Excerpts from Part 3. Power System Economics (389p) Designing Markets for Electricity 2001 Steven Stoft

Transcription

1 Power System Economics (389p) Designing Markets for Electricity 2001 Steven Stoft Excerpts from Part 3 Part 1: Introduction 88 p. Part 2: Price Spikes, Reliability and Investment 110 p. Part 3: Market Architecture 78 p. 3-1 Key Questions of Market Architecture 5 p. 1 Spot Markets, Forward Markets and Settlements 2 Controversies 3 Simplified Locational Pricing 3-2 The Pure Bilateral approach 5 p. 1 No central market 2 Central Coordination without Price 3 A Pure Transmission Market 3-3 Why Have a Spot Energy Exchange 7 p. 1 A Pure Spot Market for Energy 2 Conclusions 3-4 Real-Time Pricing and Settlement 12 p. 1 The Two-Settlement System 2 Setting the Real-time Price 3 Ex-Post Prices: The Trader s Complaint 3-5 Why Have a Day-ahead Market? 9 p. 1 When Marginal-Cost Bidding Fails 2 Reliability and Unit Commitment 3 Efficiency and Unit Commitment 4 The Congestion Problem 3-6 Day-Ahead Market Designs 13 p. 1 Defining Day-Ahead Auctions 2 Four Designs 3 The Impact of Startup Insurance 4 Transmission Bids and Virtual Bids 3-7 Multi-Part Unit Commitment? 16 p. 1 How Big is the Unit Commitment Problem? 2 Market Design #1: A Pure Energy Auction 3 Market Design #3: A Unit Commitment Auction 3-8 A Market for Operating Reserves 11 p. 1 Bid-Based Pricing. 2 Opportunity-based pricing Part 4: Market Power 56 p. Part 5: The One-Line Network 63 p.

2 Chapter 3-1 Questions of Market Architecture MARKET ARCHITECTURE CONCERNS THE KEY DESIGN ELEMENTS. While Part 2 abstracts from all questions of market design to focus on market structure, Part 3 considers alternative designs for the realtime market, the day-ahead forward markets and the relationship between the two. It also discusses several controversies, such as the degree of centralization, that have often plagued the design process. Design elements are considered in just enough detail to allow comparisons between the main alternative approaches. While Part 3 moves forward from real-time it does not move past the dayahead market, and it does not consider private bilateral markets that operate beside the markets organized by the system operator (SO). It focuses only on those markets that are typically part of an ISO design. Section 1: Spot Markets, Forward Markets and Settlements. Forward markets are financial markets while the realtime (spot) market is a physical market. To the extent power sold in the day-ahead market is not provided by the seller, the seller can buy replacement power in the spot market. This is the basis of the two-settlement system that underlies one standard market design in which the SO conducts both day-ahead and spot energy markets. Section 2: Controversies. Three major controversies have beset the design of power markets. First is the conflict over how decentralized the market should be. One view holds that both day-ahead and spot markets should be bilateral energy markets, and the SO should have no dealings that involve the price of energy but instead sell (or ration) only transmission. The second conflict arises only if the day-ahead market is to be run by the system operator (centralized). One view holds that such an auction market should utilize multi-part bids to solve the unit commitment problem in the

3 CHAPTER 3-1 The Key Questions of Market Architecture 3 traditional way. Another view holds that bids should specify only an energy price. The third conflict concerns the level of detail at which locational prices are computed. The nodal view typically argues for hundreds or thousands of locations, while the zonal view typically calls for well under one hundred. This controversy is less fundamental and is not considered in Part 3. Section 3: Simplified Locational Pricing. All markets discussed in Part 3 produce energy prices that are locationally differentiated. The theory of such prices is not presented until Part 5, so a summary of their properties is given in Section 3. These prices are competitive and thus independent of the market s architecture. Because they are competitive they have the normal properties of competitive prices; they minimize production cost for a given level of consumption, and they maximize net benefit SPOT MARKETS, FORWARD MARKETS AND SETTLEMENTS Trading for the power sold in any particular minute begins years in advance and continues until real time, the actual time at which the power flows out of a generator and into a load. This is accomplished by a sequence of markets which often overlap. The earliest markets are typically forward markets that trade non-standard long-term contracts. Futures contracts typically cover a month of power during on-peak hours and are sold up to a year or two in advance. Trading continues in less formal markets until about one day prior to real time. Typically, just as this informal trading peters out, the system operator holds its day-ahead (DA) