned above, Hogan and
194
Ring (2003) and ONeill et al. (2004) address the question of nding price signals for inducing plants to start up and shut down. We transpose and slightly extend their reasoning to the questions of locational and congestion pricing. The Grid Equilibrium Model The following primal mixed-integer program (PIP) was considered by ONeill et al. (2004) for studying price-based decentralisation of start-up decisions in unit commitment problems: max s.t.
k
PIP
ck xk Ak2zk
dk zk b0
(6.1) (6.2)
Ak1xk
Bk1xk xk 0,
Bk2zk
bk
(6.3) (6.4)
zk {0, 1}n(k).
This model is a standard mixed-integer program.3 It is easily interpreted in the context of optimal centralised investments in a regulated electricity system. We accordingly assume a system where investment decisions in the grid and in power generation are under the supervision of a benevolent, perfectly informed and perfectly rational (with unlimited computational possibilities) regulator. Let j 1, . . . , J denote the nodes of the electrical network. We cast ONeill et al.s model in the locational pricing context through the following interpretation.

Let k 1, . . ., K designate an agent, consumer or generator, that decides to connect to the network. zk is the vector of binary variables representing locational decisions of agent k. A component of zk, let zkj, is equal to 1 if k connects to the network at location j 1, . . ., J. It is zero otherwise. All components of zk are 0 if agent k does not connect to the network. The decision to connect to the network is a simplied view of a more complex decision: an agent, whether a consumer or a generator, connecting to the network indeed builds a plant of a certain type, a feature that is not represented yet in the variable z. From a welfare point of view, the decision to connect therefore also implies the decision to build
195
and hence a cost. We limit ourselves in this section to the sole locational decisions. We shall later introduce dierent technologies, i 1, . . ., I, and extend the denition of zk by considering zkij 1 when agent k connects a plant of technology i to location j of the network. xk is the vector of power injection/withdrawal of generator/consumer k. There are potentially as many components of xk as there are nodes in the network. We replace xk by signkxk in ONeill et al.s expressions (6.1) and (6.2) in order to distinguish between injections and withdrawals. In this updated expression, all xk are assumed positive and signk is a diagonal matrix whose components are 1 for injections and 1 for withdrawals; signkxk is thus the vector of the net nodal withdrawals of agent k. dk is the vector of xed costs or benets before any payment for locational charges incurred by agent k when it connects to the network. dkj is thus the xed cost of building a plant of (currently) unspecied technology in location j. Alternatively dkj can be interpreted as a vector of xed benets or costs accruing to a consumer k when it connects to the network in location j. ck is the vector of the marginal costs of the generators. Alternatively, ck is the marginal utility of the consumer k. Both marginal costs and utilities are constant as in ONeill et al.s model. A slight complication of the notation makes it possible to approximate non-linear concave utilities and convex costs. Given the interpretation of the x variables, relation (6.2) represents the load ow-based constraints that limit the injections and withdrawals in the network. Ak1 is thus identical for all generators and consumers k and can be noted A. The matrix A consists of two parts. One row of the matrix expresses the balance between injections and withdrawals in a lossless network. This is written: 1
k
signk xk
0,
(6.5)
where 1 is a row vector of 1. The other rows of A and the associated constraints express the limitations on the injections and withdrawals imposed by thermal limits on the lines. These are expressed using the negatives of the power transfer distribution factors (PTDFs) that give the ow in each line of the grid as a result of the injections and withdrawals signkxk. We keep the inequality formulation from ONeill et al., even though relation (6.5) is an equality. This facilitates the notation and can easily be justied.
196

Ak2 is taken as 0 in this application and will no longer be used in this section. The rst component of b0 is the right-hand side of (6.5) and is thus zero. The other components of b0 are the thermal limits on the lines of the network (that are written in only one direction in order to simplify notation). Thus (6.2) represents the usual DC load ow approximation of network ows. Bk1 is an identity matrix and Bk2 a diagonal matrix of the capacities (noted m) installed by generator/consumer k; bk is identically 0.
Using these notations, PIP can be restated as program PIPLOC: max s.t. A
k
PIPLOC
ck signk xk

dk signk zk

(6.6)
signk xk mkzk
b0 0
(6.7) (6.8) (6.9)
xk xk Extensions and Restrictions 0,
zk {0, 1}n(k).
This program involves both investment (z) and operation (x) variables. One immediately notices that PIPLOC boils down to the usual dispatch/welfare optimisation problem extensively studied in the congestion management literature when the zk variables are xed. This reduced problem is the basis of the study of short-term locational signals. PIPLOC therefore embeds the question of congestion management covered in Articles 5 and 6 of the Regulation. The aim of this chapter is to free the zk variables so as to also cover the long-term signals of Article 4. The model can easily be extended to accommodate several time segments and contingencies. Let xk be the power generation/withdrawal of generator/consumer k in time segment or contingency . The model can be mp rewritten as program PIPLOC(where mp stands for multi-period) max s.t.
mp
PIPLOC
ck signk xk

dk signk zk

197
A
k
signk xk mk zk 0, 0
b0
xk xk
zk {0, 1}n(k).
As mentioned above, the model can also be easily extended to allow zk to represent the decision to locate at a certain node and to build a certain type of capacity. This extension only makes sense in a multi-period (time segment) model. This model would thus consider variables zkij and zkij where i is the type of plant. In order to simplify the discussion and to facilitate the establishment of the correspondence between this chapter and ONeill et al., we rst limit ourselves to the simple formulation (6.6) to (6.9). Program PIPLOC should be interpreted as the problem to be solved by a regulator operating under the following ideal conditions: 1. Perfect information: the regulator knows the marginal willingness to pay of the consumers and the marginal cost of the generators. The regulator also knows the locational xed costs (the xed cost of building a plant of given technology at some location). Perfect competition: agents are price takers. Perfect congestion management: congestion is managed by nodal prices. Simplied electrical assumption: the network is lossless and its structure is described by a PTDF matrix.
2. 3. 4.
Assumptions (6.2), (6.3) and (6.4) are common in studies of restructured electricity systems. Assumption (6.1) may look harsh, but is not more stringent than the implicit assumption that prevailed in the former regulatory regime where the regulator was also assumed to decide or accept investments according to a perfect cost-minimisation model. A main dierence is that the regulator of the former monopoly regime did not have to interpret the results of this optimisation model in terms of locational prices. The regulator could be wrong with less damaging consequences. It is on these prices that we concentrate, assuming all the above simplifying assumptions. Locational Charges ONeill et al. suppose that one can solve problem PIPLOC and we follow suit. The assumption is not unrealistic given the astonishing numerical progress made these last 20 years in mathematical programming in general and mixed integer programming in particular. Suppose this has been done
198
and let zk k be the optimal location of the generators/consumers, we state a primal linear program PLIP(z): max
k
ck signk xk

dk signk zk

A
k
signk xk mk zk 0
b0
(u0 ) k
(6.10) (6.11) (6.12)
xk
(uk ) k
zk
zk xk
(wk ) 0,
where u0, uk and wk are the dual variables of the constraints (6.10), (6.11) and (6.12), respectively. The dual DLIPLOC(z) of that problem is written as: min u0b0 u0A signk uk mk u0 0, uk uk wk wkzk k k k. (6.13) (6.14) (6.15) (6.16)
ck signk dk signk
0, wk unconstrained
Note from the above discussion that signk in relation (6.14) multiplies the nodal components of u0A by 1 or 1 depending on whether agent k is a generator or a consumer at that node. The dual variables of the problem can then be interpreted as follows:
The rst component of u0 is the price of electricity at the hub node. The other components of u0 are the values of the capacities of the lines (owgates). These components of u0 are indeed the dual variables of the thermal capacities of the lines and should be interpreted in the same sense as in usual discussions of congestion management. One can easily verify that this implies that u0A is the vector of the nodal prices of electricity. uk is the vector of scarcity premium on tight generation and consumption capacities. The constraints (6.14) express that the nodal price at some active generation node (the corresponding nodal
199
component of u0A) is equal to the variable cost (the corresponding nodal component of ck) of a generator at that node which operates at a positive level, but below capacity. This price is equal to the variable cost plus a scarcity premium if the generator operates at full capacity at that node (note that signk 1 implies (u0A)k uk ck). A similar interpretation holds for consumers. wk are locational prices in the sense that a generator/consumer k pays wkj to locate at j. Note that dk are costs for generators. A constraint (6.15) expresses that the locational price is greater or equal to the capacity margin minus the xed cost of the plant. The equality holds for a plant that is eectively located at that node. In that case the locational price and the capacity margin pay for the cost of locating the plant at that node. Note that wk is unconstrained: its components can be positive or negative. If negative, a component of wk should be interpreted as a payment to be given to the agent to locate.
Eciency Properties of Locational Charges We now transpose ONeill et al.s result in the interpretation of locational and congestion pricing. Denition Ecient short- and long-term locational prices are prices u0 (congestion) and wk (location) such that agents (generators and consumers) when charged (u0, w) behave eciently in the following sense. Suppose that these agents maximise their prots under the sole capacity constraints: xk mk zk 0.
In other words, agent k solves the following prot-maximisation problem: max(ck s.t. xk xk 0, mk zk 0 (6.18) (6.19) u0A) signkxk (dk signk wk )zk (6.17)
zk {0, 1}n(k).
Then the solutions x k, z k that they obtain are those desired by the regulator who solves PIPLOC. Moreover the injections and withdrawals x balance and comply with the network constraints:
200
A
k
signk x k
b0.
(6.20)
Note that this denition refers only to economic eciency. It corresponds to the notion of market-clearing price in ONeill et al. but does not embed the notions of cost reectiveness or non-discrimination also foreseen by the Regulation. We examine these notions in subsequent sections of this chapter. Proposition 1 Suppose that the regulator successively solves problems PIPLOC and PLIPLOC(z) where z is the solution to PIPLOC . Let u0 and wk be part of the optimal dual solution of PLIPLOC(z). Then u0 and wk are respectively ecient short- and long-run locational prices. Let x k and z k be the decision of agent k. One has (ck u0 A) signk x k (dk signk wk ) z k 0.

Proof The proposition is a direct application of Theorem 2 in ONeill et al., which is recalled in generalised form in the Appendix. In this system, the regulator or the system operator receives the congestion charges u0 A k signk x k and the location charges k wk z k. Because the statement of the problem does not contain any information on the cost of the network, Proposition 1 contains no result on the balance of the TSOs budget. Similarly, there is no information on how the cost of the network depends on the locational decision of the agents; hence the result cannot express any cost causality. It is remarkable that generators that decide to locate and produce make a zero prot. Similarly, consumers that decide to locate and consume have a zero net welfare gain. This is the standard result of a perfectly competitive equilibrium market which is recovered here by the application of ONeill et al.s Theorem 2. All other agents would incur a loss if they were to locate on the network and generate/consume electricity.

5.
DISCUSSION
The above congestion and locational prices satisfy one objective of the Regulation, namely economic eciency. They leave the decision to locate, generate and consume to the agents who pay these prices. The prices wk and u0 are also quite distinct. The payment wk results from the decision to locate on the network and can be interpreted as the long-term signals of Article 4 of the Regulation. These payments must be made irrespective of the consumption or generation levels, provided that the decision is made to
201
connect to the network. The nodal prices u0A entail congestion charges that are directly proportional to consumption and generation. They comply with the obligation to use market-based congestion management methods of Article 6 of the Regulation. Besides oering the two price signals required by the Regulation, u0A and w also form true two-part taris in the sense commonly understood by economists. As in ONeill et al., the approach extends the commodity space beyond the sole energy space. Both energy and the locational rights are priced. One should note that it is of the essence not to convert the xed part wk of the tari into a proportional charge that would come on top of u0A. Doing this would indeed convert the two-part tari into two single-part linear prices, which cannot sustain the equilibrium in the presence of indivisibilities. This important remark appears neither in the regulation nor in the allocation mechanisms discussed by the TSOs and regulators in the Florence Forum. While Proposition 1 of Section 4 states that these non-linear taris satisfy the eciency objective of Article 4 of the Regulation, none of the other desired properties (cost reectiveness and non-discrimination) required by the Regulation is achieved. This is not specic to this construction as twopart taris are often discriminatory in two senses. By construction they lead to average prices that are decreasing with quantity. This second-degree (for example, Tirole, 1998) price discrimination is accepted by Courts. Two-part taris are also rst-degree (ibid.) discriminatory prices. They dier by agent in the sense that, except for their proportional part which is equal to the marginal cost, the xed part does not reect the cost incurred by the generator but the willingness to pay of the agent. This is unlikely to be accepted by competition authorities without further justication. We therefore elaborate on this discrimination:
The wk are rst-degree discriminatory in the sense that two generators that locate at the same site may not necessarily pay the same locational price. The only possible justication of that discrimination is that, in contrast with the common wisdom of the Regulatory Forum, price discrimination is sometimes necessary in order to achieve economic eciency. This is what happens here: the signals are much more than simply locational; they are individual signals, targeted at each candidate in each location. The wk are not cost reective. They bear no relation to the cost of the network which is completely absent from the statement of the problem.
These prices also lack other important desirable properties. First, there is no guarantee that the sum of the payments accruing from congestion and
202
long-run locational prices recovers the cost of the network, let alone induces an optimal development of the network. Second, the locational prices do not reect the incremental cost that the TSO would incur because of the decision of an agent to locate at some node of the network. This is not surprising as the model PIPLOC does not contain any representation of the cost of the network. In short, eciency is achieved at the cost of sacricing non-discrimination, cost reectiveness and other desirable properties. One may thus search for a more involved tari structure that exhibits more components, some of them related to the cost of the network. Before turning to this point, it is useful to review Hogan and Rings (2003) comments on ONeill et al.s model and to see whether they can be transposed to the problem of locational and congestion prices. Hogan and Ring (2003) note that there may be many wk capable of supporting the dispatch equilibrium in a unit commitment problem. We want to show that the same holds for the locational equilibrium. To see this, recall rst that uk is the scarcity rent captured by the optimally located plants when operating in a perfectly competitive system. There may be several uk but take one of them. Transposing Hogan and Rings reasoning to locational pricing, we rst consider the minimal payment wkj (u) that should be paid to agent k in location j in order to induce a decision compatible with the z desired by the regulator. Using Hogan and Rings notation and considering the case of a generator as an example, one denes: Pkj Pkj max( dkj dkj zkj ukj mkj zkj {0, 1}. (6.21) (6.22)
ukj mkj )zkj,
Pkj is the prot collected by agent k at location j if it behaves according to the regulators plan. This prot consists of a capacity scarcity margin ukj mkj minus the incurred xed cost dkj zkj. kj is the prot that the same agent would collect if it made its own decision to invest in location j on the basis of the expected capacity scarcity margin ukj and its investment cost dkj. The prot kj is what would be driving the generator locational decision in a pure nodal system. Dene:

wkj (u)
min(0; Pkj
kj ) .
(6.23)
Suppose wkj (u) is strictly negative, then one can verify that a payment of wkj (u) to generator k at j) compensates this agent for:

203
the opportunity cost kj of not following its own optimal strategy; the cost of following the strategy of the regulator.
Any higher payment would overcompensate the agent. Conversely, one may also be interested in computing the maximal positive charge that can be levied on the generator that decides to invest before it modies its decision. Dene: wkj (u) .
kj
(6.24)
One can verify that a levy wkj wkj will not induce a generator k that spontaneously decides to invest in location j to change its decision. One can thus imagine that a regulator could try to nance (part of) its payments wkj (u) from charges wkj (u) . More generally, let KJ and KJ be, respectively, the pairs (k, j) for which wkj and wkj have been dened. A w supports the locational equilibrium if: w kj w kj This implies that: wkj (u) wkj (u) (6.27) wkj wkj (k, j) KJ (k, j) KJ (6.25) (6.26)
(k,j) KJ
(k,j) KJ
is the minimal net payment that the regulator must be prepared to make in order to induce the dierent agents to abide to its locational strategy z. Nothing guarantees that this minimal payment is negative, that is, that the regulator will break even. Any decit should be covered otherwise. A similar reasoning applies to the consumers with a like conclusion. The overall conclusion in terms of covering revenue requirements is more positive as soon as one considers generators and consumers together. Supposing that the total surplus of the market, after paying the cost of the network, is positive (a very mild assumption), then it is always possible by playing on the w according to relations (6.25) and (6.26) to satisfy the revenue requirement. The drawback is that this will increase price discrimination. The obligation for the regulator to break even may thus lead one to increase the discrimination between agents with respect to the original w. This suggests introducing constraints on the w that are directly inspired by commonly found regulatory objectives. One may for instance wish to achieve:
204
a budgetary balance. This would imply: w km zkm 0 (6.28)
or w km zkm Fixed Network Cost. (6.29)
Non-discrimination constraints. These impose some equalities between the negative w kj and/or other equalities between the positive w kj.
Needless to say there is no guarantee that imposing these regulatory constraints on top of (6.25) and (6.26) will lead to feasible w k. However, it is always possible to try to get close to regulatory objectives by minimising the violation of the constraints. We shall not elaborate any further on these variations in this section and return to the problem of cost causality.
6.
COST CAUSALITY AND NON-LINEAR LOCATIONAL PRICE
Section 5 concludes that the wk derived from problem PLIPLOC are discrimatory and not cost reective. This is not surprising. Non-discriminatory prices are well known not to be able to support perfect competition equilibrium when there are indivisibilities and economies of scale. The absence of cost reectiveness could also be expected as the model PIPLOC used to compute the wk does not link agents locational decisions to the grid structure and hence to the grid costs. Also, note that the above tari prices only energy and location, while taris that include a capacity charge and hence a third component in the tari are common in practice. The following develops a richer set-up with the view of introducing cost causality into the model. Representation of the Network Consider an extension of problem PIPLOC, constructed as follows:
There exists a set of possible network congurations n 1, . . ., N each of cost d n and PTDF matrix PTDFn. The TSO selects one of 0
205
n these congurations. Let z0 be an N-dimensional vector with z0 1 if the TSO selects the nth network conguration. The locational price takes the form of a two-part tari, namely a xed charge for accessing the network and a proportional charge for reserving capacity (a capacity or demand charge). This reserved capacity can be a single variable valid for the whole year, or a vector if one implements seasonal reservations, for example, summer and winter (see Prez-Arriaga and Smeers, 2003). We rst work with a single annual capacity reservation in order to facilitate the discussion. With this interpretation xk plays both the role of energy generated or consumed and reserved capacity. One introduces some causality between the decision to locate, the reserved injection and withdrawal capacity and the network structure. This causality is complex in the real world. Strictly speaking, it derives from a network expansion planning problem. Because our methodology uses optimisation-type techniques to explore economic problems, we adopt an approach theoretically justied in optimisation terms but practically still to be elaborated and represent causality through two types of constraints:
(a)
The skeleton network constraints: F0z0 Fk 0zk 0 (6.30)
(b)
imply that the decisions of the agents to locate (the zk, k 0) impose a certain minimal structure on the network. This results in a set of allowed network structures z0 (there may be several z0 for a given set of location decisions zk). The incremental network constraints: G1k xk G20z0 0 (6.31)
k 0
imply that the capacities reserved by the dierent agents (the xk, k 0) further restrict the set of allowable z0 of the network: The selection constraint:
N n 1
zn 0
(6.32)
implies that the TSO can select only a single network conguration.
206
Together these constraints restrict the set of possible congurations of the network compatible with agents decisions to locate and reserve capacities. This might be sucient to fully determine the conguration of the network, but need not be so. We therefore also leave the possibility that the TSO optimally (that is, through cost minimisation) selects the conguration. One could obviously argue that the constraints (6.31) can be merged with (6.30) and hence that one needs to write only one set of constraints. A reduction of the number of constraints serves no useful purpose whether computationally or economically. From a computational point of view, the formulation that explicitly retains the rst two constraints is likely to be tighter in an integer programming sense (see Wolsey, 1998) and hence more ecient. From an economic point of view, the two constraints express dierent cost causalities and hence should be made explicit. We recognise that the construction of the constraints (6.30) and (6.31) may be dicult in practice, but leave it to further research to investigate how they can be inferred from both engineering practice and mathematical programming models. Before leaving this point, we note that the absence of any agent connecting to the grid (zk 0, xk 0, k) eliminates the need for a network and hence allows the solution z0 0. This justies setting the right-hand side of these constraints to zero. This also means that these constraints dene a cone when the integer restrictions are relaxed. Constraints (6.30) and (6.31) delineate the set of acceptable network congurations. To each of them, one associates a matrix of PTDF coecients, PTDFn, and a vector of thermal limit constraints bn. We dene the set of acceptable injections and withdrawals in network conguration n as: Kn {xn | PTDFn xn 0 0 bn, 1 xk 0 0} (6.33)
where 1 is a row vector of 1 of appropriate dimension. Note that we ignore the constraint bn PTDFnxn in order to simplify the 0 presentation. Kn is thus the set of balanced injections and withdrawals that result in ows that do not exceed the thermal limits of the lines in conguration n. We also introduce a cone Cn associated with the network conguration n as: xn 0 (xn, zn ) | zn 0 0 0
Cn
Kn, 1 xn 0
if
zn 0
0, xn 0
if
zn 0
0 (6.34)
207
(recall that zn as dened above is equal to 1 if one selects network 0 conguration n and is zero otherwise). Dene: x0 and the cone: C0 {(x0, z0 ) | (xn, zn ) Cn, n 0 0 0 1, . . ., N} (6.36) (xn, n 0 1, . . ., N), z0 (zn, n 0 1, . . ., N) (6.35)
It is obvious that (6.34) denes a convex cone. We also know that the Cartesian product of convex cones is a convex cone. (x0, z0) C0 is thus a convex constraint that we shall impose on the network company. The TSO selects a single zn 1 that minimises the cost d n and satises 0 0 (6.30) (6.31). Given this selection, the TSO can oer injection and withdrawal services x0 such that (x0, z0) C0. The agents k request injection and withdrawal services xk such that: signk xk xn 0 (6.37) (6.38)
(x0, z0 ) C0
N n 1
zn 0
1,
z0 {0, 1}N.
(6.39)
Relation (6.37) equalises the production and consumption of the network services (injection/withdrawal). It also transforms the dierent xk into a single non-zero vector x0 of injection/withdrawal services for the selected network conguration. Relation (6.38) expresses that the TSO can only oer network services that are compatible with the selected conguration of the grid. This is the convex part of the production set of the transmission company. In contrast, (6.39) describes its non-convex part, namely that the TSO can select only one conguration of the network. In order to simplify the notation, we rewrite relation (6.37) as E0x0 Ek xk 0.
k 0
208
In order to fully use conic duality, we also dene the cone Ck for each generator or consumer k: Ck {(xk, zk ) | xk
mk zk
0, xk
0, zk
0}.
The Regulators Global Problem Introducing a diagonal matrix sign0 with only 1 components, the extension of the PIPLOC problem can be stated as: Problem EPIPLOC max
k 0
ck signk xk

dk signk zk

(6.40)
E0x0 F0x0
k 0
Ekxk
0 0 0
(6.41) (6.42) (6.43) (6.44)
k 0
Fkzk G20z0 1
k 0
G1k xk 1 z0
(xk, zk ) Ck, zk {0, 1}n(k).
(6.45) (6.46)
In this model, (6.41) expresses the equality between produced and consumed network services; (6.42) and (6.43) relate the selection of the network conguration and the locational decisions of the agents; (6.44) states that only a single network conguration is allowed; (6.45) describes the production sets of all agents, TSO, generators and consumers. Specically, C0 imposes the ows to be feasible for the selected network conguration and Ck requires that an agent rst needs to locate and reserve some injection or withdrawal capacity before generating or consuming. Note that in contrast with problem PIPLOC which is a linear mixedinteger program, the conic constraint (6.45) makes problem EPIPLOC a non-linear mixed-integer program. This non-linearity should not be a real
209
concern though as EPIPLOC can easily be restated as a linear mixedinteger program of the type precedingly developed by power engineers for network capacity expansion (see Latorre et al. (2003) for a survey of these models). The non-linear version is used here because of its tractability for extending ONeill et al.s Theorem 2.4 We therefore again assume that this problem can be solved (for instance by solving the linear mixed-integer version of it) and note (x0, z0 ) and (xk, zk ) , k an optimal solution. The Continuous Version of the Regulators Problem Let z be the locational vector selected by the regulator after solving EPIPLOC. Dene: Problem EPLIPLOC(z) max
k 0
ck signk xk

dk signk zk

(6.47)
s.t. E0x0 F0x0

k 0
Ekxk
0 0
(u0 ) ( f) 0 (
g)
(6.48) (6.49) (6.50) (6.51) (6.52) (6.53)
k 0
Fkzk
k 0
G1kxk 1 k
G20z0 (
0)
1 z0 zk zk
(wk ) k (x*, z*) k k
(xk, zk ) Ck,
Note that c0 0, G10 0 and G2k 0 for k 0. We slightly generalise ONeill et al.s Theorem 2 by relying on the extension of standard linear programming (LP) duality to (convex) conic programming duality (Ben Tal and Nemirovski, 2001) presented in the Appendix. Program EPLIPLOC(z) is a conic program which comprises the conic constraint (6.53). As usual we write the dual variables of the constraints at the right of each of them. The dual variables x*, z* are associated with xk and zk appearing k k in the conic constraint (6.53). The conic dual of EPLIPLOC(z) may be written as:
210
Problem EDLIPLOC(z) min s.t. u0E0 u0Ek

f F0 f Fk gG1k gG20 0 0
wk zk
(6.54)
x* 0 ck signk
0
(6.55) x* k z* 0 z* k (6.56) (6.57) (6.58) (6.59) (6.60)
w0
d0
wk
dksignk
(x*, z*) C*, k k k

f, g
u0,
0,
wk
k unconstrained
where 1 designates a vector of 1 of appropriate dimension. Our interest in this extended model is to nd prices that induce the TSO to invest in the appropriate network conguration and the generators and consumers to invest in the right location and operate eciently. Short-term Locational Charges Before getting into this development, we rst interpret the dierent constraints of the dual of EPLIPLOC(z) and particularly those involving dual cones. Consider rst the dual of the cone C0 associated with the TSO. Recalling the relations (6.34) to (6.36), one can write: C0
n
{(xn, zn ) | PTDFn xn 0 0 0 if zn 0 0, xn 0 0 if
bnzn, 1 xn 0 0 zn 0 0.
The following lemma characterises the dual cone C*: 0 Lemma 1 C* 0

n
C*n 0
n n bn,
{(x*n, z*n ) | x*n 0 0 0 for some

n
1
n
n PTDFn,
z*n 0
and some
0}.
211
Proof
Consider the following linear program: min x*n xn 0 0 z*n zn 0 0 bnzn 0 0

n
(6.61)
s.t. PTDFn xn 0 1 xn 0 0
n
(6.62) (6.63) 0
n
xn unconstrained, zn 0 0
(6.64) and
n,
where the dual variables of the constraints are The dual of this problem can be restated as: max 0 s.t.
nPTDFn nbn n n1
respectively.
x*n xn 0 0 zn 0
(6.65) (6.66)
z*n 0
n
0,
unconstrained
(x*n, z*n ) belongs to C*n i the minimal value of the primal problem is 0 0 0 attained and is zero. This happens i there exists a dual solution n 0, n unconstrained, such that x*n n1 - n PTDFn and z*n nbn. This 0 0 completes the proof. The proof of the lemma immediately suggests an interpretation of the constraint (x*n, z*n ) C*n. The condition (6.65) indeed implies that x*n is a set 0 0 0 0 of nodal prices where n is the electricity price at some hub and n are the prices of the owgate capacities in network conguration n. Condition (6.66) implies that z*n is bounded below by the merchandising surplus. 0 Because the value of the primal and the dual are equal, one also has: xn x*n 0 0 zn z*n 0 0 0
nb when zn and by duality z*n 0; z*n is then exactly equal to the 0 0 0 0 merchandising surplus. The above can be summarised by saying that (x*n, z*n ) C*n i x*n is a vector of nodal prices associated with ow xn in 0 0 0 0 0 conguration n and z*n is the corresponding revenue of the TSO accruing 0 from congestion charges.
212
The TSOs Problem We are now equipped to interpret the role of the TSO. Denote with an upper the dual variables that are solutions of the dual problem EDLIPLOC(z). Consider the TSOs problem CIP0: max
CIP0
u 0E0x0
(d0
lf F0
lgG20
w0 )z0
(6.67) (6.68) (6.69)
(x0, z0 ) C0 1 z0 1; z0 {0, 1}N.
Note by an upper and a prime the optimal solution of CIP0 (for example, x 0 ). We briey discuss the interpretation of the dierent expressions appearing in this problem. Note rst that c0 is identically zero because the TSO does not have variable operating costs. By (6.55), u0E0 is equal to x* and hence, as shown in 0 Lemma 1, is a vector of nodal prices in network conguration n. As a nbn is the merchandising surplus in result, u0E0x 0 x*n x 0n z*n z 0n 0 0 network conguration n, and hence the congestion revenue of the TSO in that conguration. d0 is the vector of investment costs of the dierent network congurations. w0 is a vector of payments/charges imposed by the regulator to induce the TSO to select the adequate network conguration. Finally, lf F0 and lgG20 are locational payments, respectively, resulting from the generators/consumers decision to locate and reserve certain injection and withdrawal capacities. These two last terms do not carry much intuitive interpretation. This is not really surprising: one could not expect to nd an easy interpretation of cost causality when the causal relationships that drive the development of the grid are themselves murky. We shall come back to these terms later in the discussion. In conclusion, relations (6.67) to (6.69) represent the problem of a TSO that maximises the prot accruing from short- and long-term payments when investing in the grid. The Generators and Consumers Problem We now turn to the sets Ck. Recall that Ck is dened as: Ck {(xk, zk ) | mkzk
xk
0, xk
0, zk
0}.
We have the following lemma:
213
Lemma 2 C* k Proof {(x*, z*) | z* k k k
z* 0, mk k
x* k
0}.
Consider the following linear program: min z* zk k x* xk k xk 0, xk 0 0. (6.70)
s.t. mkzk zk
k
(6.71) (6.72)
The corresponding dual can be stated as: max 0 s.t.

k mk k k
z* k x* k 0
(x*, z*) C* i the primal problem has a solution and the objective k k k function value is equal to zero. This requires that the dual problem also has a solution and hence that: x* k Note that zk
kmk
z* k mk
0 and z* k
0.
0, xk
0 implies by the complementarity conditions that:

k
z* and k
x* and hence that x* k k
z* k mk
0.
We are now equipped to interpret the actions of agent k. This latter solves the problem: max s.t. (xk, zk) Ck.
CIPk
(ck signk
lgG1k
u0Ek )xk
(dk signk
lf Fk
w*)zk k
214
Denote again with an upper and a the optimal solution of problem CIPk (for example, xk ). The dierent expressions appearing in this problem can be interpreted as follows. ck signk xk is the generation cost or the willingness to pay of agent k. u0 Ek x k is the revenue collected/paid by agent k for injection/withdrawal of electricity in a nodal price system. lgG1k x k is the demand charge component due to the long-term locational prices. It can be positive or negative depending on the matrix G1k. dk signk z k is the cost incurred by agent k in order to locate on the network and to build the equipment. lf Fkz k is the charge accruing from the locational decision. This charge can also be positive or negative depending on the matrix Fk. w*z k is k a charge/payment levied by the regulator in order to induce agent k to select the right location. Again CIPk represents the problem of an agent that maximises the prot accruing from short- and long-term payments when deciding to locate and to operate on the grid. Compatibility between the TSO and Other Agents Decisions We need to show that the decisions of the generators and consumers, when solving their respective subproblems, combine to give energy ows that satisfy the constraints of the conguration selected by the TSO. We also need to explore whether nancial ows balance. Finally, we also hope that the obtained behaviours do not depart from those found in the usual theory of congestion management. These properties are obtained from the solutions of the primal and dual problems EPLIPLOC(z) and EDLIPLOC(z) as we shall now show. We begin with the less usual result coming from conic duality. The following lemmas cast our model in the standard theory of nodal pricing and hence directly relate to congestion management. Lemma 3 Let (x 0, z 0 ) and (x 0*, z 0*) be respectively part of the primal and dual optimal solutions of EPLIPLOC(z) and EDLIPLOC(z). We have: x 0n If z
n
if
z 0n
0.
(6.73)
0, then x 0n solves: max x*n xn 0 0 (6.74) bn 0
s.t. PTDFnxn 0 1 xn 0 (6.75) (6.76)
215
where x*n is a vector of nodal prices in network conguration n. 0 Proof Let n be the network conguration selected by the TSO (zn 1 in 0 (6.52)). Then (6.73) follows immediately from the denition of C0. The result for n n follows by noting that the problem (6.74), (6.75) and (6.76) is identical to the problem (6.61) to (6.64) of Lemma 1 after replacing zn by 1. x*n can be interpreted as a vector of nodal prices as shown 0 0 in Lemma 1. Lemma 3 recovers a fundamental result of congestion management in nodal pricing (Hogan, 1992). Given a structure n of the network, x*n is the 0 vector of nodal prices and the TSO selects an oer x 0n of injection and withdrawal services that maximises the value of the network. Then (6.73) requires that the energy ows can only be dierent from zero in the selected conguration n. and (6.75) requires that the thermal limits of the lines in the selected conguration are not violated. Altogether problem (6.74)(6.75) requires that the system operator maximises the value of the network capacities, using the nodal prices x*n. 0 Lemma 4 u0Ekxk is the congestion charge paid by agent k.
Proof As argued in the discussion following Lemma 1, the constraint (6.55) at optimum boils down to: u0 x*n. 0 signk.
The result then follows by noting that Ek
Combining Lemmas 3 and 4, one sees that the congestion management part of the problem is identical to the one described in the usual theory of nodal pricing when the network conguration is given. In contrast, the remaining expressions lTF0, TFk, lgF20 and lTG1k do f f g not have any standard interpretation. These are the cost causal parts of the taris that price the impact of the decision to locate and to reserve capacity on the network. They reect the cost incurred by the TSO because of these decisions. The lack of interpretation comes from the fact that constraints (6.30) and (6.31) are of a combinatorial nature and do not have any standard economic meaning. But we shall see that they induce the right behaviour on the TSO and the agents k. In no way do these locational prices satisfy the criterion of transparency demanded by the Regulation. But that should be expected, as the network expansion process is itself a complex problem that is not transparent. Before
216
exploring the global compatibility of the behaviour of the dierent agents, we note that complementarity slackness implies at least some accounting relations between the charges imposed on the agents and the revenue of the TSO. Lemma 5 (i) lf F0z 0 l F z . The sum of the locational charges k 0 f k k lf Fkz k paid by the generators and consumers is equal to the locational revenue lf F0z 0 received by the TSO. (ii) lgG20z 0 klgG1kx k . The sum of the locational charges lgG1kxk paid by the generators and consumers is equal to the locational revenue lgG20z0 received by the TSO. Proof Obvious from (6.49), (6.50) and complementarity conditions.
Decentralisation of the Decisions It remains to show that the decentralised solutions of the TSO problem and of the dierent agent k problems solve the regulators problem, that is, that they induce the TSO to select the network conguration desired by the regulator and similarly that they incentivise generators and consumers to locate and develop as desired by the regulator. Finally, we want to be sure that the electricity and transmission markets clear and that all agents, including the TSO, break even. The following proposition is a direct consequence of Theorem 1 stated in the Appendix: Proposition 2 Suppose all agents k, when solving problem CIPk, are charged/pay the price u0, lf, lg, wk found in the solution of EPLIPLOC(z). Then the solutions found by these agents are identical to the solution of problem EPLIPLOC. Moreover energy markets and the transmission service market clear. Proof The result immediately derives from Theorem 2 in Appendix 6A.
This implies the following corollary: Corollary 1 Lemmas 3, 4 and 5 hold if the solutions of problem EPLIPLOC(z) are replaced by those of the CIPk. This implies that prices u0, lf, lg, w0 and wk allow one to fully decentralise decisions among agents. Congestion management boils down to nodal pricing and retains its usual properties. TSO maximises the value of the network and collects the merchandising surplus. Also the demand charges
217
components of the location prices paid by the users of the network and received by the TSO balance. Most remarkable, the budget of the TSO, including receipts from long- and short-term prices and taking investment costs into account is balanced.
7.
NON-DISCRIMINATORY PRICES
The Regulation and the work of the Florence Regulatory Forum require non-discriminatory prices in order to induce competition. The above taris are only partially non-discriminatory. The line congestion charges in the vector u0 are the same for all agents k and hence are non-discriminatory. This is the usual result of congestion management by nodal prices. The same is true for the demand charges of the long-term locational signals, at least if one accepts the description of causality given by relations (6.30) and (6.31). In contrast the wk are truly discriminatory. This discrimination is justied by the objective of economic eciency. In contrast with the common wisdom underlying both the Regulation and the work of the Forum, there are indeed situations where discrimination is necessary for achieving eciency. Ramsey pricing is the best-known illustration of the usefulness of discriminatory prices in the network industries: its justication is that it makes everyone better o. Still it remains a discrimination and hence may be unlawful. One can obtain non-discriminatory prices, but at the cost of some economic ineciency. This is achieved by constructing a model where one explicitly requires that the wk are equal for all generators and consumers. We briey turn to that question. Recall that the model of Section 6 leads to three-part locational prices that are ecient and cost reective. The congestion and demand charges are non-discriminatory; the locational price is cost reective and non-discriminatory but the locational price (the w) remains both non-cost reective and discriminatory. The question we address here is the removal of the discrimination. We rst take up the question on the simple model of Section 4 and then extend the discussion to the more involved model of Section 6. Consider the problem PIPLOC and the dual problem DLIPLOC(z) of problem PLIPLOC(z). We modify DLIPLOC(z) into CDLIPLOC(z) (constrained dual LIP problem) by imposing the non-discrimination constraint that all locational charges are identical. This leads to: min u0b0 w
k
zk
(6.77)
218
s.t. u0A signk ukmk u0 0, uk uk w cksignk dk signk k k (6.78) (6.79) (6.80)
0, w unconstrained.
The only dierence between CDLIPLOC(z) and DLIPLOC(z) is the implicit introduction of the non-discrimination constraint: wk w 0 k
that expresses that locational charges are identical for all agents. The corresponding primal problem CPLIPLOC(z) is written as follows: max s.t. A
k
cksignkxk
dksignkzk
(6.81)
signkxk m zk zk xk 0 zk
b0 k
(6.82) (6.83)
xk
(6.84) (6.85)
0.
Note that the constraint xk mkzk 0 guarantees that zk 0. It is obvious that problem CPLIP(z) is a relaxation of PIPLOC(z) because it aggregates the capacity objectives zk zk k of the regulator into the single constraint (6.84). Strictly speaking, this does not solve the regulators problem but only approximates it. Nevertheless, it may be a very good approximation as we discuss now. The solution of PLIPLOC(z) aims at selecting the optimal choice of the capacities for each agent in dierent locations. It is unlikely that the regulator has enough information to dierentiate between the cost structure of the dierent agents. Suppose this is indeed so and the regulator cannot dierentiate between agents. He/she is then able to select only the best mix of technologies and capacities at the dierent locations. The vector kzk describes the total capacity of each technology built at each
219
location. The price signal w therefore depends on both location and technology but not on the agent. As a locational signal, it may still be considered as discriminatory to the extent that it dierentiates the locational price as a function of the technology. This may be forbidden by law except if it can be argued that this dierentiation is justied by cost casuality. For instance, gas and coal plants are not expected to operate in the same way and hence could imply dierent investments in the network. At least, the approach does not dierentiate the locational signal by agent, something that would certainly be forbidden. The problem of this formulation is that the prices found by solving problems (6.77) to (6.80) do not guarantee that the constraint (6.84) will be met in the decentralisation process. Agents may still build too much or too little capacity. This is the welfare loss implied by the non-discrimination constraint. This discussion can be extended to the model of Section 6. We consider a slightly dierent (but equivalent) formulation of the non-discrimination problem based on that model: infw w0 z0 w
k 0
zk
max
k
ck signk xk
(d0 0 0
w0 )z0
k 0
(dk signk
w)zk
E0x0 F0x0
k 0
Ekxk
k 0
Ekxk G20z0
k 0
G1k xk
1 z0 1 (xk, zk ) Ck zk {0, 1}n(k). This model is obtained from PLIPLOC(z) by two operations: 1. 2. one dualises the constraint (6.52) zk zk, and one imposes that the prices associated with these constraints are the same for all generators and consumers.
One immediately sees that the second equation imposes the nondiscrimination constraint on it that requires that all w are identical in the objective function of the problem.
220
This formulation will guarantee non-discriminatory locational prices but it will not ensure that the decentralisation process will lead to locations and investments that meet the objective kzk or zk zk of the regulator. This is the same phenomenon as mentioned for the preceding model. Generators and consumers may over- or underinvest, therefore entailing a loss of economic eciency. It is impossible to foresee a priori the extent of the loss of eciency that would result from this non-discrimination constraint on the sole basis of the model formulation. But it is certainly quite possible to investigate the question numerically. Finally, note that an obvious improvement of both models would be to dierentiate the locational charge depending on whether agent k injects or withdraws at some node. This implies dening w as the locational charge for all generators and w for all consumers. This extension is straightforward.
8.
INSTITUTIONAL DISCUSSION
The above discussion oers an abstract setting that is useful to try to relate to current practical proposals. We again rely on the discussion provided by Roselln (2003) that has already been invoked in Section 3. It should be clear that our formulation of the TSO problem best ts the transco model. In this model a single company is in charge of both the investments and the operations of the transmission system. This proposal, which is favoured in Joskow and Tirole (2000) is implemented in the United Kingdom. Combined with an appropriate incentive regulation, this experience is credited to be very successful. The current European institutional framework makes no recommendation in favour of that system and some of the language of the Regulation (Article 6, paragraph 6) even seems not to recommend it. What is certain is that it is impossible to impose a transcolike solution throughout Europe, let alone to adopt a single transco for the whole of Europe. Nevertheless it remains useful to retain this interpretation both because it is implemented in practice and also because that implementation is successful. Our model obviously diers from the exact implementation of the transco in the UK.5 In this chapter, the TSO receives instructions on how to set the long-term signals; it also receives some money transfers through the w0 and 0 that induce it to select the right network conguration. The model also supposes a well-informed, quite knowledgeable and very intrusive regulator. However, the regulator does not need to solve problem PIPLOC him/herself. The transco, or an independent consultant can do so on the basis of commonly agreed data and assumptions. But the regulator
221
is assumed to be comfortable with the whole process and to agree that the zk (or kzk in case of non-discriminations) are desirable objectives. The merchant line is the other approach to grid investments recalled in Section 3. It is analytically grounded in the theory of nodal prices and their extension into long-term nancial transmission rights as hedging instruments. The ISO model, where operations and ownership of the grid are separated, provides the institutional background of the approach. Hogans (2002) theory of merchant lines essentially extends the role of nancial transmission rights from operations to both operations and investments. Joskow and Tirole (2005) argue that this can only be done under drastic assumptions that are violated in practice. In this chapter we concentrate on one of these assumptions, namely the lumpiness of investments, and explore its possible consequences. Our conclusion is that it is still possible to decentralise lumpy investment in the grid provided one invokes a more complex set of prices that covers not only congestion but also access charges. In other words, there is a combination of access and congestion charges that provides the necessary incentives for investments. It is remarkable that this is exactly what the Regulation foresees, but without indicating how this can be done. The reality is that the derivation of these charges is a very demanding task. It indeed requires rst, extending the role of the ISO from the sole operation of the existing system, to include both the operation and investments in the grid. This extension raises several questions that we do not discuss here; however, it retains a key property of the ISO model in the sense that this latter does not need to own the grid. Our abstract model is thus also fully compatible with the notion of a large regional transmission operator (RTO) (for example, PJM6). The creation of RTOs is still a long way o in the European context, but it is denitely more possible than a single transco. The task of the RTO becomes formidable, however. Instead of auctioning only physically feasible nancial transmission rights that only cover congestion charges, this global RTO should now auction two types of long-term contacts, corresponding respectively to the access and congestion charges. In the same way as congestion charges need to be carefully tuned in order to get the prices right, the access charges also need to be well tuned in order to induce the right investments in the grid. Needless to say, ownership of access rights on top of congestion rights by generators as a result of the auction would exacerbate the generators market power in the sense of Joskow and Tirole (2000) and Gilbert et al., (2004). The answer may lie in Hogans suggestion (Hogan, 2002) that transmission companies would probably be the main owners of these rights.
222
9.
FURTHER QUESTIONS
The locational prices obtained in Section 6 satisfy two objectives of the Regulation. They are economically ecient and cost reective. They also allow for the separation between short- and long-term locational prices foreseen by the Regulation. Also all the theory of nodal price/owgate congestion management remains unchanged. Revenue adequacy is a key element of the theory of congestion management. It states that the TSO will normally receive a non-negative prot from congestion management. This non-negative prot occurs in a short-run equilibrium problem, that is for given network capacities, provided that certain conditions related to ex ante and ex post contingencies are met. We retain the standard result of revenue adequacy in congestion management operated under nodal pricing in our more general set-up. Moreover, by Theorem 17 all agents make zero prot at equilibrium. This would suggest that revenue adequacy can be extended to encompass both long- and short-term signals. But there is a dierence between this extended revenue adequacy and the standard congestion management result. Congestion payments between TSOs and agents k are in balance as can be seen by multiplying relation (6.48) by u0. Revenues and expenses from demand charges also balance as can be seen from the complementary condition of constraints (6.42) and (6.43). l* kwkzk is the total residual charge paid by the TSO and the other agents 0 to the Regulator. Assuming that the electricity system has a positive social value, this charge can only be non-negative. We therefore recover a revenue adequacy property where the residual prot is this time with the regulator and not with the TSO. We discussed in Section 5 how Hogan and Ring expanded the scope of prices capable of sustaining the equilibrium. Their reasoning can be extended to the more general model of Section 6. All this should be studied further, probably numerically. Still, these locational prices leave several open problems. We briey mention some of them. The realism of the models can be improved. The model of Section 6 assumes that the system is operated during a single time period, therefore assimilating the demand charge to an additional energy charge. This restriction can easily be removed. It has been a tradition among power engineers to plan the network for the peak period. This is unlikely to be correct because transmission ows are not necessarily highest during the peak. This suggests that a set of reference periods (seasons) should be introduced that are critical for designing the network and hence that can be used for determining the demand charge. This can easily be done by extending the model of Section 6. This would allow for a richer description of the cost causality and hence would lead to a richer set of non-linear taris. The
223
same can be said about the inclusion of the choice of a technology. The zk variables have been dened to refer only to locational decisions. As argued in Section 4, it is easy to expand their interpretation to encompass a technological choice. This would also lead to a richer set of non-linear taris. Recall that the models discussed in this chapter fail to achieve fully nondiscriminatory taris. Bjorndal and Jrnsten (2004) were able to construct non-discriminatory dual price functions for the unit commitment problem treated by ONeill et al. (2004). The possibility of extending their analysis to the grid problem constitutes another area of interest. The following issues may be more challenging. The current model supposes that the cost causality can be expressed through relations of the (6.42) and (6.43) type. This is certainly true in principle but does not say anything about how to derive these relations. This problem raises both power engineering and optimisation questions. Consider rst the electrical engineering issue. A main feature of model EPIPLOC is to embed a description of the production set of the TSO. This description involves both causal relations ((6.30), (6.31)) as well as a description of the transportation possibilities of the dierent network congurations contemplated by the TSO ((6.33) and (6.34)). The representations of the production set of the TSO in economic models is unusual but not totally absent from the literature (Vogelsang, 2001). Our relations ((6.30), (6.31) and (6.33)) are stylised from the following considerations. We assume that the TSO selects its investments using a capacity expansion model. There are several examples of these models in the electrical engineering literature (for example, Latorre et al., 2003, for a survey of these models). Many of these models are of the mixed-integer programming type. We suppose that one of these models was selected. Using the language of combinatorial optimisation, relations (6.30) and (6.31) are valid inequalities, derived from this mixed-integer program, that describe the polyhedron of feasible solutions of EPIPLOC. The construction of these inequalities is currently an area of intense research in combinatorial optimisation. It is not clear, however, that these inequalities have so far been derived for the type of model involved in network capacity planning. Also, current valid inequalities are limited to local description of the feasible polyhedron. In short, the introduction of relations (6.30) and (6.31) makes a lot of sense from the point of view of logic, but their construction remains to be explored. Consider now the following economic issue: the current model is established under assumptions of perfect knowledge and absence of market power. Relaxing each of these assumptions creates a whole set of additional complexities. Consider the assumption of perfect knowledge rst. Following a traditional assumption of the old theory of regulation, the regulator is supposed to know the cost of the generators, the willingness to pay of the
224
consumers and the set of possible network congurations of the TSO. The limitations of this type of assumption have been extensively explored in the new theory of regulation, using the notion of asymmetry of information. Embedding this machinery in extensions of Theorem 2 of the appendix would add a whole new dimension to the work. The absence of market power is another limitation of the model. An extension of the model discussed here to one that would accommodate market power is straightforward to formulate: the problem EPIPLOC should be replaced by a mathematical program subject to equilibrium constraint (MPEC). This model would request that the regulator selects the long- and short-term locational prices in order to induce agents operating in an oligopolistic market to behave in such a way that they maximise welfare. But it is not clear that this problem has a solution. Again, the real diculty is to extend the decentralisation result of Theorem 2 to that more general set-up. The static nature of the model is certainly one of the major shortcomings of the above discussion. Static models are common when discussing equilibrium problems but the simplication is a real drawback when it comes to implementation. Bushnell and Stoft (1996) have initialised the discussion of the dynamics of investments in the electrical grids and their consequences on the validity of existing nancial transmission rights. This has since been extensively elaborated in various papers, among them the discussion of merchant lines. It is totally absent from this set-up.
10.
CONCLUSION
The idea of separating long- and short-term locational signals in the Regulation on Cross-border Exchanges in Electricity is probably a good one. It oers the opportunity to depart from simple linear taris that are not sucient to induce the right investments and locational decisions in a system plagued by discrete decisions. But this separation leaves open how to construct the long-term locational signals. It is easy but misleading to pretend to solve the problem by cost allocation rules. Unless proved otherwise by numerical experiment, there is no reason to believe that these will generate the right signals. This chapter considers the goal of nding an alternative approach or at least to identify ideal abstract conditions that would allow for such an alternative. It retains three criteria referred to in the Regulation, namely economic eciency, cost reectiveness and non-discrimination. It nds that the three criteria cannot be achieved simultaneously, even under ideal conditions. But one can at least trade non-discrimination for the other criteria. It also nds that transparency of the long-term signals is probably hopeless. But this is not surprising as the network expansion process is itself
225
a murky one. Last and probably more important it provides a framework where both the long- and short-term signals can be cast, and the usal theory of short-term signals remains completely unaected.
APPENDIX
METHODOLOGICAL BACKGROUND
The following provides a small extension of the main theoretical result of ONeill et al. The extension is appropriate for introducing cost causality in the network problem. Consider the following mixed integer conic problem max s.t. Ak1xk Ak2zk bk b0 k (6A.2) (6A.3) (6A.4)
CIP
ckxk
dkzk
(6A.1)
Bk1xk
Bk2zk
(xk, zk ) C0; zk {0, 1}n(k)
where Ck is a convex cone. The only dierence between this CIP problem and ONeill et al.s PIP is the replacement of the constraint xk 0 by (xk, zk) Ck. Because xk 0 also denes a convex cone CIP generalises PIP. Let Ck {xk, zk, xkxk zkxk 0 (xk, zk ) Ck} be the dual cone of Ck. One knows that (strictly feasible) conic programs satisfy the same duality properties as linear programs. We then extend ONeill et al.s formalisation as follows. Let z* be the value of zk dual variables in an optimal solution to k CIP, dene the following primal conic program: PCIP (z*) s.t.
k
max
PCIP
ckxk
dkzk
Ak1xk Bk1xk
Ak2zk bk
b0 k
Bk2zk
226
zk
z* k
(xk, zk ) Ck Using duality theory in conic programming, one states the following dual problem: DCIP(z*) s.t. y0Ak1 ykBk1 ck x* k y0Ak2 ykBk2 dk wk z* k y0 0, yk 0 min y0b0 ykbk wkz* k
DCIP
wk unconstrained (xk, zk ) Ck. Theorem 1 of ONeill et al. is readily extended into: Theorem 1 * CIP * PCIP * DCIP
where * indicates the optimal solution value for the respective problems. The proof is identical to that of ONeill et al. Consider now a set of price vectors (P0, Pz ) . ONeill et al.s agent k problem can be adapted into: CIPk max
CIPk
(ckxk Bk1xk
dkzk ) Bk2zk
P0 (Ak1xk b1
Ak2zk )
Pz zk k
(xk, zk ) Ck, zk Zk. We then dene a competitive equilibrium as a set of prices {P*, Pz*} for k 0 all k and allocations {x*, x*} for all k such that at the prices {P*, Pz*}, the k 0 k 0
227
kAk1xk
allocations {x*, x*} solve PIPk for all k, and the market clears: 0 k b0. kAk2zk ONeills Theorem 2 can then be generalized as:
Theorem 2 Let {x*, z* } be the solution to CIPk and PCIP(z*) and let 0 k { y*, y*, w*} be the solution to CDIP(z*). If y* P0 and w* Pz then k 0 k k 0 k the prices { y*, w*} and allocations {x*, z* } for all k is a competitive 0 k k k equilibrium. Proof The proof is almost identical to the one reported by ONeill et al. The only dierence is the replacement of the two complementarity conditions: 0 0 (y*Ak1 0 (y*Ak2 0 y*Bk1 k y*Bk2 k ck ) xk w* k dk ) zk 0 0 k k
by the generalised complementarity condition: xk zk xk zk
Ck which states that:
Ck,
xk zk
Ck xk zk y*Bk2 k w* k
y*Ak1 y*Bk1 ck 0 k y*Ak2 y*Bk2 w* dk 0 k k (y*Ak1 0 y*Bk1 k ck )xk (y*Ak2 0
Ck dk )zk 0.
These generalised complementarity conditions are used to prove ** * CIP CIP by noting that:
k k
(ck because:
y*Ak1 0
y*Bk1 )x** k k
(dk
y*Ak2 0
y*Bk2 k
w*)z** k k
228
y*Ak1 y*Bk1 ck 0 k y*Ak2 y*Bk2 w* dk 0 k k xk ** zk ** and the denition of Ck. Ck
Ck
NOTES
* 1. Support from the French Commission de Rgulation de lEnergie (CRE) is gratefully acknowledged. The Electricity Regulatory Forum of Florence was set up to discuss the creation of a true internal electricity market. The participants are national regulatory authorities, member states, the European Commission, transmission system operators, electricity traders, consumers, network users and power exchanges. See http://europa.eu.int/comm/energy/ electricity/orence/index_en.htm. See Stofts chapter in this book: See any textbook in integer programming, for example, Wolsey (1998). The theorem is recalled in the Appendix. For a description of the electricity system in England and Wales, see Joskows chapter in this book. See Joskows chapter. See the Appendix.
2. 3. 4. 5. 6. 7.
REFERENCES
Ben-Tal, A. and A. Nemirovski (2001), Lectures on modern convex optimization: analysis, algorithms, and engineering applications, Society for Industrial and Applied Mathematics (SIAM), Philadelphia, PA. Bjorndal, M. and K. Jrnsten (2004), Equilibrium supported by dual price functions in markets with non-convexities, Department of Finance and Management Science, Norwegian School of Economic and Business Administration, Bergen, Norway. Bushnell, J.B. and S.E. Stoft (1996), Electric grid investment under a contract network regime, Journal of Regulatory Economics, 10, 6179. Crew, M.A., C.S. Fernando and P.R. Kleindorfer (1995), The theory of peak-load pricing: a survey, Journal of Regulatory Economics, 8, 21548. Curien, N. (2003), Cost allocation methods, in F. Lvque (ed.), Transport Pricing of Electricity Networks, London: Kluwer Academic, pp. 73101. European Parliament and Council (2003), Regulation (EC) No. 1228/2003 of the European Partiament and of the Council of 26 June 2003 on Conditions for Access to the Network for Cross-border Exchanges in Electricity 26.6.2003, Ocial Journal, L 176, 15/07/2003, 00010010. Gilbert, R., K. Neuho and D. Newbery (2004), Allocating transmission to mitigate market power in electricity markets, Rand Journal of Economics, 35 (4), 691709.
229
Green, R. (2003), Cost recovery and the ecient development of the grid, in F. Lvque (ed.), Transport Pricing of Electricity Networks, London: Kluwer Academic, pp. 13753. Hogan, W.W. (1992), Contract networks for electric power transmission, Journal of Regulatory Economics, 4, 21142. Hogan, W.W. (2002), Financial transmission right incentives: applications beyond hedging, Presentation at Harvard Electricity Policy Group, May 31, www.whogan.com. Hogan, W.W. (2003), Transmission market design, Conference Electricity deregulation: where to go from there, Paper presented at Texas A&M University April/4, www.whogan.com. Hogan, W.W. and B.J. Ring (2003), On minimum-uplift pricing for the electricity market, Kennedy School of Government, Harvard University, www. whogan. com. Joskow, P. and J. Tirole (2000), Transmission rights and market power on electric power networks, Rand Journal of Economics, 31 (3), 45087. Joskow, P. and J. Tirole (2005), Merchant transmission investment, Journal of Industrial Economics, 53 (2), 23364. Latorre, G., R.D. Cruz, J.M. Areiza and A. Villegas (2003), Classication of publication and models on transmission expansion planning, IEEE Transactions on Power Systems, 18 (2), 93846. Lvque, F. (2003), Legal constraints and economic principles, in Lvque (ed.), Transport Pricing of Electricity Networks, London: Kluwer Academics, pp. 333. Mas-Colell, A., M.D. Whinston and J.R. Green (1995), Microeconomic Theory, Oxford: Oxford University Press. ONeill, R.P., P.M. Sotkiewicz, B.F. Hobbs, M.H. Rothkopf and W.R. Steward Jr (2004), Ecient market-clearing prices in markets with non convexities, European Journal of Operations Research, 164 (1), pp. 26985. Prez-Arriaga, I., L. Olmos Camacho and F.J. Rubio Odriz (2002), Report on cost components of cross border exchanges of electricity, http://europa.eu.int/ comm/energy/electricity/orence/index_en.htm. Prez-Arriaga, J.I., F.J. Rubio, J.F. Puerta Gutirrez, J. Arceluz and J. Marin (1995), Marginal pricing of transmission services: an analysis of cost recovery, IEEE Transactions on Power Systems, 10 (1), 54653. Prez-Arriaga, I. and Y. Smeers (2003), Guidelines on tari setting, in F. Lvque (ed.), Transport Pricing of Electricity Networks, London: Kluwer Academic, pp. 175203. Pope, S.L. and S.M. Harvey (2002), TCC awards for transmission expansions, www.pjm.com/services/trans/rtp-meetingnote3.html. Roselln, J. (2003), Dierent approaches towards electricity transmission expansion, Review of Network Economics, 2 (3), 23869. Scarf, H.E. (1994), The allocation of resources in the presence of indivisibilities, Journal of Economic Perspectives, 8 (4), 11128. Tirole, J. (1998), The Theory of Industrial Organization, Cambridge, MA: MIT Press. Vogelsang, I. (2001), Price regulation for independent transmission companies, Journal of Regulatory Economics, 20, 14165. Wolsey, L.A. (1998), Integer Programming, New York: John Wiley & Sons. Woolf, F. (2003), Global Transmission Expansion: Recipes for Success, Tulsa, OK: Penn Well.
7.
Compatibility of investment signals in distribution, transmission and generation

Ignacio Prez-Arriaga and Luis Olmos
INTRODUCTION
1.
Classical centralized optimization of expansion planning in electric power systems was a conceptually easy task, regardless of the many practical diculties in its actual implementation. All relevant decisions remained in the hands of the vertically integrated utility (Prez-Arriaga et al., 1987; Joskow and Tirole, 2002). This is true, in particular, for the expansion of generation and the transmission grid, which are closely related. Generation expansion planning took place rst, typically assuming a perfect network without losses or capacity limits. Then the utility planned the transmission grid, taking the location and characteristics of any new generation investment as input data. In some rare cases the expansion of both generation and transmission was planned jointly, with the single objective of minimizing the cost of electricity supply. Distribution network planning took place separately, where the main input data were the individual or aggregated demands of consumers both load prole and location as well as the substations connecting the high-voltage distribution grids to the transmission network where generation is feeding the electricity supply (Brown, 2002). In a liberalized regulatory environment a complete new paradigm for the expansion of power systems becomes necessary. Now, the planning of generation, transmission and distribution take place independently from one another. Generation expansion and operation is now left to the initiative of private investors with the adoption of any regulatory instruments to promote adequacy of generation capacity and security of supply remain as open issues (see Chapter 3 of this book). Distribution expansion has become more sophisticated with new ideas, such as performance-based remuneration and benchmarking or the stimulating perspective of widespread distributed generation, but the change is not necessarily associated
230
Compatibility of investment signals
231
with the liberalization of the power sector. The accumulated experience in the restructuring of the sector shows that transmission expansion critically depends on the regulatory paradigm that is adopted for the activity. By transmission planning regulation we mean the choice of the entity responsible for preparing and approving the plan, the decision on who will build the transmission facilities, the remuneration of the transmission services and the economic signals and incentives used to encourage any involved entity to perform these tasks eciently. There are dierent approaches to make compatible, in an ecient way, the two worlds of competitive generation and monopolistic transmission in the context of a liberalized regulatory framework. One possible option is leaving some specialized public-service type of entity to keep performing centralized transmission expansion in the classical manner. Under this scheme, there are several aspects that now become really important and deserve to be discussed specically. A clear and objective process is needed the so called regulatory test to determine which new investments must be made. This process must avoid being discriminatory, that is, it should not benet some agents at the expense of others. Also relevant is how to allocate the cost of new lines among market agents. The advisability of establishing incentives for the system operator to plan the expansion of the grid in the most ecient way must also be investigated. Another option is to let market agents participate actively in the decisionmaking process of transmission investment. There are several possibilities. One has to decide up to what extent to leave transmission investment in private hands and how to make private investment compatible with some level of centralized decision making or regulatory control. Now the role of transmission rights in this process usually becomes very relevant (see Bushnell and Stoft, 1996; Joskow, 2005). In order for the development of the system to be ecient it is necessary that economic signals are sent to coordinate the investment decisions made by the institution in charge of the network expansion or the coalitions of network users, depending on the adopted approach and generation investors. A major unsolved problem is the interaction between transmission and generation expansion. The links between transmission and distribution expansion also need to be studied. First, the reasons for regulating distribution and transmission independently have to be examined and it must also be evaluated whether they are still justied in the new regulatory environment. One has to bring here the experience that has been obtained from existing sound regulatory schemes of transmission and distribution network planning (see Khator and Leung, 1997 and Brown, 2002). This chapter examines a number of issues that are frequently faced by
232
transmission planners in the new regulatory environment and that still do not have a conclusive answer:
Whether there should be an underlying coincidence between the outcome of transmission investment under the traditional and the market-based regulatory approaches. The role that transmission-related economic signals and incentives when separately applied to the activities of generation, distribution and transmission can fulll in achieving joint eciency in investment and operation of the complete power system. Whether the set of transmission-related economic signals and incentives that are useful under a given regulatory paradigm can also make sense under a dierent one. The relative strength that transmission-related locational signals can have when compared to other competing economic signals or to the distortions that are introduced by lack of regulatory harmonization in tari setting. In particular it will be examined whether transmissionrelated locational signals as they are currently used in some power systems could be enhanced in an ecient and cost-responsive way so that they may result in more ecient location decisions by the network users.
Following this introduction, Section 2 reviews and classies the several paradigms that have been proposed to regulate transmission investment. This section also introduces and discusses the regulatory test, a construct to provide conceptual support to any decision related to the regulatory approval of a new transmission investment. Section 3 examines which locational signals make sense to encourage good practices in transmission planning under dierent regulatory paradigms. Section 4 analyses the compatibility of economic signals for generation and transmission and how these signals may complement one another. Section 5 examines this very same issue but now concerning distribution and transmission. Section 6 presents several case examples, based on actual power systems, which help to illustrate the principles that have been presented in the preceding chapters. Section 7 concludes.
2.
REGULATORY PARADIGMS IN TRANSMISSION INVESTMENT
With the introduction of competition in generation and supply, the regulatory environment in which the planning of transmission takes place has
233
changed radically. Even the organization model of the transmission activity is under scrutiny. Most systems maintain traditional planning approaches that correspond to a regulated monopoly, although some incentive schemes may be added. And there are those who favor the restructuring of this activity by leaving some responsibility in the expansion of the grid to the private initiative. Conceptually this can be done either by allowing coalitions of users to promote or even decide the construction of some lines subject to some sort of regulatory approval and the application of the same general rules of open access, remuneration and tari setting as for any other lines or by allowing private investors to build a line and sell its transmission services, by buying electricity at one end of the line and selling it at the other. It is important to realize from the outset that the incentives to invest in transmission do not exist per se. They depend on the regulatory framework that has previously been adopted for the transmission activity as a whole. Ideally, those agents that benet from the expansion of the grid, that is, the market agents, should be left to decide, or at least propose, the construction of new lines. However, due to the fact that market benets are frequently very dicult to identify and allocate or are very widely distributed, such an approach may turn out to be problematic. One cannot say much on the locational signals that are derived from transmission until the several regulatory paradigms for transmission investment have been carefully characterized. This is why this section summarizes the main regulatory schemes for transmission investment previously presented in Chapters 2 and 5, in a format that is suitable for the discussion on investment signals in transmission, distribution and generation that will take place in the next sections. One can distinguish several cases. First, the system operator is responsible for proposing a reinforcement plan, which has to be authorized by the regulator. Then, the construction of this reinforcement may be assigned by competitive bidding or (more frequently) to the incumbent transmission company which in most cases is also the system operator and in European countries is thus named the transmission system operator (TSO) under cost-of-service remuneration. A more detailed description: The system operator must propose a plan for reinforcements of the transmission network, after taking into consideration (or rejecting, with due justication) any proposals made by network users. The regulatory authorities approve the plan (or not) and, if it is approved, they authorize the construction of individual new facilities. Here the system operator can be termed passive, because it proposes plans for the expansion of the transmission network, but does not make the nal decisions regarding approval of the plan and construction (for example, REE in Spain1). Both the system operator, when proposing the construction of a line, and the regulator, when deciding upon the approval of the
234
project, need some kind of rule to apply in order to determine whether the investment is justied and what is the net benet associated with it. This rule is called the regulatory test. See below for further details regarding the concept of the regulatory test. Appendix 7A1 presents the mathematical formulation of the problem of optimizing the expansion of the grid within the traditional framework, and how it is conceptually equivalent to the regulatory test to be applied by the system operator and the regulator in the new competitive environment. However, one cannot ignore that both the regulator and the system operator have a natural inclination to accept overinvestment in grid facilities, since both can be held responsible for the overall security of the power system in one way or another. Therefore the more prominent the role of these institutions in making the nal decisions on transmission investments, the greater is the degree of overinvestment to be expected with respect to what would be optimal with the assumed utility function of consumers. It is debatable whether this is a positive feature of this method. Construction, operation and maintenance of each facility are here the responsibility of the incumbent transmission company (usually the TSO). An alternative is competitive bidding. If competitive bidding is adopted, the winner of the bidding process is paid as bid, and the bid becomes the regulated cost of the line or lines built. Since the contract will have a limited duration (the economic life of the asset, typically), the operation and maintenance of the facility may be auctioned again at the end of the period. Availability targets may be set for each facility and penalties or credits can be applied, depending on the actual performance. Second, a for-prot (private, usually) company is awarded the transmission license and it is regulated as a monopoly: investment and operation is subject to a prescribed grid code, while the remuneration is based on some price-control scheme of the type RPI-X.2 A more detailed description: A private company is awarded the transmission license and regulated as a monopoly. The transmission company must follow prescribed design requirements (a mandatory grid code) and it will be subject to incentives to meet performance targets for the transmission system. This is similar to the more customary regulation of distribution networks. This private company can be called an active TSO because it is ultimately responsible for the expansion of the network (for example, National Grid Company (NGC) in England and Wales). In this approach the scheme that is typically adopted to determine the global remuneration of the complete transmission network is of the type RPI-X where, given that transmission network investments can easily be examined individually, the nal remuneration should also take explicitly into account the actual new investments, the economic lives and the depreciation
235
of existing investments, the nancial situation of the transmission company and the expected improvements in eciency. The combination of a global application of RPI-X to the entire network plus consideration of individual investments and economic incentives to promote better performance makes transmission remuneration more of an art than anything else. One concern with this otherwise sound approach is that optimality of investments will not be attained in general. It is not a trivial task to develop remuneration mechanisms that encourage the system operator to pursue those investments that are most benecial for the system, while pursuing its own benet, since the incentives may have implications in dierent aspects of the security and economics of the power system. For instance, economic incentives to reduce congestion costs might result in reduced security margins. Or economic incentives to reduce network losses might result in operating measures that could be detrimental to the eciency of the generation dispatch. A carefully designed combination of moderate economic incentives and regulatory supervision might be successful in practice. Experiences in countries where this type of approach has been adopted show that certain well-justied reinforcements interconnections in particular are not attractive to the planner because they would not earn any income for their construction (see CEER, 2003). Third, some coalition of network users proposes a reinforcement, which has to be authorized by the regulator. In order for the regulator to authorize it, and according to what has been explained before, both the positive and negative eects of the line on the outcome for all the market agents would matter. The construction is assigned by competitive bidding, which determines the total regulated cost to be recovered by network taris. As is the case for alternative 1, where a passive system operator is responsible for proposing new investments that must be approved by the regulator, a regulatory test is needed for the regulator to be able to decide whether a new line is justied (see Littlechild and Sherk, 2004). A more detailed description: Here the initiative to propose and in some cases also to build network reinforcements corresponds to coalitions of network users. Several options are possible and some have been tried in the Argentinean system. In a rst option the coalition builds and pays for the reinforcement, which needs authorization from the regulator, has to provide open access and receives the proceeds of general transmission charges resulting to the users of the facility. In a second option, coalitions for and against the building of the line have to go through a quasi-judicial process, and the regulator decides whether the reinforcement is justied. If justied, its construction is assigned by competitive bidding and paidas-bid to the winner of the auction. Availability targets may be set and penalties or credits will be applied, according to the actual performance.
236
The license will have limited duration and will be auctioned again for operation and maintenance for the next period. The regulated cost of the reinforcement will be charged to all network users by means of transmission taris. The third option is known as investments at risk. As in the second option there is a quasi-judicial process. If the reinforcement is nally found to be justied and the proprietary coalition is properly dened, construction is assigned by competitive bidding. The regulated remuneration for the line is determined by the auction. A fraction of it, which may be related to the fraction of the line that is actually used, is charged to all network users by means of regulated transmission taris. The remaining fraction is covered by the proprietary coalition, who also receives nancial rights (rm transmission rights, FTRs) for part of the congestion rents of the line. Fourth, merchant lines, whereby their unregulated remuneration is based on the market value of their transmission services. A more detailed description: The basic idea is to regulate the transmission activity as any other competitive business. Therefore, the owner of the line could use it to buy energy at one end of the line (where the energy is cheaper) and sell it at the other (where it is more expensive) at marketdetermined prices. Network constraints most frequently line congestions are at the origin of the dierences in energy prices between nodes. The remuneration of a merchant line comes from the dierence in prices between the end nodes of the line or from the revenues resulting from the sale of the corresponding transmission rights (in some DC lines it has even been proposed that the network capacity be bid in a short-term market). Here FTRs may be seen not only as a risk-hedging mechanism, but also as a way to nance investments in the grid. Promoters considering investing in a new line or corridor will have more certainty of recovering the investment if they are able to sell transmission rights for part of the capacity of the line in advance of its construction. Some of the diculties in generalizing this approach to a signicant number of lines are: (a) insuciency of market-driven revenues to pay for the total cost of a well-developed network,3 resulting in the fact that only those lines where high congestion rents are expected will be built under this approach; (b) high exposure to risk in revenues for the network investor, in particular if regulated lines can also be built which reduce the amount of congestion on the line; (c) underinvestment in a merchant line is a way of abusing a market power situation, given the highly discrete nature of transmission investment. Once a merchant line is built that is smaller than what is advisable, according to the general interest, it may not be justied that the central planner, supposing there is one, installs a new line, since the total investment in the corridor may be excessive.
237
Therefore, when merchant lines are allowed as a possibility for building reinforcements they should always be just an alternative to another approach such as the rst or second case above, since they are capable of building any line that is justied because of economic and security reasons. If some kind of centralized network planning and merchant investments coexist, rules for priority between both approaches have to be established. In some systems the installation of a merchant line must be subject to the regulatory authority giving its consent to it. The consent might be subject to verication that the proposed merchant line is not detrimental to the system and, it also seems appropriate, that it is not included in the plan of regulated investments that has been prepared by the responsible centralized institution. Note that it should be more favorable for the network users to pay the regulated cost of a line rather than its congestion rents, which supposedly are expected to be higher, since the merchant investment in the line appears to be attractive. See CEER (2003) for details on this type of regulation, in the context of the process of formation of the internal electricity market of the European Union. Broadly speaking, the drawbacks of most of these methods stem from the diculty in aligning the interest of the grid promoter, whoever it may be, with that of the system as a whole. The regulatory test, to be discussed next, makes it possible to verify whether any proposed transmission investment increases the global welfare. The Regulatory Test The regulatory test is the rule that should be applied by the system operator or the institution with the responsibility of proposing a network expansion plan when preparing transmission expansion plans and by the national regulators (if this is the case, depending on the specic domestic regulation) when deciding whether any proposed new investments should be authorized. Conceptually, the rule should distinguish a justied investment from another one that is not justied. What is more, if the expansion of the grid is centrally planned, it should help the regulator to identify the most ecient investment among a set of possible ones. In order to do so the test should also evaluate the opportunity cost corresponding to the alternatives to the investment under consideration (see Australian Competition and Consumer Commission, 2003). Under traditional regulation the rule is as follows: One should invest in transmission network assets only while the additional network investment cost is still smaller than the additional saving in system operation costs. Obviously, as has been mentioned above and expressed in simplied terms,
238
when alternative investment possibilities exist, the network planner should choose the combination of investments that maximizes the ratio operation savings/additional network cost. Appendix 7A1 proves that this denition is consistent with the one that should be applied in a context of competition: Invest while the additional network investment cost is smaller than the net aggregated benets (once network charges are included) of all network users (that is, all generators and all consumers). The net benet of a generator, because of a new transmission investment, is the increment in its margin of market revenues over operation costs, minus the network charges corresponding to the new investment. The net benet of a consumer, because of a new transmission investment, is the increment in the margin of its utility function over its cost of purchasing electricity, minus the network charges corresponding to the new investment.4 Note that, conceptually at least, the regulatory test should allow the justication of so-called reliability lines, that is, lines whose justication is mostly because of a general improvement of reliability conditions in the power system. Reliability can be quantied in economic terms, even if the beneciaries whose reliability conditions are improved might be widely dispersed. This may be done by making use of the utility function associated with the consumption of electricity. However, most frequently, transmission planning is subject to hard reliability constraints in network design that are imposed by some sort of national or international grid code, regardless of their economic justication. Note, also, that reinforcements in the transmission network may bring substantial benets to large aggregations of distributed generation that are or plan to be connected to distribution networks, since these reinforcements may facilitate or even make possible the secure injection into the lowervoltage grid of signicant amounts of power that otherwise could not be reliably absorbed by the local demand. Most required network investments are not large ones, such as long lines or complex substations, but minor reinforcements, replacements or modications. The nature of actual transmission investments is explained in Chapter 5 of this book, so it will not be repeated here. Therefore, although conceptually sound, the regulatory test is very dicult to apply in practice in strict terms. Dierent kinds of approximations and simplications are used in practice throughout the world by regulators.5 Note that, in order for the planner to strictly assess whether a new line is benecial or not or the best option to meet the new needs in the system one should try to compute, at least approximately, the true economic benets that are produced by the line.
239
3.
ECONOMIC SIGNALS RELATED TO TRANSMISSION INVESTMENT
We are now interested in designing incentives that could promote an optimal expansion of the grid. Quite logically, these incentives should vary according to which entity is chosen under the adopted regulatory scheme to be responsible for deciding on the lines to be built and where they should be located. In this section, the network planners and potential investors for each regulatory approach are identied, as well as the most appropriate economic incentives for each of them. Potential Network Planners and Investors There are three main categories of potential grid promoters:
The system operators (either integrated or not with the transmission rm or rms); here we shall assume that this integration exists, since this is the common practice in Europe, where most system operators are also TSOs, subject to some degree of regulatory supervision. These TSOs may be active or passive, as explained in the previous section. Coalitions of network users, with their proposals subject to regulatory approval and execution by the regulator, or just authorization and execution by the coalition itself (investment at risk). Merchant investors, who have unregulated remuneration and may be subject to regulatory authorization. Merchant investors will collect just the congestion rents of their lines, or their expected values, via capacity contracts or rm transmission rents of some kind.
Incentives for Transmission Network Investment For the three categories of grid promoters that have just been presented, we shall now examine which are the most suitable incentives to promote transmission investment. First we shall consider the case where the TSO is in charge of making the plan to develop the grid. Two situations can be distinguished, depending whether the system operator is fully responsible for the development of the grid or the nal decisions on investment rest with the regulator:
Active TSOs: The remuneration of the transmission activity for these TSOs should refer to an ecient and well-adapted network not the actual one. Economic incentives could also exist that are
240
related to the actual performance of the transmission system regarding quality of supply. Then these TSOs will invest in transmission at least in order to comply with any mandatory grid codes and only beyond this if the remuneration scheme (typically a mix of RPI-X, cost-of-service and performance-based incentives) appears to make the new investment protable. Compliance with the minimum levels of quality of supply can be guaranteed by the use of administrative penalties. Assuming that the adopted remuneration scheme is such that the income of the TSO is not directly related to its actual investment, the system operator will try to cut down its expenditure in new grid equipment as much as possible. This avoids the risk of overinvestment that is typical of more regulated approaches, but it is dicult that an optimal level of investment may be achieved this way. Passive TSOs: TSOs propose plans for the expansion of the transmission network, but they do not make the nal decisions regarding approval of the plan and construction of the facilities. It is dicult to design correct incentives for a TSO that is not fully responsible for the expansion of the grid that nally takes place. Here the nal responsibility for the new network reinforcements actually corresponds to the regulatory authorities. The remuneration of the transmission activity for these TSOs should refer to the existing network and, in principle, any incentives either penalties or credits should only depend on the availability record of the network equipment. Typically, these TSOs will tend to propose plans with some excess investment, since this will reduce the potential problems of security of supply that may be derived from lack of transmission capacity and maintaining a secure operation is the main responsibility of the TSO. Besides, if the TSO builds more lines, the size of the company will grow with assets that have a guaranteed and attractive remuneration (if the allowed rate of return is reasonable). Therefore, under this approach one should not expect that the TSO will propose the most ecient grid that it is possible to build. But it must be remembered that here the ultimate responsibility for transmission planning corresponds to the regulator. However, it may also be presumed that it will be dicult for the regulator to reject investments that have been proposed by the TSO, therefore resulting in some level of overinvestment.
Coalitions of network users constitute the second category of potential network investors, with their proposals subject to regulatory approval and assignment of constructors, or just authorization (investment at risk). Nodal energy prices and transmission taris with locational signals, when
241
they exist, will provide incentives to the network users to organize coalitions to promote new transmission investments that are benecial for them, if the regulation allows this participation in the process. The success of this scheme critically depends on the existence and correctness of the locational signals in nodal prices and/or transmission taris. Lines that are economically justied but whose benet is widely dispersed (for example, lines whose main purpose is to improve the overall reliability of the power system) will never get a coalition of users that is interested in promoting or building them. In a well-meshed network, as is generally the case in Europe, the economic benets of new transmission reinforcements will probably be quite dispersed. Therefore this approach has to be supplemented by a centralized back-up scheme that makes sure that these lines are built. It remains a controversial issue whether a line should be accepted when it is proposed by a coalition whose benets because of the line are higher than the cost of the line, even if the construction of the line harms other users and the line is not benecial for the system as a whole. If it is agreed that only those lines that have a net benet for the system should get approval for construction, then the coalition should bear a fraction of the cost of the line, so that the net benet of the line for all users is positive. Alternatively, transmission charges could be designed so that agents would pay not only for the cost of the line they are beneting from but also for the harm the construction of the line is doing to other agents in the system. The third category of network investors comprises the merchant investors: their income will consist of the congestion rents of their lines, or their expected values, via application of nodal energy prices or the sale of some kind of rm rights of transmission capacity. Open network access by buyers and sellers is obviously needed. The expectation of these congestion rents is the only driver behind network investments by merchant grid promoters. Merchant lines will only be built if there is an underlying structure of energy prices that reect dierences between locations because of network constraints. These dierences in energy prices may take place within a single power system or in between power systems. Some Initial Conclusions We therefore conclude that regulated investment, which may adopt a variety of formats, must play a predominant role in the future development of almost every real-world transmission network. This is especially true when the system network is already well meshed, as it is mostly the case in Europe (at least within each country).
242
When TSOs are responsible for decision making in transmission expansion and construction of the new facilities, some performance-based incentives may be given to the TSOs so that network reinforcements go beyond the minimum that is required by any mandatory grid code. When TSOs only propose the expansion plans and the regulatory authorities make the nal decisions about the realization of all investments, the natural incentive of the TSO will be towards overinvestment, since this creates a comfortable security margin for the operation of the power system, and also the opportunity to grow in size. The regulator must be aware of this tendency, although the regulator itself is not free from the same impulse towards more security. If economic signals are provided by nodal prices and transmission taris with locational content, coalitions of network users receive the correct economic incentives to promote the construction of new transmission facilities. However, this sound mechanism can only be trusted for those lines with clearly dened and relatively few economic beneciaries. Also interesting is the idea of coalitions of TSOs who could propose the development of interconnections or other facilities of interest for crossborder trade. The facilities could be authorized by the corresponding regulators after using some kind of extended regulatory test. These regulators should make sure that these cross-border investments are properly remunerated. This appears to be a pragmatic approach to getting some lines built, when they are mainly justied in terms of their impact on cross-border trade. This is particularly true of those lines that are built within a country that does not happen to be the main beneciary of their construction. Nodal prices and/or congestion rents are natural incentives for merchant investors who want to appropriate these price dierences. However, the investment that would maximize the prots of a merchant investor is typically of a lower capacity than the investment that the regulator would have chosen. A larger investment would reduce too much the remaining congestion rents. Besides, if merchant and regulated network investments coexist, then the merchant investor always runs the risk of losing his congestion rents because of a new regulated investment. This is why merchant investment can contribute to the development of a transmission network only in a few specic instances. One major problem faced by the grid promoters in most liberalized systems is that the certainty about generation investment and the dierence in time scales in construction time of transmission and generation during the good old times of traditional central planning no longer exist. This adds a new level of complexity to transmission network expansion, which will be discussed next.
243
4.
COMPATIBILITY OF INVESTMENT SIGNALS IN TRANSMISSION AND GENERATION
The Diculties The objective now is to make compatible, in an ecient way, the two worlds of competitive generation and monopolistic transmission in the context of a liberalized regulatory framework. Strong evidence supports the claim that, despite being independently managed, transmission and generation (and to a lesser extent also demand) inuence each other. As a consequence of this, the optimal development of the system can only be achieved if coordinating signals are employed that indicate to network users what the impact in terms of network costs of new generation or demand being installed at each point of the system will be and, vice versa, how the agents will be aected by the addition of a new line. Contrary to what happened in the traditional regulated environment, new investment in generation does not result from a centralized process but from the non-coordinated decisions of a multiplicity of investors. Therefore, the central planner no longer knows with certainty what the development of generation in the system will be. The construction of new lines may have a signicant inuence on the competitive position of some generators in the wholesale markets, therefore guiding the installation of new generators in each area. For instance, some reinforcement may eliminate a network constraint that was forcing the operation of an expensive plant that, otherwise, would remain idle most of the time. Or, on the contrary, it may eliminate a bottleneck and allow a generator to increase its output and export its power to another region. Thus, for example, the reinforcement of the corridors connecting the Iberian system to the rest of Europe could allow less-expensive generation in a very wide area to access the future Iberian market and the average energy price in this local region should presumably fall below its present level. Locational transmission signals (a combination of losses, congestion and transmission network taris) may be inuential in the ecient election of network connection sites for new investments in generation and, perhaps, large consumers as well. In order for the agents to make ecient investment options, they should face the real cost that the system will incur because of the installation of a new generation or consumption facility at a given point in the network. This can be achieved by sending locational signals in the form of nodal energy prices on the one hand, and transmission taris with longer-term locational content on the other. Although the existence of transmission-related charges is just one among several considerations when
244
choosing a site for a new investment, in some cases it may have a decisive inuence. This is particularly true when there are actual choices to be made, such as building a plant near a convenient fuel supply (a gas eld or a highpressure pipeline) and shipping the power via the transmission network to the load center, or building the plant near the load center and bringing the gas to the plant via a pipeline. It would be clearly inecient to apply the same transmission-related charges to two identical new power plants such that: (a) one would be located near the major load center in the system, so that losses and congestions are small and no network reinforcements are needed and (b) the other would be located far from the main load centers, so that losses will increase, congestion problems will appear and the network will need to be reinforced. A numerical case example will be presented in Section 6, showing some of the major considerations to be included when determining transmission-related locational signals. Last but not least, and mainly due to growing environmental concerns raised by local and regional administrations and organizations in general, the process for obtaining the environmental permits and rights of way needed to begin the construction of a new line is signicantly longer than what was usual before. Consequently, construction time for the prevalent generation technologies in most countries (combined-cycle gas turbines, CCGTs, typically) is often shorter than the construction time of new transmission lines. As a consequence of all this, the transmission planner nds it increasingly dicult to predict where the new generation will be installed with a sucient time margin. For instance, according to the Spanish system operator, the average time to get a line built is about ve years now, far more than the one to one-and-a-half years strictly needed to build the line and also more than the two or three years that the construction of a CCGTs plant ideally takes (in practice at least three and sometimes even up to ve or six years). In some cases the construction of highly controversial lines has been delayed by more than 20 years or even indenitely. Signals Derived from Transmission Pricing Short- and long-term signals Transmission pricing is the allocation of the regulated annual revenues of the transmission activity to the network users. The rst attempt to design these prices should be to resort to nodal energy prices, since nodal prices are perfectly ecient short-term signals, that is, geographically dierentiated short-term marginal costs of energy (see Schweppe et al., 1988). Nodal prices of energy implicitly include the eect on prices of losses and congestion in the network. They send adequate signals for decisions concerning the economic operation of generators and loads.
245
Strict application of nodal prices to generators and loads results in a net amount of revenues, which should be applied to partly pay for the cost of the network. Under ideal circumstances, impossible to nd in practice, these revenues would suce to pay the total network costs fully. However, as indicated before, these revenues are usually very insucient to cover the total network costs (cost recovery by nodal energy prices typically does not exceed 20 per cent of total transmission costs) (see Prez-Arriaga and Rubio, 1995). Thus, additional signals are needed to recover the remaining transmission network costs. These costs have to be assigned to the network users so that distortion of economic eciency is minimized. Therefore, in the rst place, these signals must be long term, so that they will not interfere with the nodal prices. This can be achieved by designing them as annual charges (although they may be distributed monthly, for instance). Ideally these long-term signals should be consistent with the underlying cost function of the transmission activity, so one should design them according to the driver behind transmission investment. This implies that cost causality, that is, cost responsibility, should be applied as much as possible, (see Chapter 6 of this book by Smeers, and Prez-Arriaga and Smeers, 2003). In the new competitive regulatory framework, investment in a new transmission line is justied whenever the present value of the aggregated benets of all the network users (generators and consumers) is larger than the present value of the cost of the line. No existence of market power is assumed. Transmission cost allocation criteria Then, conceptually, the solution is to charge the network cost that is not recovered through nodal prices in proportion with the benets that the transmission network (either globally or line by line) provides to each one of its users.6 The resulting long-term economic signals have no purpose in the operation (that is, short-term) timeframe; they are only meant to provide locational signals to new generators and loads or to those considering retirement that is, to inform them about the transmission network costs that are incurred because they locate or have located in one part of the network instead of another one. In the long term the response of the potential new network users to the transmission charges (that is, whether they will decide to install) will depend on their expected prots, after transmission charges are duly included. Unfortunately, allocation of transmission costs to the economic beneciaries is plagued with diculties in practice. Most of the problems arise from the lack of adequate information about the generators in a competitive setting and the need to estimate the future behavior of the system, but also because it is dicult to evaluate the economic impact on the market
246
agents of each individual line in a well-developed network with some level of reliability-driven redundancy. This is why some measure of electrical use has frequently been adopted as a pragmatic approximation to benets (and it is also much easier to compute) (see Prez-Arriaga, 2002; Prez-Arriaga and Smeers, 2003). If we refer to the case of the internal electricity market in Europe, this is the prevalent line of thought in the Florence Forum, a series of meetings where representatives of the regulators and the main groups of stakeholders in the European system gather to discuss and agree on the regulation that will be applied at European level. Several algorithms have been proposed to compute the electrical use that each agent makes of the system grid (see Rubio, 1999). At present, in most systems transmission network costs are socialized and charged uniformly to all network users, or to generators and consumers separately, according to some prescribed ratios of the total network cost. Non-transaction-based transmission charges An important practical conclusion that is derived from the criteria of allocation of the long-term signals is that transmission taris should not be transaction based. Indeed, the adopted criterion of cost allocation has nothing to do with the commercial transactions that the agents are engaged in at a given moment in time, under the assumption of a working market that is competitive and with perfect information. Transmission taris may depend on the connection point to the network, on the nature of the agent producer or consumer on the amount of power injected into or retrieved from the network and on the time of injection or withdrawal, even on the economic benets that ideally a market agent could obtain because of the development of the network, but not on whether the agent, in a particular moment in time, is buying from or selling to a power exchange or via a bilateral contract, be it with a local or with a foreign agent. Avoid tari pancaking In the context of a regional market it is very important to recognize that what intuitively appears to be a fair transmission pricing rule may lead to completely wrong results. This is the case of the still prevalent rule worldwide of charging to an international power transaction that crosses N countries the corresponding charge of each country as if it were a national transaction, therefore leading to a piling up or pancaking of transmission taris. This seems a fair treatment from each individual countrys viewpoint, but it results in a tarication system that critically depends on the shape of political borders, rather than on the physical reality of networks and ows. This pricing rule has two major defects: (a) it is transaction dependent; and (b) the transmission tari that is applied to a cross-border
247
transaction is the accumulation of the taris of all the countries that have been crossed (the pancaking eect referred to above), instead of some kind of average regional tari which would have been applied in a truly open regional market without political borders. The correct approach to an ecient system of regional transmission pricing is the single system paradigm, that is, a pricing scheme that tries to get as close as is practicable to the transmission taris that would be applied if the entire region were considered as a single country. Transmission rights and market power There is the general belief among some experts in transmission regulation that transmission rights over transmission capacity should be allocated to those who pay this transmission capacity, in proportion to their network charges. We are of the opinion that transmission rights should never be explicitly allocated to those agents who contribute to the recovery of an already sunk cost. They should only initially be assigned to a private party, in proportion to its share in the payment of the cost of the line or lines, when this party is promoting the construction of the new line and as a way of helping the agent to partly nance its investment. Even when congestion rents or the revenues resulting from the sale of transmission rights contribute to the reduction of the fraction of the sunk cost of a line to be recovered by other means, there is the possibility that market agents enjoying market power get transmission rights that provide them with an incentive to exercise this power (see Gilbert et al., 2002). If there are congestion rents associated with a line and these rents are used to reduce the transmission tari of those agents that pay for the line, then the network users are already receiving the same economic benets as if they were holding the same transmission rights. Therefore, it is not advisable to use the congestion rents from a line whose owner has the recovery of its investment guaranteed, for any other purpose than to reduce the amount of network charges in the system as a whole. However, this topic is still largely unexplored. Recommendations for Practical Implementation Some recommendations can be derived from the basic principles that were presented in the preceding subsection. If the transmission network is well meshed and there are no clear locational signals to be sent, because generation and load are more or less evenly distributed and no systematic congestions are likely to occur, then the beneciaries (or major users) of the network cannot be clearly identied on the basis of their location. According to economic theory, in the absence of a clear indication from the
248
underlying transmission cost function, it makes sense to refer to the inverse price elasticity rule (that is, the concept behind Ramsey pricing) in order to minimize the loss of eciency. This rule must provide an indication of how to split the global charge between generators and consumers and then also how to charge individual consumers on the one hand and generators on the other. Regarding allocation to the individual consumers, the inverse elasticity rule would advocate charging more to the least elastic consumers.7 Note, however, that this may be considered to be an unacceptable discrimination (see Chapter 6 in this book, and Prez-Arriaga and Smeers, 2003). Assuming that there is strong competition on the generation side and that generation is perfectly adapted to the system, due to totally free entry and exit by generators, so that whenever a generator is no longer protable it will leave and accordingly generators will enter the market whenever there are business opportunities to take advantage of, the rule advocates charging transmission costs mostly to consumers, since generation will be very elastic to prices and in the long run the large elasticity of generators will result in a complete transfer of the charges to the consumers (energy prices will rise or decrease as necessary in order for the generators to recover the transmission charges they are paying from consumers so that they will make no prots in real terms, once the average rate of return for an investment in a sector with the same level of risk is deducted). Note that this is not a trivial or universal rule, although it is a common misperception that: consumers always pay all network charges in the end. A simple example will illustrate this fact: consider a project to build a new generator with noncontested access to an inexpensive energy source in a remote location. Assume that transmission pricing rules are such that the generator is charged a large fraction of the transmission line that would connect it to the major load centers without the project becoming unprotable. In this case the network charge will not ultimately be transferred to consumers, since it will be fully absorbed as a cost of the new generation project. If the transmission network is such that long-term locational signals are needed and they can be more clearly identied because of systematic structural limitations of the network then the allocation of transmission costs should pay attention to location. Note that these long-term signals are no longer useful for existing generators and loads (except for those considering retirement because of economic reasons); they are meant to encourage new facilities to select adequate sites and to fully recover network costs.8 However, for the sake of simplicity and to avoid any appearance of discrimination, most regulators choose that both the existing and the new network users must be subject to the same charges. Following on from what was said in the previous point, if generation is perfectly adapted due to free entry and exit, then it is not very important how much is recovered through
249
generators and how much through consumers, since consumers will end up paying the entire bill. What would matter in this case would be the dierences in charges among generators when they are placed in dierent locations, so that they have the right incentive to locate in the network and, similarly, the dierences in charges among consumers. However, if the conditions presented before are not fullled, the absolute value of transmission charges will matter, since generators will not be able to completely transfer these charges to consumers and the absolute value of taris could also aect the decision of a generator to install. Both situations may take place in the individual system or at national level. In those countries where it is deemed that there is little need for long-term locational signals in transmission, transmission costs may be allocated to generators and consumers without any geographical dierentiation. This seems to be the case in most European countries. On the other hand, in those countries where long-term location signals appear to be necessary (for example, England and Wales, Norway or Sweden in Europe; Chile, Argentina and Australia are also good examples), transmission charges could have geographical dierentiation. These criteria are equally valid in a regional or multinational context. If geographical dierentiation of the long-term signals is not a major concern, then uniform regional transmission charges for generators and consumers could be applied in strict application of the single-system paradigm. However, this would require a very high level of regulatory integration and a pragmatic alternative could be to let each country charge its national taris to its network users, who in this way would automatically gain access to the entire region. However, the opposite situation may also be possible. At regional level, one may also want to send long-term signals in order to indicate the most appropriate and inappropriate zones to locate new generation and load. If the locational problem is a serious one that is, the economic utilization of generation resources at regional level to meet the regional load causes much stress in the existing transmission network then the long-term locational signals are needed. A rigorous approach would consist of assigning the cost of each one of the lines in the region to those agents that use it (or benet from it) while ignoring any political borders. However, this regional tarication scheme may only be possible in markets with a very high level of integration. Less radical alternatives are possible, such as replacing the nodal allocation of transmission costs at regional level by compensation mechanisms among countries, which would be based on how much each country uses (or benets from) the networks of other countries. As indicated above, complete socialization of transmission taris (that is, the postage stamp tari) does not contain any locational signal. More
250
cost-reective approaches, based on either the computation of the electrical use such as the areas of inuence method in Chile or Argentina or some estimate of the responsibility in network investment such as the investment cost-related pricing (ICRP) method in the UK or Colombia allocate the annual transmission network charge among all network users, according to some procedure that typically involves a high level of cost socialization among all users. A stronger locational component could be obtained if the transmission tari could reect the positive or negative incremental cost impact that the new network users impose on the transmission network. Thus, we may want to develop a more aggressive set of transmission network charges, that is, one with a stronger locational component. This topic is discussed in the next subsection. A Proposal to Compute Incremental Transmission Charges for New Network Users The basic concept The emphasis here will be to identify the incremental grid cost that a generator causes the system to incur when locating at a given node, while keeping in mind that the nal driver behind the network charges should be cost causality (see Prez-Arriaga and Smeers, 2003). One should be aware of the frequent discrepancy between network charges and the volume of network usage or benets that are obtained from the network or individual lines by its users. On one hand, the total amount of income that has to be collected to remunerate the owner of any of the lines must equal the regulated annualized cost of the line, which is the same for every year. However, due to the existence of economies of scale and the lumpiness of investments in the grid, the lines that are actually built may not coincide with the volume of investment that is strictly needed at a given time. In principle, one should avoid making new agents responsible for the transmission capacity in excess of what is necessary just because of their connection, since this excess capacity is actually meant for future network users that obviously cannot be charged now for it. For the purpose of this analysis grid reinforcements will be classied into three types: (a) connection lines, that is, lines that are built for the exclusive use of a generator (or consumer) so that it can be connected to a suitable point of the common network; (b) lines that are built to meet the needs of two or more network users; and (c) investments that are needed to extend the useful life of transmission facilities (since they are almost always replaced but not retired). Only a reduced subset of all the lines in the system are connection lines, their length is typically small when compared to the remaining lines and
251
their use and therefore their cost can easily be allocated to specic generators or demands. Network reinforcements of type (c) are meant to replace existing facilities that arrive at the end of their physical life; new agents should not be assigned more responsibility in the investment for these lines than already existing agents. The cost of these lines could be allocated in proportion to some measure of the average use of the line by each agent. It is more dicult to assign the responsibility for lines in the second group (b). Here it is argued that the allocation mechanism should reect the fact that new generators are more responsible than the older ones for the construction of those lines that have recently been built in response to the need for new network expansion in the system. In the same way, a generator or a load leaving the system (in generators this is typically because of retirement) may alleviate a network constraint or, on the contrary, increase its severity, and this should also be reected in a transmission credit or charge. The decision by a generator to install in a mainly importing area (or a demand to install in a mostly exporting one) may reduce the need for new investment in the grid. In this case, a sound transmission pricing mechanism may result in a negative charge (that is, a credit) for that generator or demand, thus making it attractive for new generators (loads) to install in importing (exporting) areas, since this will promote a more ecient use of the existing grid. The proposal The method that is proposed here employs the installation time of lines and generators and some measure of how much each agent uses or benets from the network as the input to compute transmission charges. The method is applied only to lines of type (b) above. Any agent requesting connection to the grid should pay the cost of its direct connection line. When the connection line has extra capacity so that future agents may connect to the same node, the cost of this extra capacity can be socialized to demand. As for those network reinforcements of type (c) that are built to extend the useful life of a given line, the same transmission pricing rules that were applicable to the original investment should be applied now. Figure 7.1 graphically shows the process of computation of the locational signals that result from the allocation of the regulated revenue of the lines of type (b) in the system to the network users. Several choices are oered to the regulator regarding the share of the cost to be jointly borne by consumers and generators. General criteria to decide on how to make use of these choices in each particular case have been presented before, although in practice this has a certain degree of arbitrariness. A rst choice
252

Cost of the unused fraction of the line is socialized (preferably to demand) 1 G1 Paid by generation Cost of the used fraction of the line G2 1 CG2 G1 G2 1 + kl kG2 G1 G2 CS
Total cost of line l (CT)
....
Gng1 C Gng1 Gng1 1 + kl kGng1 Gng Gng D1 D2

Modify participation factors according Dnd1 to how old is the line and each generator Dnd Scale down participation factors
....
....... Same process for loads
Gng1 CS Gng
Paid by demand
Allocation of the use made of the line Cgi: usage factor for generator i
Figure 7.1
Process of computation of locational signals
concerns the possibility of socializing to demand the cost of the fraction of the transmission capacity of each line that is not actually used,9 according to some agreed denition. In Figure 7.1, is the used fraction of the line and 1 the unused part. Next, the cost of the used fraction of each line is divided into a fraction to be paid by generators and the remaining fraction 1 to be paid by demand. Again this is a choice to be made by the regulator using the general criteria that were discussed before. The cost of each one of these two parts is initially allocated to the corresponding market agents in proportion to their share in the utilization of the line (CGi for generator i and CDj for load j). What is new in the method being proposed here, is that these participation factors are modied according to how long ago the line and the agents (generators and loads) were connected to the system. This is accomplished through the use of coecients representing how old the line and each one of the agents are (kl for the line, kGx for generators, where x is the generator number, and kDy for loads, where y is the load number). Thus, the participation factor of an agent in the recovery of the corresponding part
....
....
253
of the cost of line l (where this corresponding part is .(1 ) CT1 for generators and . .CT1 for loads) comprises three components. The rst one is simply the agents share in the use made of the line: CGi for generator i, for instance. The second component is a function of the time for which the line, and the agent itself, have been in service (kl and kGi, respectively, for line l and generator i, where these are input parameters whose value has to be dened a priori).10 Finally, these participation factors are scaled down so that the sum of the participation factors of all the generators is 1 and the same for the loads. Therefore, the expression for the total usage charge, corresponding to line l, that generator i must pay is: PGi CT1 (1 ) CGi (1 k1kGi ) CS, (7.1)
where CTl is the total cost of line l, is the used fraction of the line, is the fraction of the cost of the used part of line l to be recovered from consumers (and therefore 1 is the fraction to be recovered from generators), CGi is the usage factor corresponding to generator i, kl is a factor representing how old line l is, kGi is a factor representing how old generator i is and CS is the scale factor used to modify participations so that all of them add up to 1. The expression for the scale factor is: CS
ng
(1
i 1
1 . kl kGi )CGi
(7.2)
We are proposing here that cost causality should be interpreted in an incremental rather than a plain average way, so that new agents are assigned more responsibility for the most recent network reinforcements. This results in the use of coecients kl and kGi so that new generators and loads pay a larger part of the cost of new lines than those market participants who have already been operating for a number of years. Thus, the factor used to compute generator is contribution to the recovery of the cost of a line l results from increasing generator is usage factor (CGi) by the proportion represented by the product of coecients kl and kGi. Discussion The major dierence between marginal methods of transmission pricing and average ones is that the former result in both positive charges (when the installation of the agent creates the need for new grid investments) and negative ones (when the agent reduces the amount of network investment that is needed). Marginal methods send stronger locational signals typically, although they require a higher level of arbitrary assumptions than
254
average methods (see for details: Prez-Arriaga, 2002; Prez-Arriaga, et al., 2002). The approach that is proposed here can be used with any type of method for network cost allocation, either marginal or average, but it introduces the notion of temporality in the allocation, so that the link of causality between recently installed network facilities and new agents is emphasized. The method is supposed to produce stronger locational signals, where strong does not mean that the numerical value is larger, but that the relationship of causality has been reinforced. The method can easily be adjusted to emphasize more or less this temporal factor. The method can equally be applied to compute transmission charges for new generators and consumers. However, it is not expected that consumers will be very responsive to these economic locational signals. Besides, many countries have decided to apply uniform electricity taris for every class of end consumer. If this is the case, at charges may be applied to consumers and the proposed method could be applied only to generators. The concept of introducing the temporal dimension in the causality relationship when computing transmission charges can be found in the regulation of PJM in the United States (see Joskow, 2005). Transmission charges for generators corresponding to new grid investments may dier widely in PJM from one area to another, thus sending important locational signals to new generators. However, unlike the scheme of signals proposed here, grid charges to generators in PJM are computed once (at the moment the generator is installed). Furthermore, charges paid by a new generator cover only the cost of facilities built because of the construction of the generator as well as those lines built shortly before the installation of the generator from which it is beneting. In other words, new generators are not held responsible for the cost of lines built after the installation of these generators unless these lines were already projected when the generators decided to install. Responsibilities in network cost are evaluated by computing the incremental impact of the installation of each new generator on the ow over congested lines or corridors that must be reinforced. Temporal dierentiation plays an important role as well in the proposal for the computation of grid charges for the Peruvian power system by a group of international consultants that included the rst author of this chapter (see Mercados Energticos, 2005). An interesting feature of the approach that has been proposed for Peru is that it employs two dierent network cost-allocation schemes, one for the existing lines and the existing network users (method A) and a second method for the new lines and network users (method B). The point here is there is little use in sending economic signals to consumers and generators that are already connected to the grid, so there is not much interest in perfecting any current transmission pricing method A that is already in use for existing lines and
255
network users. However, one could aim even for the most adequate method B allocation to the economic beneciaries when it concerns only the new lines and the new network users. The major diculty with the method of allocation to beneciaries going deep into the past and the future guessing what the economic benets were or will be disappears if there is no need to go into the past and if the exploration into the future is limited in time. Limiting the exploration into the future can be achieved by returning to method A once some years have passed and what were new lines and network users can be considered to be existing ones. Another approach the one that was adopted in the Peruvian study consists of computing the future transmission charges of the new lines for the new users once and for all, not subject to future revision. This also has the positive eect of eliminating any uncertainty of the new users with respect to any possible changes in transmission charges in the future. Locational signals for retiring generators In order to encourage market agents to make the right decisions in the long term we must send them economic signals so that market agents realize the cost that the system incurs as a consequence of either their installation or their retirement. The previous subsection discussed a system of economic signals aimed at encouraging market agents to install in the best location from the system point of view. This subsection presents a scheme of signals that is meant to induce those agents considering the possibility of retiring a power plant to take into account the grid costs that may result from this decision. Computing the eect that the decision of retiring a plant has on the number and nature of future grid investments is very dicult. Besides, the application of economic signals that are based on costs not yet incurred by the system is very problematic from a regulatory point of view. Therefore, it seems more appropriate to derive signals to market agents from the impact that their decision to leave the system would have on their current electrical usage of the existing grid. It is proposed here to follow for retiring plants exactly the same basic approach that has been proposed for new entrants. There are two major possibilities. The rst one is to hold the agent responsible only for the change in the use of any congested lines, which will probably have to be reinforced in the near future if the plant is retired. The second one is to consider that the agent is responsible for the change in the use of any line in the system, whether congested or not. It seems better to compute the grid charges taking into account only the congested lines, since there is no evidence that other lines will have to be reinforced because of the decision of the agent to leave.
256
In summary, the same method presented above to send strong locational signals to new generators is also proposed to compute the grid charges associated with the retirement of an existing plant. The retiring generator would be regarded as new as well as the congested lines and the lines that have actually been installed recently. The signal should be implemented as a single payment to be made when the agent retires the plant. Other Locational Signals for Generators and Network Planners The previous subsections have discussed the design of a system of transmission charges that could send meaningful economic signals to existing and future network users, while recovering the total network costs. However, there are other locational signals of a dierent nature that generators and also system planners could send and receive. These signals are mostly related to information that helps in removing the uncertainty that accompanies investment decisions in both network and generation. As we have seen, the development of the grid must be planned in conjunction with that of the generation and the demand in the system. The location of new power plants and loads inuences the decision by grid planners on which are the most interesting reinforcements. Likewise, market agents should take into account the limits that are imposed by the existing grid and the investments in new lines or other transmission facilities that are planned for the future. Both the network planners and the potential investors in new generation need to know what the other party will do with as much certainty as possible. Therefore, it is important on one side for the network planner to have a reliable indication of the level of commitment of new agents that ask for connection to the grid, since network expansion should be specially addressed to meet the needs that result from new network users. Consider as an example the case of the Spanish power system, with a peak load of 43,000 MW in 2005 and where the system operator has standing requests for new connection of more than 50,000 MW of wind generators and more than 50,000 MW of CCGT power plants. Most of these requests will not materialize, since their purpose is just to make sure that a particular site has met all the administrative requirements to install new capacity a procedure that usually takes several years. A regulatory measure that can help the system operator to sort out the more serious requests from the merely speculative ones is to require from the potential investor a nancial guarantee, in order to start the administrative process of authorization of the connection to the grid plus any environmental or other kind of permits that may be necessary. The guarantee will be returned only if the power plant is actually built within a prescribed time limit.
257
On the other hand, the system operator should provide the market agents that are considering the installation of a new plant, with as much information as possible regarding the future development of the grid including the projected importing and exporting capacity with neighboring systems, the expected value of transmission charges at each node as well as the loss factors and the nature, the location and severity of the restrictions that the grid is anticipated to impose on the operation of the system. An adequate time horizon for these estimations should be about ve or ten years. This information should be made available in some kind of N-year-ahead transmission report to be issued annually. It is also very important that the regulator and the system operator make clear the network access rules to all potential generation investors. Access rules encompass the management of network constraints as well as the criteria to grant connection to the transmission network. Ill-conceived access rules may deter new generation investment by creating too much uncertainty or, at the other extreme, they may provide excessive advantages to the rst new entrants with respect to future ones. For instance, a reasonable access rule for generators may specify that connecting rst to a given node of the transmission network does not provide any advantage regarding priority of access to the network, in case the network cannot accommodate all the power that the generators that are connected to that node are willing to inject. An investor that considers building a new power plant and to connect it to this particular node should expect competition with the generators already connected just for the use of the limited capacity of injection at the node. FTRs may be used by the existing and the future power plants to manage the economic risk of being displaced from producing at a node by more competitive generators, therefore creating more favorable conditions for investment. Note that, in general, if there is a bottleneck, the network planner will decide to reinforce the grid if it is economically and technically justied so that the congestion is conveniently reduced or even eliminated. Therefore one should not expect this type of situation to be frequent, unless some of the existing generators at the node are very inecient and network reinforcement in this case is not justied. Each power system may choose dierent access rules from the ones that have been presented here as an example. The message is that access rules are an essential ingredient of the information that potential generation investors need to make sensible decisions whether to invest in a given power system and whether to connect at a given node of the transmission network. Access rules and transmission charges can be used to guide the process of connection to the grid of new loads in a meaningful way. The applicable legislation in many countries allows agents to apply for connection at any voltage they choose and the network planner is then forced to comply.
258
Some agents may take advantage of this situation to get connected to voltage levels where they know they will pay fewer network charges or have other types of advantages, but which may not be convenient for the overall design of the distribution or transmission network. One possibility to guide the location of new agents is to charge them according to the voltage level where they should be connected depending on their peak demand and total consumption, for instance instead of the voltage level where they are actually located.
5.
COMPATIBILITY OF INVESTMENT SIGNALS IN TRANSMISSION AND DISTRIBUTION
The separation between transmission and distribution networks is not a clean one. Most European countries have adopted the simple and, up to a point, arbitrary decision to dene transmission as encompassing all network facilities operating at 220 kV and higher (400 kV, typically) and, also frequently, all interconnection lines with other countries. Distribution would correspond to all other networks at lower voltages. Obviously, there are some interactions at the interface between both networks. For instance, depending on the physical conguration of each particular network, some further development of the 220 kV grid may help the underlying highvoltage distribution network to avoid some investments. Similarly, a strong high-voltage 132 kV subtransmission grid may replace some investments at 220 kV level. Traditionally, the regulatory approaches to the transmission and distribution activities have been very dierent. There are some good reasons for them to remain so. First, transmission and distribution grids perform dierent functions. Transmission makes sure that major generators are well connected to major load centers. Flows over the lines in the transmission grid change direction frequently. Transmission has a great impact on the implementation of the wholesale market. On the contrary, if we leave aside the existence of distributed generation which in some areas may outweigh local demand so that the corresponding distribution network is a net exporter power over the distribution grid always ows in the same direction: downstream from transmission to the end consumers. The operation and planning of distribution networks as a whole are mostly inuenced by the location and load proles of nal consumers. On the other hand, the internal conguration and operation of distribution networks has a negligible impact on the wholesale market and the conguration of the transmission network. However, at higher-voltage levels distribution grids tend to be meshed and
259
their conguration can be adapted to the particular operating conditions by opening and closing the corresponding devices. In some cases the power ows in the transmission network cannot be fully understood without considering the impact of the subtransmission network. Besides, more and more generation is connecting to the distribution grid at all voltage levels. Therefore, one can easily conclude that, in many cases, those parts of the traditionally considered distribution grids that are immediately below transmission substations are really performing transmission grid functions. The dierence between transmission and distribution grids, especially at the border between them, is becoming increasingly blurred. Second, the number of lines and other network facilities in distribution grids is much higher than that of transmission grids. In transmission each individual investment can be considered separately for authorization and remuneration whereas in distribution, particularly at lower voltages, this would be an extremely tedious and meaningless task. Performance-based regulation is a preferable choice for distribution networks. Third, transmission has an important role in maintaining the overall security of the power system. Failures at transmission level are very infrequent. However, when they occur, there is usually a large disruption in power supply. Failures at distribution level are much more frequent. The quality of supply of most consumers mostly depends on the distribution network. Therefore, the regulation of distribution should heavily relate to quality of supply considerations. The above-mentioned broad dierences that exist between transmission and distribution in many aspects should result in a specic treatment of remuneration, planning of the expansion, operation and, quite importantly, the short- and long-term signals to be sent in each one of the two cases:
Regarding the remuneration of the activity, there are implications both for the computation of the regulated revenues of the owner of the grid and for the split of these revenues into the part to be collected from consumers and the one to be collected from generators. In the absence of distributed generation, consumers should pay 100 per cent of the distribution network costs. According to cost causality, generators connected to transmission should not pay distribution network costs, since the design of distribution networks is not inuenced at all by the nature or location of the generation power plants. On the other hand, the costs of the transmission network should be shared by generators and consumers (here we mean both large consumers that are directly connected to the transmission network and smaller consumers that obtain their power from the
260
distribution networks). When distributed generation (that is, generation that feeds into the distribution network) cannot be absorbed locally and feeds into the transmission network, it should also pay transmission network charges. Given the ambiguity that exists when dening both the border between transmission and distribution and the responsibilities in expanding the network, the approaches used to compute the remuneration of the transmission and distribution companies at least should be consistent with each other.
As said before, performance-based regulation should be employed to determine the revenues of the owner and developer of the distribution grid. In the case of transmission, currently most of the countries compute the income of the transmission company or companies through some sort of cost of service regulation. This lack of coherence may result in a suboptimal development of both grids. We have already mentioned the natural inclination of system operators and regulators to overbuild the transmission network. Moreover, when faced with a remuneration scheme that is based on the cost of individual transmission facilities or on the winning bids in a public auction, transmission companies or TSOs will be happy that as many lines as possible are built. On the contrary, provided that the remuneration of the high-voltage distribution company is some kind of RPI-X, they will prefer that the needs of reinforcement of the subtransmission network could be met whenever possible by reinforcements at transmission level, in order to reduce costs without hurting the overall network service. As a consequence of this, too much transmission grid may be built that will perform a function that would have been better accomplished by the distribution grid. Regulators should be aware of these distortions in order to minimize them by adjusting the corresponding remuneration mechanisms. However, a complete elimination of these distortions does not seem to be possible because of the discontinuity of the regulatory treatment at the border between the two types of network. Ideally, planning transmission and distribution grids should be coordinated. For instance, the decision on the location and number of the transmission/distribution substations at the border between both grids is critical in this regard. An iterative process, similar to the method that is used in some cases to determine the location of substations and transformers between dierent distribution grid levels, could provide the location and number of the substations between transmission and distribution as well (see Peco, 2001). In the absence of any joint planning in practice, coordinating economic signals might be useful. These locational signals could be applied to the end
261
consumers and also to the distribution utilities. As shown before, in general network charges for consumers in the distribution grid should include both transmission and distribution costs, whereas those aimed at consumers and generators at transmission level could ignore the distribution grid. This is true as long as distribution grids do not contribute to transporting the power output from distributed generation to load centers located in other distribution networks. Locational signals in relation to transmission investment have already been discussed in Section 3. The next paragraphs explain how nodal prices could be employed to send incentives for investment both to end consumers and to the distribution companies. Since the distribution and retailing activities should be unbundled in a sound regulatory framework, the distribution utility does not purchase electricity. However, the point here is that regulatory incentives for the distribution utility should be designed so that the price of acquiring at each moment in time from the transmission network the power that is needed to physically supply the end consumers of the distribution company is as low as possible. A regulatory mechanism to accomplish this is to make the distribution company responsible for the dierence between the actual cost of purchasing electricity at hourly nodal prices from the corresponding transmission nodes and the amount of electricity consumed by the end consumers of this distribution company multiplied by some average loss factor and some average hourly energy system price. This is a generalization of the regulatory mechanism that is frequently used to provide incentives to distribution utilities to reduce network losses. The rst level of implementation of this scheme of locational signals, with eects over the development (and also operation, see below) of the distribution grid, would consist in letting the distribution company assume the economic dierence between some reference purchasing hourly price for the energy demanded by the consumers of the distribution company (aected by some correction loss factor) and the pretended purchase (since the power is actually bought by the retailing company) of the energy from the transmission system at the corresponding nodal price, which would be dependent on the point of interface with the transmission grid. Under this scheme, the distribution company would react by planning the development of the grid in order to minimize not only the cost of losses over the grid, which has traditionally been its major objective, but also the cost of purchasing wholesale power. Nodal prices can also encourage ecient operation of the distribution network. Similarly to what happens with the planning of the grid, pretending that the distribution company buys the energy for its customers at the corresponding transmission hourly nodal prices would lead the distribution utility to operate the grid in order to minimize the cost of purchasing the wholesale power at any given time.
262
By operation of the grid we mean here the possibility of modifying the conguration of the subtransmission grid. Normally, this grid is well meshed because of security, quality of supply and stability reasons. However, at least part of this grid is operated as a radial network by suitable switching. Changing the network conguration during the operation allows the distribution company to modify the incurred losses and, if they buy power from the transmission grid at nodal prices, to modify also the amount paid for this power as well. Some indication of the quantitative relevance of these locational signals can be found in Carillo-Caicedo and Prez-Arriaga (1995). A second level of implementation of a scheme of nodal prices would be to pass through the wholesale power prices to the end consumers. Computation of nodal prices at the end consumer nodes of the distribution grid has to be discarded because of the high volatility these prices may experience, which would make them unacceptable (see Outhred and Kaye, 1996). It would be possible, however, to charge each end consumer a zonal price based on the nodal prices at the node or nodes of the transmission grid where the distribution company buys its power. These zonal prices could be obtained by averaging, over the period of time for which the prices are computed, the nodal prices at the dierent nodes of the transmission grid weighted by the amount of energy that the considered distribution zone is obtaining from each transmission node. As mentioned before, the massive introduction of distributed generation will likely blur the dierences between transmission and distribution regulation, especially if this generation ends up being partially fed into the transmission network. Then, the development of transmission and distribution grids will become much more dependent on one another. The planning of transmission and high-voltage distribution grids would probably have to be carried out jointly or, at least, would involve a much higher level of coordination. Transmission and distribution grids would perform much more similar functions and, therefore, the regulatory treatment of transmission and distribution would have to evolve accordingly.
6.
CASE EXAMPLES
This section presents numerical examples of the most relevant concepts that have already been presented. The rst case shows how nodal transmission taris can be computed, both at national and multinational levels, and also how the same technique can be employed to determine economic compensations between countries or power systems because of the use that agents in one country make of the networks of other countries. The potential of
263
these locational signals to guide transmission and generation investment is examined. The second case example shows how nodal transmission taris are just part of an ensemble of other locational signals and what the expected weight of each one of them is. Nodal Transmission Taris in the Internal Electricity Market of the European Union This subsection presents numerical results that help to illustrate some of the ideas that have been discussed in the chapter within the context of the internal electricity market (IEM) of the European Union. As mentioned before, nodal (that is, with locational content) transmission taris, together with nodal energy prices, may inuence the decisions by market agents on whether to install new generation and consumption capacity (or retire existing capacity) and where to do it. In addition, nodal transmission taris and nodal energy prices can also encourage coalitions of network users (generators and loads) to promote the construction of new network facilities. This may happen in dierent ways, depending on who is ultimately responsible for the development of the grid and the role that the prevailing network regulation assigns to the several market agents in this process. In Europe, the adoption of a common system of transmission taris appears to be too ambitious for the time being. To start with, the regulator of each country has approved regulated transmission costs using widely dierent methods, and also has adopted widely dierent procedures to compute the transmission taris for the network users. The end result is a total lack of harmonization in the computation of transmission charges (see Prez-Arriaga et al., 2002). However, a common procedure of interTSO payments has been established to compensate each country for the external use of its network by agents in other countries. The net value of the compensations must then be used to modify the local taris calculated by the regulator of each country (ibid.). A method that is able to compute nodal transmission taris at regional level must also be used to determine the net volume of inter-TSO payments that each country has to receive. In order to do so, one only has to aggregate the resulting nodal transmission taris at country level, thus directly obtaining how much the agents within each country use their own grid and the grid of others. After exploring dierent methods of allocation of grid costs to individual network users (that is, of computation of transmission taris), all of them based on some measure of electrical usage, we have reached the conclusion that the method of average participations (AP) is the best one among those currently available for this specic application (see ibid. for a
264
description of the AP method and a comparison with other approaches). Appendix 7A2 provides a succinct description of the AP method, since this is the one that has been used in the numerical examples that will be shown below. However, one must bear in mind that this method is one among several electric usage-based methods that are currently available to compute transmission taris or inter-TSO payments and that other novel interesting methods may be proposed at any moment (see, for instance, Florence School of Regulation, 2005). We have used the AP method to obtain nodal transmission charges for each of the 3,965 nodes of a model comprising 13 countries that belong to the IEM. Twenty-four dierent scenarios, corresponding to the real operation of these 13 countries throughout the year 2002, have been used to compute the annual taris. The data have been provided by the association of European Transmission System Operators (ETSO). Some other data have been used to obtain the results here presented. We have had to estimate the annual cost per km of a 400-kV line in order to express both compensations among countries and nodal transmission taris in monetary terms. First, we computed the number of kilometers of equivalent 400 kV lines within the horizontal network of 11 European countries: Austria (A), Belgium (B), France (F), Germany (D), Italy (I), Portugal (P), Spain (E), Switzerland (CH), the Czech Republic (Z), the Netherlands (NL) and Slovenia (SLO). The gures obtained were 3,896 km for Austria, 3,930 for Belgium, 4,477 for Switzerland, 5,116 for the Czech Republic, 34,811 for Germany, 21,511 for Spain, 27,916 for France, 13,019 for Italy, 3,737 for the Netherlands, 3,627 for Portugal and 696 for Slovenia. Second, we divided the regulated cost of the horizontal network of each country (values were those used in the provisional method for the year 2002) by the number of kilometers of equivalent 400-kV lines within the network of the country. Thus, a regulated cost per kilometer of equivalent 400-kV line was obtained for each country. Finally, the standard unit cost of a 400-kV line used to express all the results in monetary terms resulted from computing the average of the regulated unit costs obtained for the dierent countries. An annual value of 0.0352 million per km of 400-kV line was obtained. From these nodal taris and for each agent and scenario an average annual tari has been obtained for every agent. In order to calculate the average tari for an agent, its nodal tari in each scenario has been weighed by the amount of power either produced or consumed by the agent. By default the AP method equally allocates the total ow over each line to all the generators and to all the loads in the system. Therefore 50 per cent of the use of the line (and consequently of its cost) is allocated to generators and 50 per cent to loads. However, the method easily allows for any other split of the cost of the grid between generation and load.

nodal_L_tariff
265
nodal_G_tariff
3.5 3 2.5 2 1.5 1 0.5 0 A B CH CZ D E F H
NL
SLO
SK
Note: A Austria; B Belgium; CH Switzerland; CZ Czech Republic; D Germany; E Spain; F France; H Hungary; I Italy; NL Netherlands; P Portugal; SLO Slovenia; SK Slovakia.
Figure 7.2
Average L and G taris in Europe
Figure 7.2 shows the average nodal L and G taris for the 13 countries that have been considered in the study. For each country, the L tari is represented by a light column and the G tari by a dark one. Figures 7.3 and 7.4 show the nodal L and G transmission taris, respectively, for all the nodes in the IEM-13 network. The average values for G and L for each country have been represented by a dotted line, whereas the average internal taris for each country (resulting from dividing half of the cost of the national grid by the total amount of load or generation in the country; this is the internal average transmission tari for each country before the application of inter-TSO payments) are represented by a dashed line. Instead of applying a system of pan-European taris like this, the main stakeholders in the IEM have agreed to establish a system of compensations among countries (inter-TSO compensations) for the use each country makes of the grids of others. The dierence between both horizontal lines in each country indicates the signicance of the impact of the inter-TSO payment mechanism on the transmission taris of each country. Numbers are expressed in /MWh. We can appreciate from the gures that the distribution of the average transmission taris per country is not too dispersed. The average values for G range between 1.03/MWh and 2.89/MWh, with an average value for the IEM of 2.02/MWh. In the same way, the average values for L range
266
Nodal tariff 10

Average nodal tariff Average internal tariff
0
A B C H C Z H E D N L P F I
Note: For country abbreviations, see Figure 7.2.
Figure 7.3
10
L nodal taris in Europe

Nodal tariff Average nodal tariff Average internal tariff
0
C H C Z N L P D A B E F H I
Note: For country abbreviations, see Figure 7.2.
Figure 7.4
G nodal taris in Europe
between 1.35/MWh and 2.89/MWh, with an average value for the IEM of 2.33/MWh. The signicant dierences among the average transmission charges in dierent countries are not only due to the dierent pattern of ows existing in each part of the grid but also to the fact that the amount
267
of transmission assets per MWh varies widely from one country to another. Inter-TSO payments among countries account for the dierence between the average transmission charge that would exist in each country if the impact of cross-border exchanges of electricity were ignored (dashed line in every gure) and the average transmission charges per country when the AP method is employed (dotted line). For instance, in this example Austria must receive a positive inter-TSO payment of 26 million. This is the net amount that other countries have to pay Austria, because of the dierence between the cost of the Austrian grid (represented by the average internal taris on G and L) and the total cost of the whole grid used by the Austrian generators and consumers (represented by the actual average G and L taris when cross-border ows are accounted for). The distribution of transmission charges within every country shows a large dispersion with respect to the average value. This is related to the dierent ways in which the same pattern of ows can aect the transmission charge computed with the AP method for two agents located not very far from each other. Due to the non-linear nature of the method, the unitary participation in the use made of the grid by an agent may critically depend on the size of G and L at the nodes. The AP method provides very reasonable results at the macro level (inter-TSO payments among countries) but at the micro level (nodal charges) the values are somewhat volatile and require some interpretation and further treatment. Now we shall focus on a single country Spain for instance in order to examine this issue. In general, generation and/or load nodal taris are similar in nodes located nearby (in the same geographical area). Outliers well above the average can be explained as corresponding to agents that either produce or consume a small amount of power and are connected to the rest of the grid through a dedicated line or lines (therefore, a large fraction of the cost of this line is charged to a small amount of power). Values well below the average are normally generators in a node where demand is predominant (mostly importing nodes) or conversely loads in a node where generation is dominant (mostly exporting nodes). One possibility to avoid the problem of excessive dispersion of the taris is to compute and apply zonal taris. In order to compute them, each nodal tari in a given zone would be weighted by the amount of power either consumed or produced by the corresponding agent and an average for predened geographical regions would be obtained. In this way, those extreme values would have a small inuence in the resulting tari to be applied to every generator or load in the area (see Figure 7.4). It is also possible to design methods to smooth out the individual nodal taris by simply taking into account the values of nearby nodes. This is the approach that has been adopted here. The next paragraphs explain the
268
method that has been used to get rid of outliers and to make the geographic distribution of transmission taris smoother. The basic idea is to compute the new transmission tari for a generation or a load node by taking into account both the original tari of this node and the ones of the nodes of the same type (either generation or load) surrounding it. Therefore, only the taris of the surrounding generation nodes are considered to modify the tari of a generation node and the same for demand nodes. A sound method for allocation of the sunk grid costs may probably result in taris that are quite dierent for the generation and the load in the same region, depending on whether the region has a decit of generation or load. G taris in a mainly importing area will probably be smaller than L taris and vice versa for a mainly exporting area, thus signaling the advisability of installing more generation in the former case and more demand in the latter. The objective is not to distort these signals by smoothing the G and L taris together. Therefore the smoothing process is carried out for generators and loads independently. The new tari for a given node A is computed as the weighted average of all the nodes of the same type as node A. The taris of its neighbors are weighted using a function that decreases with the distance to node A. The weighting factor that is used for the original tari of node A is 1 and it decreases the further away we move from this node. Therefore, nodes that are electrically closer to node A will have a stronger inuence in the new tari of this node than those which are located farther from an electrical point of view. The electrical distance between two dierent nodes is measured in terms of the equivalent electrical impedance between these two nodes. Figures 7.5 and 7.6 show, for the same IEM-13 system referred to above, the detail of the original G and L taris for the dierent provinces within Spain. This is represented by a thick line. The same gures show the corresponding new taris once the smoothing process is applied. This is represented by a thin line. Numbers are again expressed in /MWh. Finally, Figure 7.7 compares, for 17 European countries, the average transmission tari of a typical industrial customer to the net inter-TSO payment resulting from the application of the AP method to the same 24 scenarios referred to above (see Prez-Arriaga et al., 2002; Prez-Arriaga and Olmos, 2003). Numbers are expressed in /MWh. According to this gure, there can be a dierence of up to 7/MWh between the transmission taris in two dierent countries (see the dierence between the transmission tari in Sweden and Spain). This dierence is of the same order of magnitude as that between the energy prices in two power exchanges (see Figure 7.8), which has been provided by the Spanish power exchange OMEL. Since transmission taris are longterm signals that should not interfere with system operation, it is advisable

Province ORIG_T New_T
269
10
13.6 10.46
Badajoz
Burgos Cceres Cadiz Santander Ciudad Real Cordoba Guadalajara Guipzcoa Huelva
Hues La Coruna La Rioja Len
Alava
Malaga Murcia Navarra Orense Palencia
Salamanca
Tarragona Toledo
Barcelona
Figure 7.5
Original and new L taris in Spain for the IEM-13 system
to give transmission taris the format of a capacity (/MW per year) charge, or even a lump-sum charge (/year), but never an energy charge (/MWh), since this could introduce non-negligible distortions in the generation dispatch in the region. Figure 7.7 also shows that the net inter-TSO payment each country must receive (expressed again in /MWh) is much smaller than the typical transmission tari. Therefore, it is very unlikely that these inter-TSO payments will be able to send any signicant locational signal in the presence of widely dierent transmission taris in the considered countries. One must be aware of the fact that any method used to compute interTSO payments is, at the same time, allocating the cost of any line among the countries in the system. The importance of this feature should not be underestimated, since the adopted method will automatically allocate the cost of any new network investment among the dierent countries, according to the use that each one of them is making of this facility. This has been implicitly accepted when adopting the inter-TSO payment mechanism and it appears to be an excellent way of solving a problem that typically has proven to be a contentious one.
Valencia Vallodolid Vizcaya Zamora Zaragoza
Madrid
Sevilla
270

Province ORIG_T New_T 10 23.45
0
Salamanca Sevilla Tarragona Toledo Vallodolid Vizcaya Murcia Navarre Madrid Ciudad Real Granada La Coruna La Rioja Guipzcoa Palencia Pontevedra Cceres Cadiz Cantabria Barcelona Zaragoza Len Albacete Asturias Burgos Orense Huelva Alava Hues Lugo
Figure 7.6
Original and new G taris in Spain for the IEM-13 system
Case Example of Transmission-related Locational Signals This subsection compares the locational economic signals corresponding to two dierent areas in the transmission network of a ctive system where a potential investor might consider installing a new CCGT power plant. The rst location (L1) is close to a liqueed natural gas (LNG) regasication facility on the coast. The second location (L2) is close to a main load center. In both cases, the power plant is meant to meet the demand growth in the aforementioned load center. The most important locational factors can be grouped into three categories:
Locational signals related to the impact of losses, in both the electricity and the gas networks. Network constraints are ignored here since systematic congestions are infrequent in well-developed networks. Most systematic congestions are usually removed by new network reinforcements in the short or medium term since they typically have a signicant impact on the system operating costs.

12 10 8 6 4 2 0 Trans. tariff
271
Net inter-TSO payment
Note: Figures are expressed in /MWh. Per unit inter-TSO payments for Italy and France are too small to be visible since these are large countries with small transits. They are almost purely exporting (France) or importing (Italy) countries. No data were available for Greece and Luxembourg.
Figure 7.7 Comparison between the transmission tari and the net interTSO payment for 17 European countries
/MWh 70 2000 60 50 40 30 20 10 0 2001 2002 2003 2004
1 Trimestre
2 Trimestre
En
gl
3 Trimestre 4 Trimestre
an
1 Trimestre
2 Trimestre
3 Trimestre
4 Trimestre
1 Trimestre
2 Trimestre
3 Trimestre
4 Trimestre
1 Trimestre
2 Trimestre
3 Trimestre
4 Trimestre
1 Trimestre
2 Trimestre
3 Trimestre
4 Trimestre
Figure 7.8 Evolution of the energy price of several power exchanges belonging to Europex, January 2000 to November 2004
W al e Fi s nl an d Fr an c G er e m an y G re ec e Ire la nd Lu Ita xe m ly bo ur N et g he rla nd s N or w a Po y rtu ga l Sp ai Sw n Sw ede n itz er la nd

Date
APX NORDPOOL LPX POWERNEXT OMEL GME
m iu lg en D
st
Au
Be
m d an
ar k
ria
272
Those signals related to the cost of new network facilities both for gas and electricity that are built because of the installation of new agents in the system. Other locational signals that may exist because of various reasons. In this case we have to consider signals derived from the eect of the altitude over sea level on the useful generation capacity of the CCGT power plants.
With the purpose of facilitating the comparison of both options, the power plant has been assumed to provide the same amount of usable power to the system regardless of whether it is installed in one place or the other. Table 7.1 shows the eect of transmission losses, the provision of ancillary services and the altitude over sea level on the total generation capacity that is needed to supply an increment in demand of 384 MW in the load center. In both cases, the power plant must have the same useful generation capacity (see rst row of Table 7.1, Useful power delivered), which equals the assumed increase in the power demand in the consumption centre. This gure is then modied to take into account the dierent increments in electricity transmission losses that the system incurs when the same extra amount of usable power is produced in the two locations (see second row, Net power output). Including the ancillary services that each power plant must provide leads to a 4 per cent increase in its generation capacity (see third row, Nominal capacity). Finally, we consider the eect of the altitude of each location. Whereas the plant near the LNG facility is located at sea level, the power plant close to the main load center is supposed to be Table 7.1 Impact of dierent factors on the total generation capacity needed to supply a 384-MW load, located close to a main consumption center, from two dierent locations, one close to the load center and the other close to an entry point for LNG
Close to main load center (L1) Useful power delivered (MW) Net power output (MW) (transmission losses) Nominal capacity (MW) (ancillary services) Nameplate capacity (MW) (altitude over sea level) 384.0 385.1 401.1 422.2 Close to an LNG facility on the coast (L2) 384.0 407.8 424.8 424.8
273
located considerably higher, thus experiencing a reduction of 5 per cent in its total generation capacity (20 MW for a plant with a nameplate generation capacity of 400 MW). Therefore, in order to be able to produce 401.1 MW, that is, the required nominal capacity of the plant in L2, the nameplate capacity of the power plant must be 422.2 MW. On the other hand, the power plant in L1 does not experience any eciency losses related to its location because the eect of the altitude is negligible there (see last row, Nameplate capacity). Table 7.2 shows, for both locations, the dierent cost components involved in meeting an increase of 384 MW in the demand for electricity of the consumption centre close to L2. The useful life of the power plant has been assumed to be 20 years and the cost of capital per annum 6 per cent real. The price of electricity production with natural gas has been assumed to be 17/MWh for the entire time horizon of the study (20 years) and the thermal eciency of the power plant has been assumed to be 50 per cent in both cases. We estimate that in both locations the power plant would be functioning 5,000 hours a year at its full capacity. The per-unit investment cost of the considered CCGT power plant is 1.208 million per megawatt of installed capacity. This gure includes the cost of some dedicated facilities assumed to be needed to connect the power plant to the main electricity and gas systems (both the electrical transmission line and the pipeline connecting the plant to the substation and the main pipeline system, respectively, have been supposed to be 40 km long). Electricity transmission losses and the provision of ancillary services have an impact on the amount of energy produced as well as on the cost of construction of the power plant. The power plant must produce some extra amount of energy in order to supply the transmission losses and ancillary services allocated to it. This amount varies with the location of the plant. Moreover, the required capacity of the generation facility also depends on the loss of capacity because of transmission losses and ancillary services (see Table 7.1). The altitude over sea level has an eect on the required nameplate capacity of a CCGT power plant because it limits the maximum power output of the generator. Therefore, if this power output is prescribed from the outset, the higher the power plant is located above sea level the larger the nameplate capacity of the power plant must be. Consequently, all other factors being equal, generation investment costs increase with the altitude over sea level. Since the nal amount of energy produced and the fuel consumption are the same regardless of the altitude, production costs are therefore not aected by this factor. Regarding the losses incurred when transporting the gas from the point of injection to the place where it is consumed, it has been assumed that
274
Table 7.2 Comparison of the cost savings involved in supplying a 384-MW load located close to a main load center
Market value of energy produced (m) L2 Annual impact of electricity transmission losses Annual impact of the provision of ancillary services Annual impact of the altitude over sea level Annual impact of losses in the gas pipeline system Annual impact of the use of the electricity grid infrastructures Annual impact of the use of the gas grid infrastructures Overall annual value (total annual value of the energy produced or total annual cost of the facilities built) 0.2 L1 4.3 Investment in generation facilities (m) L2 0.1 L1 2.4 Investment in transmission facilities (m) L2 L1
2.9
3.1
1.6
1.7
2.1
0.5
7.9
72.7
76.5
42.0
42.3
7.9
a fraction of the transported gas is lost. The assumed percentage can only be regarded as an estimate for the purpose of this example, since we do not know about consistent statistics on the relative importance of losses in gas pipeline systems. No gas losses are considered for the plant located at L1, since the plant is very close to the LNG regasication facility. On the other hand, losses incurred when the power plant is located near the load center at L2 are assumed to account for 0.7 per cent of the total gas consumed.
275
We have taken into account both the price of gas and the thermal eciency of the power plant in order to compute the cost of these losses, which aect the production cost of the power plant. As for the locational signals related to the cost of those electric transmission lines that the new power plant must be held responsible for, we have considered only new lines here. The cost of those lines that already exist will probably be allocated according to some average method which results in a socialization of their cost thus not giving rise to any locational signal.11 In order to assign the cost of new lines the novel method presented in Section 4 has been used here, with the following simplifying assumptions:

The ow pattern in the transmission grid is such that area L1 exports power to area L2 where the large consumption center is located. Lines connecting areas L1 and L2 are already congested or close to being congested.
Then, installing a new power plant in L1 to serve the load located near to L2 would imply that a new line reinforcing the connection between both areas must be built. If the new CCGT plant is the only new generator using the line and if a split of the cost of transmission lines of 50 per cent to generation and 50 per cent to load is assumed, application of the method explained in Section 4 would result in the new plant at L1 having to pay half the cost of the line (the entire fraction of the cost to be recovered from generation). The line has been assumed to be a typical 400-kV line with a capacity of 100 MW and a cost per kilometer of line of 0.35 million. If the length of the line is assumed to be 550 km, the annual transmission charge to the plant at L1 would be 7.9 million. This gure is of the same order of magnitude as the locational signals that result from the existence of losses in the transmission grid or from the eect of the altitude on the generation capacity of the power plant. Therefore transmission charges in this example could inuence the decision on whether to install the plant in one place or the other. Alternatively, we could have used some average-use method like AP to allocate the cost of both the old and the new lines. In this case, the dierence between total transmission charges in both locations (L1 and L2) would probably not have been very large and locational signals related to the allocation of the cost of electric transmission facilities would have been weaker than those considered in this case example with the new proposed scheme for transmission pricing. Likewise, the cost of infrastructures of the gas grid that are needed by the power plant at L2 have to be accounted for. Here we shall assume that
276
the gas consumed by the CCGT plant at L2 is coming from L1 through a single pipeline whose cost is shared with gas consumers. It has been supposed that the infrastructures that are used to transport gas from L1 to L2 include 90 km of 20 pipeline, 450 km of 26 pipeline and two compression stations, both in the 26 pipeline. Given the amount of gas consumed by the power plant and the typical amount of gas transported annually through the dierent gas pipelines, it has been estimated that a CCGT power plant at L2 would be using between 10 and 15 per cent of these infrastructures on average. This results in an annual charge of 3 million for the use of gas grid infrastructures by the plant at L2. Perunit construction costs have been taken from Spanish Ministry of Industry, Tourism and Trade (2005). Locational signals relating to the existence of congestion in the grid have not been taken into account in this example since congestion is typically infrequent within any system where the grid is suciently meshed. Assuming that the lines between L1 and L2 become congested in the future and dierent energy prices are computed for both areas, locational signals resulting from this congestion would amount to the dierence in energy prices between both areas times the total power production of the power plant over the period of time during which the lines are congested. Considering a dierence in energy prices of 10 per cent of the average system price (supposed to be 36/MWh) and assuming that the line is congested 30 per cent of the time, annual revenues from the sale of energy would be about 2 million higher at L2 than at L1. This gure is comparable to the dierence between the annualized eect of the altitude on the investment cost of the power plant in both locations (which is also about 2 million) but clearly smaller than the dierence between the annual value of the losses assigned to the generator in both cases (which is about 6.4 million). However, numbers depend critically on the assumptions that have been made. From the numbers shown in Table 7.2 one can conclude that the net present value of the cost savings that are obtained annually when installing the power plant close to the consumption center, instead of installing it close to the LNG regasication facility, would amount to 9 million. This gure is the dierence between the total annual costs associated with each one of the two locations (see last row in Table 7.2). Numbers in this row correspond to either the total annual value of the energy produced in the two locations or to the total annual cost of the facilities built in the two cases (9 (76.5 72.7) (42.3 42) (7.9 3)). Note that, when computing these numbers, we have assumed a certain pattern of electricity ows in the grid that remains the same during the entire time horizon of the study. Therefore, transmission charges and
277
losses also remain the same. The breakdown of costs in Table 7.2 shows the relative importance of each one of the locational factors for this case. Obviously, these results may be dierent for other cases, but the approach to be followed in the evaluation should be basically the one that has been shown here.
7.
CONCLUSIONS
Restructuring and liberalization of the electric power industry has resulted in the unbundling of the activities of generation, transmission, distribution and supply. Investment decisions for generation and network infrastructures are no longer in the same hands. The ensuing loss of eciency due to unbundling may be partly compensated by coordinating locational economic signals at the dierent interfaces. Ecient investment in generation can be encouraged by the use of nodal energy prices and transmission taris with locational content. These signals should be designed on the basis of responsibility on the incurred network costs. The usefulness of network-related locational signals for transmission investment critically depends on the adopted regulatory approach for transmission and, in particular, on which one is the entity in charge of the expansion of the grid: the regulatory authority, the transmission system operator, merchant investors or the network users themselves. Other useful locational signals of a non-economic nature, whose purpose would be to decrease the level of uncertainty that both potential generation investors and transmission network planners must face, consist in providing reliable information on future investment decisions (generators) and expected network conditions (system operator). Economic locational signals that are based on cost causality are always useful to guide the siting in the network of new generation and load facilities. The issue here is whether these locational signals may or may not be strong enough to have any actual impact on the location of new network users or even on the decision to retire a power plant or a load from a given site. The chapter has provided criteria and numerical examples showing the nature of these signals and what can and cannot be expected from their application. A novel scheme to design transmission charges with a special emphasis on cost causality has been proposed. Locational signals may also be useful at the interface between transmission and distribution networks in order to minimize the distortions that result from having separate network operators under widely dierent regulatory rules.
278
APPENDIX 7A1
MATHEMATICAL FORMULATION OF THE PROBLEM OF CENTRALLY PLANNING THE EXPANSION OF THE GRID BOTH WITHIN THE TRADITIONAL AND THE COMPETITIVE REGULATORY FRAMEWORKS
Centralized Planning of Network Investments in a Traditional Regulatory Setting The objective of transmission planning in the traditionally regulated environment is to maximize the global social welfare resulting from the expansion of the grid. Given that generation was traditionally planned together with transmission, maximizing the social welfare was equivalent to minimizing the cost incurred when supplying the system load while complying at the same time with certain technical and reliability criteria. In order to do so one has to take into account the investment, maintenance and fuel costs of the power plants, as well as the investment, operation and maintenance costs of the grid. When deciding how to supply the demand, it was possible to consider the joint cost of the transmission and generation involved in a number of dierent possible expansion plans and to choose the most ecient alternative. The central planner had to consider a number of uncertain inputs to the planning process, which basically were the demand forecast, the estimation of hydrological conditions, the costs of fuel and capital, and the expected availability of the already existing and new generation and transmission equipment. From all this information the central planner had to decide where and how much to invest in new transmission and generation facilities. In other words, contrary to what happens now that competition has been introduced, the central planner knew about the generation expansion plans and could modify them. It is possible to write the mathematical formulation of this optimization problem. For the sake of simplicity, a static optimization model is presented, where the optimal mix of generation and transmission is determined for a single given future year, ignoring the trajectory of investments along the planning horizon. The objective function to be minimized is:
Kl l k 1 Mlk Wn N Ei
hlk
m 1
Xlkm rs,i
w 1
hnwYnw
D
s
ps
i 1c 1
cie gs,ie (7A1.1)
CEN
i 1
279
where is the subset of branches (that is, corridors) where new investments are allowed, hlk the investment cost of a line of type k in corridor l, Kl is the number of dierent types of line (220 or 400 kV, single or double circuit, number of conductors per phase, overhead or underground) that are allowed as expansion options in corridor l, Mlk is the maximum number of new lines of type k that can be added to corridor l, Xlkm is a binary variable whose value is 1 if line m of type k is built in corridor l and it is 0 otherwise, is the subset of nodes where new investment in generation can be located, hnw is the investment cost of a power plant of type w at node n, Ynw is a binary variable that is 1 if a plant of type w is built at node n and it is 0 otherwise, D is the duration of the planning horizon, the subset of considered production cost scenarios, ps is the probability of occurrence of scenario s, N represents the total number of areas, Ei is the total number of generation blocks in area i, cie corresponds to the variable production cost of generation block e in area i, gs,ie is the power output of generation block e in area i for scenario s, CEN is the cost of unserved energy and rs,i is the unserved power in area i for scenario s. The objective here is to minimize the total cost of supplying the demand, while taking into account the xed costs of the newly installed generation, the variable power production costs and the transmission costs. Ohmic losses can, for instance, be taken into account by adding a ctitious load at both ends of each line with every load being equal to one half of the ohmic loss of the line. A number of constraints have to be considered: maximum total number of lines that can be installed in each corridor (7A1.2), maximum number of lines of a given type that can be installed in each corridor (intrinsic to the formulation), maximum total investment per corridor (7A1.3), or for the entire network (7A1.4), maximum output for every thermal generation block and for the equivalent hydroelectric unit in each area (7A1.5), maximum unserved energy in every area (7A1.6), electric network equations (7A1.7) and the transmission capacity limit for every corridor (7A1.8).
Kl Mlk k 1m 1 Kl k 1
Xlkm
Ll l
(7A1.2)
Mlk
hlk
m 1
Xlkm
Mlk
Il
(7A1.3)
Kl l k 1
hlk
m 1
Xlkm
TI
(7A1.4)
280
0 0 Sfs fs gs
gs rs rs
s
gs ds ds 0
s
(7A1.5) (7A1.6) (7A1.7a) (7A1.7b) (7A1.8)
DST | fs|
fs
where Il is the maximum investment in corridor l, Ll is the maximum number of lines to be installed in corridor l, TI is the maximum total invest1 T ment in the network, is the vector representing the ohmic losses 2S for each one of the areas, represents the vector of ohmic losses for each one of the corridors, g is the vector of active outputs from every thermal generation block and every hydroelectric equivalent unit in each area, |g| is the vector of maximum active power outputs from every thermal generation block and every hydroelectric equivalent unit in each area. This vector is a function of the investment variables in generation Ynw. Finally, r is the vector of area load curtailments. Given that the construction times and the investment costs of power plants used to be much higher than those for transmission lines, generation planning normally took precedence over transmission planning. Once the expansion of generation had been planned, investments in the transmission grid were decided while considering the investments in generation as given. In this case, the mathematical statement of the problem can be simplied since the investment variables in generation Ynw are now input data. As for the planning of the distribution grid, the objective remains the same as that used for transmission. The planner must minimize the cost incurred by supplying the forecast demand while complying with certain reliability criteria. Reliability can be incorporated into the objective function to be minimized, it can be taken into account through reliability constraints or it can be included both in the objective function and the set of constraints. Traditionally, planners developed the distribution grid independently from the transmission grid. Despite the fact that the planning of transmission and distribution grids deserve a specic treatment, some level of coordination between them is also advisable, as has been explained. Here we leave aside the fact that the volume of distributed generation is growing very fast in many countries. The cost factors that play a role in the optimization process are mainly investment in distribution grids, ohmic losses and the cost of non-supplied energy (for details, see Peco, 2001).
281
Dierences and Similarities in Centralized Transmission Planning Under Traditional and Competitive Regulatory Frameworks When competition in the generation activity is introduced, the entity in charge of optimizing the development of the grid no longer knows what the evolution of generation in the system will be. This fact adds much more complexity to the task of deciding which lines to build, due to the higher degree of uncertainty that the network planner must face. It is generally agreed that, also within a competitive framework, the ultimate objective of the process of planning the expansion of the grid should be to maximize the global social welfare. At the same time, generators are competing among them with the purpose of maximizing their prot, like consumers also do. Therefore, assuming that there is a central network planner, his/her objective will be to maximize the joint net prot made by generators and consumers because of the utilization of the grid. This means that both the benets and the adverse eects of any type that are incurred by the construction of a line must inuence the decision by the planner on whether to construct a line. If it is accepted that generation has precedence over transmission both in the traditional and the liberalized setting then the theoretical equivalence of the outcome of transmission planning under both regulations can be proved, when the former objective function is replaced by the new one. Therefore, it will be shown that it is possible to provide a sound conceptual answer to the problem of the objective function to be pursued by transmission planning in the new competitive environment. This has practical implications in the denition of the regulatory test (see Section 2). The simple proof of this statement follows: within the traditional approach, investment in transmission and generation is jointly optimized. Since generators are paid their cost of service, the objective is just to maximize the consumers welfare, that is, the utility they obtain from the use of electricity minus the cost of producing and delivering electricity. Therefore, the objective function can be simply expressed as: Max{U(D) CFG CVG CT}, (7A1.9)
where U(D) is the utility function of consuming a demand D, CFG represents the generation xed costs, CVG are the generation variable costs and CT are the transmission costs (capital plus operation and maintenance costs, which altogether can be considered as xed costs). Assuming an inelastic demand (in the range of prices that may be modied by transmission investment) and if generation planning is also
282
prescribed from the outset, transmission planning becomes the typical minimization of generation operation costs via network reinforcement: Min{CVG CT}. (7A1.10)
Within the competitive regulatory framework, the entity in charge of transmission planning (the system operator, typically, under regulatory supervision) must apply the following optimization criterion in order to identify the network reinforcements that must be proposed to the regulatory entities for authorization: Max{Net benefit of consumers Net benefit of generators} (7A.11)
where the total cost of any justied investments is implicit in these net benets as network charges to consumers and generators. In general, it is a good sign in the design of the rules for competitive markets that the ideal outcome coincides with the one that the traditional approach would produce under the same circumstances. This is exactly what has been accomplished here, as is shown next. In a competitive wholesale market, the following expression holds: PD IG IVT RNC 0, (7A1.12)
where PD is the total payment (at wholesale level) of consumers, IG is the total income of generators (net of any network payments), IVT is the global variable income of the transmission network (based on the application of nodal prices to both consumers and generators) and RNC is the residual network charge of the transmission network (that is, the part of the total network cost CT that is not recovered by IVT). The preceding expression allows one to replace the objective function of the maximization problem in the traditional approach by this one that is entirely equivalent: {U(D) PD} {IG CVG CFG} {IVT RNC CT}, (7A1.13)
which shows that the maximization problem in the traditional approach can be replaced by the following equivalent problem in the context of the competitive approach: Max{Net benefit of consumers Net benefit of generators} (7A1.14)
283
since the transmission network is regulated so that CT IVT RNC. Note that, embedded in the net benet of consumers and generators, are the complete payments for any justied investment in transmission facilities. One must be aware that, in order for the proof to be valid, the assumed entity in charge of the centralized development of the grid in a competitive environment must pursue those grid expansions that maximize the global surplus resulting from them. The same outcome in general will not be attained under dierent regulatory schemes, such as leaving the initiative to build any new network facility to coalitions of network users willing to pay its cost or letting private promoters invest in new lines.
APPENDIX 7A2
Background
THE METHOD OF AVERAGE PARTICIPATIONS
The basic intuition behind the AP method is that the sources of the supply to loads and the destination of the power injected by generators, as well as the responsibility for causing the ows in all lines, can be assigned by employing very simple heuristic rules that only make use of the actual pattern of network ows. Although this procedure does not intend to capture the details of the physics of the problem, one could argue that, in an electricity market that works reasonably well, the power ows from nodes where it is less expensive towards nodes where it is more expensive. Thus, using the actual network ow pattern may be a way of assigning sources and sinks to loads and generators, respectively, in a reasonable and economically meaningful manner. It is not the only possible way, but it is intuitive, and simple to explain and to compute. The earliest reference that the authors of this chapter know about of the AP method is a handbook on transmission pricing by the New Zealand transmission company, Transpower, dated late 1980s. The AP algorithm is simple and robust, and it uses as its basic input the data of historical or computed network ows, where there are obviously no simplications, and the actual network topology. The method is based on a proportionality assumption: the inows to a node are distributed proportionally among the outows. Causality, or attribution of responsibility, directly results from this assumption which allows one to trace the ows upstream or downstream. Robustness, that is, little volatility with respect to input data, or the absence of arbitrary decisions, such as the choice of a slack bus, which critically inuence the results in other more sophisticated methods, is a very desirable characteristic of this method. This is even more
284
so when the application of the method has signicant economic implications for the network users. The AP method is obviously not free from assumptions such as the rule of proportional allocation of ows at each node that is not fully supported by engineering principles (and it cannot be, as power does not really ow in the networks as a uid in a pipeline) but this rule is of a physically intuitive nature, very much associated with the robustness qualities of the method, and it basically guarantees the rationality of its results, which cannot easily be challenged. The algorithm of average participations has been used, with minor variations, in real systems such as New Zealand (in the late 1980s), Poland or South Africa, as well as in specic transmission pricing studies in at least Chile, Central America, Romania, Spain and the IEM of the European Union. It can be programed easily and it has been studied thoroughly in the technical literature. It is not free from objections, but it appears to be a reasonable procedure whose level of sophistication is well adapted to the problem under consideration. Description The method requires as its basic input data a complete snapshot of the network power ows corresponding to the specic system conditions of interest. The algorithm is based on the assumption that electricity ows can be traced or the responsibilities for causing them can be assigned by supposing that, at any network node, the inows are distributed proportionally between the outows. Implicit in this rule is the additional assumption that the physical ow of electricity in a line is not the combination of many ows in opposite directions, since only ows in the same direction as the nal physical ow are assumed to exist. Under these assumptions, the method traces the ow of electricity from individual sources to individual sinks; that is, the model identies, for each generator injecting power into the network, physical paths starting at the generator that extend into the grid until they reach certain loads where they end. Symmetrically, the paths from loads to generators can also be found, yielding exactly the same results of allocation of responsibility of ows to generators and loads. Then, the cost of each line is allocated to the dierent users according to how much the ows starting at a certain agent have circulated along the corresponding line. This is how the method works: for every individual generator i, a number of physical paths are constructed, starting at the node where the producer injects the power into the grid, following through the lines as the power spills over the network, and nally reaching several of the loads in the system. An analogous calculation is also performed for the demands,
285
20 MW G1
L3
L1 D1
15 MW
L4 40 MW G2 Figure 7A2.1 D3 = 10 MW
L2 D2 35 MW
Proportionality principle in average participations
tracing upstream the energy consumed by a certain user, from the demand node until some generators are reached. One such physical path (with as many branches as needed) is constructed for every producer, and for every demand. In order to create these paths, a basic criterion is adopted: in each node of the network the inows are allocated proportionally to the outows. This makes the method easier to implement. A simple example is shown in Figure 7A2.1. According to the proportional distribution rule, generator G1 would contribute 15 20/(20 40) MW to the ow over line 1, 35 20/(20 40) MW to the ow over line 2, as well as the total of 20 MW owing over line 3 and nothing to the ow over line 4. Similarly, demand D2 would contribute 20 35/(10 15 35) to the ow over line 3, 40 35/(10 15 35) to the ow over line 4, all the 35 MW owing over line 2 and nothing to the ow over line 1. The demand D3 of 10 MW has a contribution of 20 10/(10 15 35) to the ow over line 3, 40 10/(10 15 35) to the ow over line 4 and no contribution to the ows over lines 1 and 2. The participation of agent i in the use made of line j is obtained as the fraction of the ows starting at agent i that passes through line j. The method implicitly results in a 50/50 global allocation of costs to generators and demands. However, if desired, an ad hoc weighting factor could be used to modify this percentage.
NOTES
* The material in this chapter has beneted from countless discussions with many colleagues in dierent parts of the world and for many years. We should mention at least the members of the Infrastructure Working Group of the Council of European Energy Regulators (CEER), the many participants in the discussions leading to the
286
transmission regulation for the Central American Electricity Market (SIEPAC project), the professionals from Red Elctrica de Espaa (the Spanish grid company) with whom we have had discussions for several years about transmission investment issues and who have provided the data for the second case example on locational signals, the participants in the meetings of the CambridgeMIT Consortium, the members of the Working Group on Transmission Taris of the French Electricity Regulatory Commission (CRE) that for two years met regularly to discuss transmission issues, the sta of the Spanish national energy regulator from whom we learned about the tarisetting process in the gas sector and our colleagues Javier Rubio, Carlos Vzquez, Michel Rivier, Toms Gmez, Juan Rivier and Jess Peco from the Instituto de Investigacin Tecnolgica (IIT). 1. REE stands for Red Elctrica de Espaa, the Spanish Electricity Grid. 2. The alternative regulatory frameworks that are presented here are representative cases of the existing possibilities, since an exhaustive enumeration is out of question. For instance, a combination of the rst and second cases is possible: the TSO may have to present a global transmission expansion plan to the regulator to be approved (under a technical viewpoint) as in the rst case, but the TSO will be responsible for the actual implementation of the plan and be subject to RPI-X remuneration, as in the second case. 3. This is a well-known fact that has proved to be true in numerous actual networks and that also has a theoretical support (see Prez-Arriaga and Rubio, 1995). 4. This is how The Victorian Supreme Court has expressed the regulatory test: A new interconnector or transmission system augmentation satises this test if it maximizes the net present value of the market benet, having regard to a number of alternative projects, timings and market development scenarios. Market benet here means the total net benets to all those who produce, distribute and consume electricity in the electricity market. See Littlechild (2004, p. 20). 5. For instance, benets are frequently replaced by some measure of electric usage and very often the traditional version of the rule is used even in the context of competitive wholesale markets. 6. Since nodal prices can generally only recover a small fraction of the total transmission costs, the problem of determination of transmission taris that pay for transmission costs will be considered from now on to be tantamount to the problem of determination of the long-term signals, regardless of whether nodal prices of energy are actually applied in a system or not. 7. In most European countries it is probably true that most domestic consumers are among the least elastic ones, while large industrial consumers are typically very elastic to electricity prices. However, this statement may be wrong in many developing countries, where industrial consumers generally subsidize domestic consumers. 8. Remember that signals that are derived from losses and congestions are short-term ones; they cannot generate the complete revenues for the required investment since: (a) in general they will be too weak for that, due to the typical overinvestment in transmission; and (b) these signals will typically be much reduced even almost disappear once the reinforcement is built. 9. Assume a line with a nominal transmission capacity (in MW) that is dened as C1 according to some agreed procedure. However, because of system security reasons, the maximum real power that the line can transmit under normal operating conditions is C2. Then, if the actual real power owing through the line is P, the fraction of the line that is actually used is simply P/C2. 10. For instance, we could use coecients that are much larger than 1 (like 5, for instance) at the time the corresponding line (in the case of kl) or generator (in the case of kGi) enters into operation. Then, these coecients would gradually decrease with the passing of time until they become zero after ve years for generators and 10 years for lines, for instance. 11. Locational signals are caused by the dierence among the transmission charges paid by generators placed at dierent locations.
287
REFERENCES
Australian Competition and Consumer Commission (2003), Review of the regulatory test, www.accc.gov.au/content/index.phtml/itemId/142. Brown, R.E. (2002), Electric Power Distribution Reliability, New York: Marcel Dekker. Bushnell, J. and S. Stoft (1996), Grid investment: can a market do the job?, The Electricity Journal, 9 (1), 749. Carrillo-Caicedo, G. and I. Prez-Arriaga (1995), Optimal reconguration of distribution networks for a diversity of regulatory frameworks, IEEE/KTH Stockholm Power Tech Conference, Stockholm, Sweden, August 1921. Council of European Energy Regulators (2003), CEER Guidelines on Regulatory Control and Financial Reward for Infrastructure, www.ceer-eu.org/. Florence School of Regulation (2005), Study on the inter-TSO compensation mechanism pursuant to article 3 of Regulation (EC), n.1228/2003, www.iue.it/ RSCAS/ProfessionalDevelopment/FSR/. Gilbert, R., K. Neuho and D. Newbery (2002), Allocating transmission capacity to mitigate market power in electricity networks, Massachusetts Institute of Technology, Cambridge, MA, www.econ.cam.ac.uk/electricity/. Joskow, P.L. (2005), Transmission policy in the United States, Utilities Policy, 13, 95115. Joskow, P. and J. Tirole (2002), Transmission investment: alternative institutional frameworks, in Wholesale Markets for Electricity, November 223, Toulouse, France. Khator, S.K. and L.C. Leung (1997), Power distribution planning: a review of models and issues, IEEE Transactions on Power Systems, 12 (3), 11518. Littlechild, S. (2004), Regulated and Merchant Interconnectors in Australia: SNI and Murraylink revisited, CMI Working Paper 37, Cambridge, UK, January. Transactions on Power Systems, 9 (4), 188694. Littlechild, S.C. and C.J. Sherk (2004), Regulation of transmission expansion in Argentina. Part I: State ownership, reform and the Fourth Line, Department of Applied Economics, CMI electricity project, University of Cambridge, Working Paper ep 61, www.econ.cam.ac.uk/electricity/publications/wp/, p. 75. Mercados Energticos for Osinerg (Peruvian Energy Regulatory Agency) (2005), Plan Estratgico para Modernizacin del Marco Regulatorio, www.osinerg. gob.pe/. Outhred, H.R. and R.J. Kaye (1996), Incorporating network eects in a competitive electricity industry: an Australian perspective, in M. Einhorn and R. Siddiqui (eds), Issues in Transmission Pricing and Technology, Boston, MA: Kluwer Academic, pp. 20728. Peco, J.P. (2001), Modelo de Cobertura Geogrca de una Red de Distribucin Elctrica, Universidad Ponticia Comillas, Doctoral Thesis, Instituto de Investigacin Tecnolgica, Madrid. Prez-Arriaga, J.I. (2002), Methodology for cross-border tarication in the internal electricity market of the European Union, Power Systems Computing Conference (PSCC), Seville, June 248. Prez-Arriaga, J.I. and L. Olmos (2003), Extension of the project on cost components of cross border exchanges of electricity, European Commission, Directorate-General for Energy and Transport, http://europa.eu.int/comm/ energy/electricity/publications/index_en.htm.
288
Prez-Arriaga, J.I. and F.J. Rubio (1995), Marginal pricing of transmission services: an analysis of cost recovery, IEEE Transactions on Power Systems, 10 (1), 54653. Prez-Arriaga, J.I. and Y. Smeers (2003), Guidelines on tari setting, in F. Lvque (ed.), Transport Pricing of Electricity Networks, Boston, MA: Kluwer Academic, pp. 175204. Prez-Arriaga, J.I. et al. (1987), Situacin del estado del arte en la planicacin de redes de transporte de energa elctrica, Report prepared for Red Elctrica de Espaa (in Spanish). Prez-Arriaga, J.I., L. Olmos and F.J.R. Odriz (2002), Cost components of cross border exchanges of electricity, European Commission, Directorate-General for Energy and Transport. http://europa.eu.int/comm/energy/electricity/ publications/index_en.htm. Rubio, F.J. (1999), Metodologa de Asignacin de Costes de la Red de Transporte en un Contexto de Regulacin Abierta a la Competencia, Universidad Ponticia Comillas, Doctoral Thesis, Escuela Tcnica Superior de Ingeniera, Madrid. Schweppe, F.C., M.C. Caramanis, R.D. Tabors and R.E. Bohn (1988), Spot Pricing of Electricity, Boston, MA: Kluwer Academic. Spanish Ministry of Industry, Tourism and Trade (2005), Orden ITC/102/2005 por la que se establece la retribucin de las actividades reguladas del sector gasista, www6.mityc.es/energia/archivos/Orden%20Retribuciones%202005.pdf.
Index
AC interconnectors 182 AC power lines 109, 118 access charges 205, 221 access rights 221 access rules 153, 233, 235, 241, 2578 active TSOs 234, 239 incentives for 239 airline industry 65 Alberta mixing planned and merchant transmission in 119 MW-mile charges in 107 transmission network organization in 148 zero-congestion policy in 91, 102, 119 algorithms 246 alternating current (AC) power lines 109, 118 altitude over sea level 2733, 274 aluminiumzirconium wires 115 amortization charges 162 ancillary services, cost of provision of 146, 272, 273, 274, 275 Anderson, D. 67 arbitrage 16, 155, 188 areas of inuence method 250 Areva 72, 79 Argentina transmission investment in 235 transmission network organization in 148 transmission pricing in 249, 250 association of European Transmission System Operators (ETSO) 187, 264 asymmetry of information 1214, 1512, 192, 193, 224 attributes of transmission investments 13647 auctions 88, 109, 111, 118, 163, 168, 191, 221, 233, 234, 235, 236 Australia energy only market in 3940 HVDC link with 141, 155 transmission investment in 155 transmission pricing in 249 Australian Competition and Consumer Commission 237 Austria net inter-TSO payment of 267, 271 transmission pricing in 265, 266, 271 average cost of transmission 2534 average participation (AP) method 11, 2634, 267, 268, 275, 2835 Averch, H. 667 Bailey, E. 66 balancing market 158, 160, 163, 166 Bar-Ilan, A. 334 Barthold, L.O. 115 base-load plants 34, 2230, 43, 49 Beesley, M. 66 Belgium cost of capital in 76 generation investment in 70, 72 net inter-TSO payment of 271 transmission pricing in 265, 266, 271 Ben-Tal, A. 209 benchmarking 135, 162, 164, 180 benet/cost ratios 172, 1745, 238, 245 BETTA (British Trading and Transmission Agreements) 182 Bjorndal, M. 193, 223 blackouts see power cuts Borenstein, S. 108, 126 Bower, J. 67, 71 Brennan, M.J. 35 British Electricity Association 59, 66 British Energy 81 Brown, R.E. 230, 231 Brunekreeft, G. 107, 118, 124, 127 budget-balance constraint 151, 152, 159 289
290 bundled transmission services, prices for 1656 Bushnell, J.B. 110, 187, 224, 231
Index competitive bidding see auctions complementary goods 2 computer systems 15, 124 congestion 4, 8, 10, 16 economically optimal level of 7, 14 EU regulation on management of 189, 190191, 196, 201 growth of 1312 pushed across borders 15, 140 revenue adequacy in congestion management 222 social cost of 172 three distinct costs associated with 14, 8890 see also congestion costs; congestion rents; cost of congestion to load unhedgeable 172, 173, 1745, 177, 178 zero-congestion policy 14, 913, 95, 106, 119 strategic manipulation of 1023 congestion charges bilateral schedules liable for 166 eciency properties of 199202 FTRs as hedge against 191, 192, 221 incentive eect of 193, 221 in New York ISO 132 non-discriminatory 217 in PJM 1312, 1734 and revenue adequacy 222 congestion costs compared with congestion rent and cost of congestion to load 14, 8890 denition of 89 economic models focusing on 139 LMPs reecting 166 measurement of 120121, 172 trade-o between security and reduction in 235 trade-o between transmission investments and reduction in 113, 139, 1756, 187 ways to reduce 125 congestion rents compared with congestion costs and cost of congestion to load 14, 8890 denition of 89
California congestion in 132 generation investment in 57, 58, 62, 65 independent system operators 54 power shortages in 1, 31, 37, 45, 126 capacity charges 204, 205, 214, 217, 222, 269 capacity expansion model 223 capacity factor 75, 79 capacity margins 4, 416, 47, 199 capacity obligations 6, 1112, 166 capacity payments 6, 357, 40, 158 capacity reserves 45 capital costs 1314, 55, 68, 69, 70, 7280, 81, 120, 1612, 273 capital intensity 1314, 55, 667, 74, 81 carbon price trading 71, 80 categorization of transmission investments 13743 CCGT see combined-cycle gas turbine (CCGT) plants CEER (Council of European Energy Regulators) 235, 237, 285 Central Electricity Generating Board (CEGB) 41, 157, 164 Chao, H.-P. 138 Chile, transmission pricing in 249, 250 circulated uidised bed combustion technology 69 coal-red plants 219 technological switch to/from 58, 60, 61, 62, 63, 64, 65, 66 factors aecting 67, 6872, 81 coalitions of network users, investment by 233, 235, 239, 283 incentives for 240241, 242 combined-cycle gas turbine (CCGT) plants 4, 412, 46, 54, 244, 256 factors aecting location of 270277 technology change to/from 1314, 545, 69, 70, 712, 73, 745, 79, 80, 81 combined heat and power producers 62, 64 Comitology 189
Index economic models and 139 and merchant investment 9, 110118, 156, 171, 177, 236, 237, 241, 242 optimal transmission investment and 956, 99102, 1067, 11216 PJM projects nanced from 144 transmission rights yielding 9, 89, 110113, 115, 191, 236, 247 congestion revenue rights (CRRs) 10910, 111, 112, 115 constant returns to scale 11314, 187 construction costs electricity transmission lines 275 gas grid infrastructures 276 location and 171 plant type and 3, 67, 71, 734, 76, 79, 81 consumer surplus 116, 117 control areas 140 costbenet analysis 172, 173 cost causality, locational signals based on 16, 188, 18990, 200, 201, 202, 20417, 222, 223, 224, 225, 245, 250, 253, 254, 259, 277 cost minimization 120121, 122, 197, 206, 230, 278, 279, 280, 282 cost of capital 1314, 55, 68, 69, 70, 7280, 81, 120, 1612, 273 cost of congestion to load 14, 88, 90 cost of service regulation 119, 135, 156, 161, 166, 179, 233, 240, 260 cost of unserved load 120, 121, 122, 125 cost reectiveness see cost causality, locational signals based on Council of European Energy Regulators (CEER) 235, 237, 285 Credit Suisse First Boston (CSFB) 70, 71 credits 234, 235 Crew, M.A. 124, 187 Curien, N. 190 CUSC (Connection and Use of System Code) 159 Czech Republic, transmission pricing in 265, 266 day-ahead markets 166, 188 DC power lines 11415, 118, 171, 236
291
see also high-voltage direct current (HVDC) transmission links de Luze, G. 77 deep interconnection policy 134, 142, 16870 delivery to load criteria 1678 demand charges see capacity charges demand schedules 158 demand uncertainty 3031 Denmark net inter-TSO payment of 271 transmission pricing in 271 Department of Energy see DOE (Department of Energy) depreciation rates 161 DGEMP (Direction Gnrale de lEnergie et des Matires Premires) 68, 71, 72 digital technology 66 direct current (DC) power lines 11415, 118, 171, 236 see also high-voltage direct current (HVDC) transmission links discount rate 97, 116 discrete decisions 190, 193, 224 distributed generation 230, 238, 258, 259, 260, 261, 262, 280 distribution networks coordination between transmission investment and 17, 138, 231, 25862, 277, 280 denition of 258 regulation of 234 Dixit, A.K. 313 DOE (Department of Energy) Annual Energy Outlook 71 see also EIA-DOE (Energy Information Agency and Department of Energy) dual red plants 5966 dual price functions 193 economic eciency principle 16, 18990, 199202, 217, 220, 222, 224, 245, 248 economic models of transmission investment 15, 1323, 139, 154, 181 economic transmission investments 15, 133, 139, 1423
292
Index net inter-TSO payment of 271 regulatory framework for transmission in 15962, 180 transmission investment in 132, 142, 1589, 1624, 175, 177 transmission network facilities in 136 transmission network organization in 1478 transmission pricing in 12, 37, 142, 158, 15960, 1612, 168, 169, 182, 249, 271 wholesale market arrangements in 12, 15762, 179, 220 Enron 77, 81 Erie West HVDC 16970 ETSO (association of European Transmission System Operators) 187, 264 EU see European Union (EU) European Commission 189, 228 European Parliament and Council 187 European Regulation 1228/2003 on Conditions for Access to the Network for Cross-border Exchanges in Electricity 18792, 196, 200, 201, 215, 217, 220, 221, 222, 224 European Union (EU) Directives 54, 164 inter-TSO payments in 2634, 265, 267, 268, 269, 271 internal electricity market of 54, 189, 237, 246 nodal transmission taris in 26370, 271 Regulations see European Regulation 1228/2003 on Conditions for Access to the Network for Cross-border Exchanges of Electricity regulatory responsibilities in 135 transmission investment policy in 146 exit fees 175 expected amount of unserved energy 36 expected net present cost 978, 104 expected value of capacity 36 externalities 7, 87, 102, 108, 187
in PJM 1434, 1713, 1745, 1767 versus reliability-driven transmission investment 133, 1757, 180 economies of scale 8, 9, 10, 14, 87, 187 compared with lumpiness 1012, 11316 and market power 108 transmission pricing with 151, 204, 250 ecient size of network 89 ecient use of network 8 eort, reward for 120, 121, 1234 EIA-DOE (Energy Information Agency and Department of Energy) 58, 65, 68, 69, 70 see also DOE (Department of Energy) electric power network models 1323 Electric Transmission Week 181 Electricit de France 24, 75, 767, 79, 81, 148 electricity utilities, investment by 5664, 67 emissions 41, 43 emissions trading 71, 80 energy balancing costs 160161, 162 energy charges 269 energy only markets 12, 3940 engineering reliability criteria in England and Wales 158, 160, 1623 factors aecting 154 ignored by economic models 15, 1323, 181 ISO responsible for applying 148, 165 minimizing cost while complying with 278, 280 PJMs 165, 1679 reliability investment to restore 142 transmission investments driven by 134, 1445, 1623, 176, 177, 180 England and Wales capacity payments system in 6, 357, 40 cost of capital in 76 generation investment in 4, 358, 412, 47, 54, 57, 59, 65, 66, 67, 68, 69, 70, 72, 78, 80
Index feasible sets of transmission rights 10911 Federal Energy Regulatory Commission (FERC) 135, 141, 164, 1656, 170, 172, 173, 191 Felder, F. 118 FERC (Federal Energy Regulatory Commission) 135, 141, 164, 1656, 170, 172, 173, 191 nancial guarantees 256 nancial transmission rights (FTRs) 3, 10, 110112, 167, 168, 169, 171, 172, 1912, 221 nes see penalties Finland cost of capital in 76, 79 generation investment in 423, 71, 72, 79, 81 net inter-TSO payment of 271 transmission pricing in 271 Finon, D. 77, 80 Firecone Ventures Pty Ltd 118 xed costs of generation 3, 22, 23, 40, 195, 197 recovery of 4, 6, 24, 26, 39 see also construction costs xed costs of transmission 14, 94, 95, 119 recovery of 9, 10, 11, 14, 15, 16, 923, 100102, 107, 108, 11718, 144 see also construction costs Florence Regulatory Forum 18990, 201, 217, 246 Florence School of Regulation 264 Ford, A. 37 forward contracts 160, 192 France cost of capital in 75, 767 HVDC link with 141, 157 transmission network facilities in 136 transmission network organization in 1478 transmission pricing in 265, 266, 271 free riders 108, 11618, 127 fuel costs congestion inuenced by 139 technology choice inuenced by 30, 68, 69, 70, 71, 73, 78, 79, 80, 81
293
correlation between electricity price and 14, 75 fuel mix see generating technologies, choice of functional separation 1356, 1478 future grid development, information on 257 game theory 103 Gans, J. 120 gas-based electricity generation 412, 43, 126, 219, 244, 256 factors aecting location of 270277 peaking plants 34, 22, 46 technological switch to/from 1314, 546, 5765, 66, 80 factors aecting 6572, 73, 745, 7780, 81 gas bubble 78, 81 gas grid infrastructures 16, 274, 2756 gas losses in pipeline 2735 gas only plants 58, 59, 60, 61, 62, 63, 64, 65, 66 gearing rate 14, 38, 75, 76 generating technologies, choice of 1314, 30, 5481 generation, investment in 37, 1214, 2152, 5481 coordination between transmission investment and 1012, 1617, 1027, 125, 187228, 230231, 24358, 277, 2783 case examples 26377 generation capacity investment cycles and 359 investment in liberalized markets 416 irreversibility and uncertainty and 3035, 47 modelling optimal level of 2130, 4752 paying for capacity in practice 3941 generation investment strategies 1027 generator deliverability investments 166, 1689 generator interconnection investments 1378, 167 geographic scope of TSOs 179 Germany generation investment in 70
294 net inter-TSO payment of 271 transmission pricing in 265, 266, 271 Gilbert, R. 126, 221, 247 Glachant, J.-M. 77, 80 gold-plating 123 Gollier, C. 79 government subsidies 13, 16, 54, 67, 73, 81, 82, 150, 151 graft 1234 Greece net inter-TSO payment of 271 transmission pricing in 271 Green, R.J. 39, 67, 190 Gribik, P.R. 110 grid codes 162, 234, 238, 240, 242 grid equilibrium model 1946 GRTN (Gestore Rete Transmissione Nazionale) 59, 66
Index Finances & Conseil Gnral des Mines) 75, 76, 79 incentive alignment 1512 incentive regulation see performancebased regulation (PBR) incomplete contracts 152 independent power producers (IPPs) 73 investment by 5664, 778, 81 independent system operators (ISOs) congestion revenue rights (CRRs) issued by 109, 110, 111 creation of 54 FTR option rights issued by 111 model of 148, 221 planning by 102 rationale for 1489 regulatory challenges associated with 14950 see also PJM (PennsylvaniaNew JerseyMaryland) independent transco model 1478, 179, 1923, 220221 indivisibility see lumpiness information rent 1214 installed capacity markets 4041 integrated gasication combined cycle technology 69 inter-TSO investments 15, 132, 133, 134, 13942, 147, 150, 157, 1634, 176, 179 in PJM 1735 inter-TSO payments in the EU 2634, 265, 267, 268, 269, 271 interconnection charges 158, 159 interconnection rules 134, 167, 170 internal electricity market 54, 189, 237, 246 nodal transmission taris in 26370, 271 International Energy Agency (IEA) 57, 62, 64, 68, 71, 73 intra-TSO investments 15, 132, 133, 134, 139, 147, 150, 176, 179, 180 in PJM 1713 inverse price elasticity rule 248 investment at risk 236, 239, 240 investment cost-related pricing (ICRP) 250 investment cycles 359
harmonizing transmission pricing practices 15, 133, 232, 263 Harvey, S.M. 88, 191 Hawdon, D. 35 hedging 3940, 52, 78, 79, 81, 11112, 1912, 221, 236 unhedgeable congestion 172, 173, 1745, 177, 178 Henney, A. 157 Hirst, E. 132 Hogan, W.W. 88, 110, 118, 188, 191, 1934, 202, 215, 221 horizontal integration of TSOs 1334, 150 Hungary, transmission pricing in 265, 266 Hunt, S. 56, 67 HVDC (high-voltage direct current) transmission links 134, 141, 1556, 157, 16971, 173, 182 hydro-electric generators 31, 43, 44, 58, 60, 61, 62, 65, 66, 69 Hydro-Quebec 182 Iberian System 243 ICRP (investment cost-related pricing) 250 IEA (International Energy Agency) 57, 62, 64, 68, 71, 73 IGF-CGM (Inspection Gnrale des
Index investment lags 334 investment strategies 23, 98, 1027 Ireland net inter-TSO payment of 271 transmission pricing in 271 irreversible investment under uncertainty 13, 315, 47 Ishii, J. 67 ISO New England 143, 176 ISOs see independent system operators (ISOs) Italy generation investment in 4, 57, 59, 65, 66, 70, 80 net inter-TSO payment of 271 transmission network organization in 1478 transmission pricing in 265, 266, 271 Johnson, L. 667 Jrsten, K. 193, 223 Joskow, P.L. 41, 66, 67, 72, 77, 78, 102, 112, 113, 125, 127, 132, 134, 139, 1456, 152, 154, 156, 164, 165, 166, 169, 170, 175, 177, 183, 191, 220, 221, 230, 231, 254 Kaye, R.J. 262 Khator, S.K. 231 King, S. 120 Laont, J.-J. 151, 163 Larsen, E. 35 Latorre, G. 209, 223 Lautier, T.-O. 120 legacy infrastructure considerations 1456 Leung, L.C. 231 levelized cost methodology 55, 68, 71, 72, 73, 79 Lvque, F. 190 lignite 67 line type 7 linear lumpy technology 11316 linear mixed-integer programs 2089 Littlechild, S.C. 66, 67, 118, 235, 286 load-duration curve 4, 225, 48, 49, 512 load reduction programs 138 lobbying 108, 118, 127
295
locational signals see locational transmission signals, long-term; nodal energy pricing locational transmission signals, long term information-related 2567, 277 transmission charges 187228, 240250 access rules and 2578 case examples 26277 incremental charges for new network users 250255 for retiring generators 2556 LOLP (loss of load probability) 5, 6, 357, 40 Long Island, HVDC links with 155, 171, 173 Long Island Power Authority (LIPA) 155, 171, 173 long-term contracts 13, 14, 389, 78, 79, 80, 134, 1556, 157, 171, 173 two types of 221 long-term locational signals see locational transmission signals, long term loss of load probability (LOLP) 5, 6, 357, 40 lump-sum charges 269 lumpiness 78, 9, 10, 14, 87, 95, 97, 154, 187, 188 compared with economies of scale 1012, 11316 merchant transmission investment with 119, 1567 transmission pricing with 1912, 201, 204, 221, 250 Luxembourg net inter-TSO payment of 271 transmission pricing in 271 MAAC (Mid-Atlantic Area Council) 165 MacKerron, G. 75 marginal cost of generation 195 equilibrium energy price compared with 4, 13, 226, 37, 39, 40, 956, 165, 166 geographically dierentiated 8, 95, 244
296
Index mixing planned and merchant transmission 9, 11819, 180181, 237, 242 in PJM 155, 16971, 173 policy initiatives based on 1267 regulatory framework accommodating 150, 1557, 191, 221, 237 Mid-Atlantic Area Council (MAAC) 165 Midwest ISO (MISO) 165 Mishan, E. 67 MISO (Midwest ISO) 165 MIT (Massachusetts Institute of Technology) 55, 71, 72, 736 monitoring and control equipment 15, 144 mothballing 1213, 31, 33, 35, 378, 45 National Grid Company (NGC) 12, 42, 15764, 234 Nemirovski, A. 209 net investment denition of 21 in selected countries 42, 43, 44, 45 net local demand curve 89 net power output 272 net present value 98, 276 net social benet 1056 NETA (New Electricity Trading Arrangements) 12, 158, 162 Netherlands net inter-TSO payment of 271 transmission pricing in 265, 266, 271 network operating practices 133, 140, 144 network upgrade costs 142, 16870, 172 New Electricity Trading Arrangements (NETA) 12, 158, 162 New England, HVDC links with 141, 171, 182 New England Power Pool 146 New England regional expansion plan 143, 176 New York City HVDC links with 155, 156, 171, 173 market power in 126 New York ISO 112, 132
known to regulator 197 wind power 103 marginal cost of transmission compared with congestion cost 139, 175, 187 compared with congestion rent 100, 1012, 106, 113 compared with nodal price dierential 8, 956 transmission prices reecting 153, 2534 marginal locational pricing see nodal energy pricing marginal opportunity cost of consumption 4, 25 market design 140, 181 market power absence of 188, 193, 224, 245 diversication as insurance against 412 and xed cost recovery 14, 108, 11718 of local suppliers due to transmission limitations 119 price distortions caused by 5, 39, 42, 88 transmission rights and 110, 112, 247 underinvestment as a method of abusing 113, 236 market windows 172, 173, 1745, 1767 Massachusetts Institute of Technology (MIT) 55, 71, 72, 736 McDaniel, T. 124 mean load factor 75 mean reversion 345 measuring transmission capacity 145 Mercados Energticos for Osinerg (Peruvian Energy Regulatory Agency) 254 merchant plants 38, 73, 74, 77, 81 merchant transmission investment 14, 88, 102, 10818, 236, 239 congestion rents and 9, 110118, 156, 171, 177, 236, 237, 241, 242 nancing costs for 156 long-term contracts undertaken by 134
Index New Zealand average participation (AP) method used in 283, 284 transmission network organization in 1478 Newbery, D.M. 38, 67, 78, 79, 132 NGC (National Grid Company) 12, 42, 15764, 234 Nimby constraints 143, 171 nodal energy prices congestion managed by 197, 1989, 201, 211, 212, 21415, 216, 217, 222 denition of 8 nancial transmission rights and 10, 112, 191, 221 locational signals in 11, 240245, 261, 262, 263, 277 and optimal investment mix 10, 88, 90 in PJM 8 nominal capacity 2723 non-discriminatory prices 16, 18890, 200, 201, 202, 204, 21720, 223, 224, 248 non-linear mixed-integer programs 2089 Nord Pool 43, 45 NordNed Cable 182 Norway generation investment in 434, 54, 57, 59 net inter-TSO payment of 271 transmission network organization in 148 transmission pricing in 249, 271 nuclear power plants base-load plants 34, 22, 43 technological switch to/from 1314, 54, 55, 58, 60, 61, 62, 63, 64, 65, 66 factors aecting 67, 69, 712, 7380, 81 nuclear Pressurized Water Reactor (PWR) 712 NVE (Norwegian Energy Regulatory Authority) 59 ONeill, R.P. 10, 188, 194, 223 obligation capacity 6, 1112, 166 Ofgem (Oce of Gas and Electricity
297
Markets) 12, 162, 164, 182, 159, 161, 162, 164 oil-red plants 60, 61, 62, 63, 64, 65, 66 oil tankers 35 Olmos, L. 268 OMEL 268 open access 233, 235, 241 open-cycle gas turbines 22, 69 operating and maintenance costs 68, 69, 75, 79, 81, 146, 1612 opportunity cost of investment 237 optimal transmission cost recovery for 99102 dynamic 967 option value 978 static 936, 1023 strategic manipulation of optimal transmission planning 1057 option rights 110111 option value 6, 13, 31, 32, 33, 34, 79, 80, 978, 102 Outhred, H.R. 262 participation constraint, TSO 150151 participation factor 2523 passive TSOs 233, 235, 239 incentives for 240 path ratings 125 payback period 172, 173, 1745 peaking plants 34, 2230, 40, 46, 49, 512 Peco, J.P. 260, 280 penalties 6, 40, 125, 166, 234, 235, 240 Prez-Arriaga, J.I. 190, 193, 205, 230, 245, 246, 248, 250, 254, 262, 263, 268, 286 perfect competition 33, 90, 112, 113, 188, 197, 200, 204 perfect information 188, 193, 197, 2234 performance attributes of transmission networks 1467 performance-based regulation (PBR) distribution networks 259, 260 transmission investment 88, 102, 11925, 135, 180, 240 diculties for transcos 1415, 1245 direct approach to 120121 NGC revenues 1612
298
Index plant decommissioning see retiring generators plant size 79, 81 Pollitt, M. 67 pool system 6, 12, 357, 40, 43, 45, 146, 158, 165 Pope, S.L. 191 Portugal net inter-TSO payment of 271 transmission pricing in 265, 266, 271 postage stamp tari 142, 249 power cuts 1415, 36, 41, 115, 120, 121, 124, 125, 126 Power Exchange 37 power ow studies 163 power transfer distribution factors (PTDFs) 195, 197, 204, 206 practical transmission planning policy 1057 pre-construction formalities 33 price capping 56, 7, 12, 14, 3941, 125, 154 reducing power of incentive mechanism 1225 price discrimination 16, 18890, 200, 201, 202, 204, 21720, 223, 224, 248 price-duration curve 4, 23, 25, 49, 51 price elasticity of demand 5, 10, 11, 119 price-taking agents 193, 197 primal linear program (PLIP) 198 primal mixed-integer program (PIP) 194 privatization 65 production set 223 productivity improvements 162 prot maximization 2, 47, 106, 192, 199, 214, 242, 281 prot sharing 1224, 135, 152, 180 property rights 154 proportionality principle in average participation 2834, 285 public consultation 161 public goods 41, 154, 177 public interest 1, 5, 18, 134, 135 pulverised fuel technology 69 Quebec, HVDC links with 141, 182
policy initiatives based on 1267 two approaches to reducing information rent 1215 performance norms for TSOs 135, 180 permits for new transmission links 149 physical attributes of transmission network components 1435 Pindyck, R.S. 313, 35 PJM Interconnection 1278, 167, 170, 175, 178, 181 PJM (PennsylvaniaNew JerseyMaryland) capacity markets operated by 6, 11, 40, 166 congestion charges in 1312, 1734 creation of 54 expansion of 140, 165, 173 nancial transmission rights (FTRs) used by 110112, 167, 168, 169, 171, 172 HVDC projects involving 155, 16970, 171 industrial organization and wholesale market design in 1112, 1656 nodal energy pricing by 8 Operating Agreement 167, 168 Regional Transmission Expansion Plan (RTEP) 167, 168, 169, 171, 172, 173 Reliability Assurance Agreement 167, 168 reliability criteria in 165, 1679 reports 58 transmission investment in 132, 1434, 16779 transmission pricing in 142, 1646, 167, 16870, 173, 254 PJM West (Allegheny Power Systems) 140 planning approach, traditional distribution investment 230, 280 generation investment 230, 27880 transmission investment 14, 15, 878, 1027, 11819, 230, 27880 compared with competitive regulatory framework 234, 2378, 2813
Index R&D costs 81 RAENG (Royal Academy of Engineering) 68, 69, 72, 73 Ramsey rule 11, 217, 248 ratchet mechanisms 122, 123, 135, 152, 1612, 180 rate of return (ROR) on investment guaranteed 67 high 114 investment risk premiums aecting 93 normal 123 regulation 88, 102, 122, 125, 126, 127, 166 rationing 25, 36, 41, 138, 154 RAV (regulatory assets value) 161 real option 31 real-time markets 138, 166, 188 reconductoring existing lines 15, 144 Rede Electrica de Espana (REE) 59, 66, 233 redispatching costs see congestion costs reference periods (seasons) for network design 222 regional transmission investment planning process 135 regional transmission operators (RTOs) 1278, 141, 165, 179, 221 regulatory assets value (RAV) 161 regulatory framework for distribution 261 regulatory framework for transmission 1346, 17980, 230231 in England and Wales 15962, 180 European see European Regulation 1228/2003 on Conditions of Access to the Network for Cross-border Exchanges in Electricity principles to guide 150157 regulatory paradigms 17, 2328, 27883 in US 135, 1656, 167, 170, 180 regulatory hold-up 152 regulatory ineciencies 667 regulatory risk 93 regulatory test 231, 232, 234, 235, 2378, 242, 281 relays and switches 15, 124, 144
299
reliability assessments 1678 reliability criteria see engineering reliability criteria reliability-driven transmission investments 15, 133, 1423, 154, 241 in PJM 1679, 175, 178 regulatory test allowing 238 versus economic transmission investment 133, 1757, 180 reliability of supply see engineering reliability criteria; reliabilitydriven transmission investments; security of supply remote monitoring and control equipment 15, 144 remote supply function 88, 89 renewable energies 13, 58, 59, 60, 61, 62, 63, 64, 65, 66, 69, 82 see also hydro-electric generators; wind power rent extraction goals 151 rental cost of transmission lines 94, 113, 115, 116, 120, 124 request for proposals (RFP) 155 Rseau de Transport de lElectricit (RTE) transmission network 136, 157 retailers, long-term contracts with 389 retiring generators economic factors determining 1213, 21, 31, 33, 47 growing number of 175 locational signals for 2556 returns to scale see constant returns to scale; economies of scale revenue adequacy 222 revenue requirement 161 Ring, B.J. 188, 194, 202 risk attitudes to 18, 153 regulatory 93 and technology choice 1314, 7380, 81 Rose, N. 67 Roselln, J. 125, 191, 220 Rotger, J. 118 Roulet, M. 77 Royal Academy of Engineering (RAENG) 68, 69, 72, 73
300
Index transmission investment in 146 transmission network organization in 1478 transmission pricing in 265, 266, 267, 268, 269, 270, 271 Spanish Ministry of Industry, Tourism and Trade 276 spark spread 77 stability ratings 128 standard market design (SMD) 1645, 191 standards 12, 158, 1623 State Aid 73, 81 static models 224 stock market values 77 Stoft, S.E. 52, 108, 110, 119, 126, 187, 224, 231 stranded costs 3 Strange, W.C. 334 strategic behaviour 23, 98, 1027 strategic generation investment problem 103, 1045 subcontracting 1234 subsidies 13, 16, 54, 67, 73, 81, 82, 150, 151 substation facilities 144, 230, 260 Sun, H. 124 sunk costs 30, 31, 33, 38, 47, 923, 247, 268 supernormal prots 26, 28, 47 Svenska Kraftnt 45 Sweden generation investment in 445 net inter-TSO payment of 271 transmission pricing in 249, 268, 271 Sweeting, A. 157 Switzerland net inter-TSO payment of 271 transmission pricing in 265, 266, 271 system balancing costs 160161, 162 tari pancaking 2467 Tasmania, HVDC link with 141, 155 technology, locational price as function of 21819, 223 technology changes see generating technologies, choice of telecommunications industry 66 TenneT 182
RPI-X remuneration 2345, 240, 260 RTEC (Rseau de Transport de lElectricit) transmission network 136, 157 RTOs see regional transmission operators (RTOs) Rubio, F.J. 245, 246, 286 Santaholma, J. 72, 79, 80 scale factor 253 scarcity rent 202 Scarf, H.E. 190 Schmalensee, R. 66, 67, 139, 146, 152 Schwartz, E.S. 35 Schweppe, F.C. 244 ScotlandEngland interconnector 182 second-best regulatory mechanism 151 security of supply 147 capacity margins and 46 transmission investment and 1415, 108, 119, 1245, 132, 137, 138, 234, 235, 240, 242, 259 see also engineering reliability criteria; reliability-driven investments sensitivity analysis 73 SERP 107 Seven-Year Forward Statements (NGC) 1589 shallow interconnection policy 134, 142, 170 share prices 46, 77 Sherk, C.J. 235 shirking 123 simulation studies 132 single system paradigm 247 Slovakia, transmission pricing in 265 Slovenia, transmission pricing in 265 SMD (standard market design) 1645, 191 Smeers, Y. 190, 205, 245, 246, 248, 250 SO procurement behaviour 154 social cost of congestion 172 social welfare maximization 278, 281 software technologies 147 Spain generation investment in 4, 57, 59, 65, 66, 70, 80 net inter-TSO payment of 271 requests for connection in 256
Index Texas congestion in 132 generation investment in 57, 58, 62, 65, 77 independent system operators covering 54 thermal eciency 41, 46, 71, 273, 275 thermal limits 195, 198, 206, 215 thermal power plants 16, 44 timeframe of investments 14, 745, 81, 124, 242, 244 Tirole, J.J. 41, 112, 113, 127, 134, 139, 151, 154, 156, 163, 169, 175, 177, 191, 201, 220, 221, 230 total revenue, generating plants 256 transaction costs 140 transcos performance-based regulations for see performance-based regulation (PBR) planning by 102 see also independent transco model transfer payments 89 transformer upgrades 144, 171 transmission investment in 79, 1415, 87127, 23242 coordination between distribution network investment and 17, 138, 231, 25862, 277, 280 coordination between generation investment and 1012, 1617, 1027, 125, 187228, 230231, 24358, 277, 27883 case examples 26377 see also regulatory framework for transmission transmission charges England and Wales 12, 37, 142, 158, 15960, 1612, 168, 169, 182, 249, 271 harmonization of pricing practices 15, 133, 232, 263 locational signals in 187228, 240250 access rules and 2578 case examples 26277 incremental charges for new network users 250255
301
for retiring generators 2556 non-transaction-based 246 PJM 142, 1646, 167, 16870, 173, 254 transmission cost functions 14, 93, 94, 100, 1012, 106, 112, 245, 248 transmission licences 158, 163, 234, 236 transmission line relief orders (TLRs) 131 transmission losses 8, 24, 146, 166 in distribution grids 280 economic incentives to reduce 235 in economic models of transmission investment 139, 175 locational signals related to 270 magnitude of 87, 100, 273, 274 mechanisms to account for 279 transmission network facilities, denition of 1367 transmission network organization 14750 transmission planning process, requirements of 1545 transmission rights 9, 88, 89, 10810, 138, 153, 163, 236, 241, 257 congestion revenue rights (CRRs) 10910, 111, 112, 115 nancial transmission rights (FTRs) 3, 10, 110112, 167, 168, 169, 171, 172, 1912, 221 and market power 110, 112, 247 paradox of 11213 planning incorporating 154, 231 transmission service price structures 1523 transmission signals, locational see locational transmission signals, long term Transmission System Security and Quality of Service Standards 158, 1623 transparency of long-term signals 21516, 2245 Transpower 283 tree pruning 15, 124 trigger price investment rule 323 Turvey, R. 67 TVO 14, 79 two-part taris 201
302
Index use of system charges in England and Wales 142, 158, 15960, 161, 168, 169, 182 regulatory framework for determining 1612 useful life of a power plant 273 useful power delivered 272 value of lost load (VOLL) 57, 36, 40, 176 vertical integration model 147 regulatory challenges associated with 149 viability constraint, TSO 150151 Vickers, J. 66 Victoria (Australia), HVDC link with 141, 155 Vogelsang, I. 107, 125, 192, 223 VOLL (value of lost load) 57, 36, 40, 176 voltage increases 144 voltage level, transmission charges related to 2578 wholesale market prices 389, 42, 74, 77, 120, 133, 140 willingness to pay 5, 10, 41, 153, 197, 201, 214, 2234 Wilson, R. 128, 138 wind power 65, 69, 103, 1067, 256 wireless network 66 Wolfram, C.D. 67, 157 Wolsey, L.A. 206, 228 Woolf, F. 191 Yan, J. 67 Yarrow, G. 66 yield to equity 14, 756 zero-congestion policy 14, 913, 95, 106, 119 strategic manipulation of 1023 zonal prices 3, 262, 267
uncertainty irreversible investment under 13, 315, 47 signals to reduce 255, 2568, 277 types of 3031 underwater links 155 unhedgeable congestion 172, 173, 1745, 177, 178 unit commitment problem 193, 202, 223 United Kingdom see England and Wales; ScotlandEngland Interconnector United States blackouts in 124, 126 cost of capital in 75, 76 generation investment in 2, 4, 1314, 37, 456, 47, 5365, 67, 68, 71, 72, 734, 778, 80, 81 legacy infrastructure in 1456 market power in generation in 126 obligation capacity in 6 regulatory framework for transmission in 135, 1656, 167, 180 reliability of supply in 1415 transmission congestion in 1312 transmission investment in 142, 1645, 180 see also PJM (PennsylvaniaNew JerseyMaryland) transmission network facilities in 136 transmission network organization in 148, 179 transmission pricing policy in 1645, 170 see also PJM (PennsylvaniaNew JerseyMaryland) unreliability 92 unserved load, cost of 120, 121, 122, 125 upgrading lines 125 US Energy Information Administration (EIA) 164

Investments in Electricity

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Investments in Electricity

Uploaded by

Copyright:

Available Formats

Competitive Electricity Markets and Sustainability

2.1 2.2 2.3 2.4 2.5 2.6 2.7 2A.1 3.1

Investments in competitive electricity markets: an overview

Competitive electricity markets and sustainability

Investments in competitive electricity markets: an overview

Competitive electricity markets and sustainability

Investments in competitive electricity markets: an overview

Competitive electricity markets and sustainability

Investments in competitive electricity markets: an overview

Competitive electricity markets and sustainability

Investments in competitive electricity markets: an overview

Competitive electricity markets and sustainability

. . . AND BACK TO THE COORDINATION BETWEEN INVESTMENTS IN GENERATION AND TRANSMISSION

Investments in competitive electricity markets: an overview

Competitive electricity markets and sustainability

OVERVIEW OF THE BOOK AND SYNOPSIS OF THE CONTRIBUTIONS

Investments in competitive electricity markets: an overview

Competitive electricity markets and sustainability

Investments in competitive electricity markets: an overview

Competitive electricity markets and sustainability

Investments in competitive electricity markets: an overview

Competitive electricity markets and sustainability

Investment and generation capacity

A SIMPLE MODEL OF CAPACITY

Investment and generation capacity

(1) Peaking Base load

The determination of electricity capacity and prices

Investment and generation capacity

Investment and generation capacity

(1) Peaking Base load

T* /MW-year (2) Peaking Base load /MW-year (3)

Hours/year /MW-year Base load

How the capacity mix aects revenues

WHAT IF THERE IS THE WRONG LEVEL, OR MIX, OF CAPACITY?

Investment and generation capacity

IRREVERSIBILITY AND UNCERTAINTY

Investment and generation capacity

Investment and generation capacity

Investment and generation capacity

THE VALUE OF LOST LOAD AND THE LOSS OF LOAD PROBABILITY

Investment and generation capacity

Investment and generation capacity

PAYING FOR CAPACITY IN PRACTICE6

Investment and generation capacity

INVESTMENT IN LIBERALISED MARKETS

Proportion of peak demand 0.3 0.2 0.1 0 0.1 1991/92

Capacity margin Gross investment Net investment

Source: National Grid Company (NGC).

Investment in England and Wales

Investment and generation capacity

Proportion of peak demand 0.3 0.2 0.1 0 0.1 1992

Capacity margin Gross investment Net investment

Proportion of peak demand 0.5 0.4 0.3 0.2 0.1 0 1992

Capacity margin Gross investment Net investment

Capacity margin Gross investment Net investment

Investment and generation capacity

Capacity margin Gross investment Net investment

Investment in the United States

Investment and generation capacity

A MATHEMATICAL MODEL OF CAPACITY AND PRICING

Alternatively, we can write it as a Lagrangean, which gives us: Max L

This gives us a rst-order condition for output: 0 qit Bt qit ci 0. (2A.6)

Investment and generation capacity

(1) Peaking Base load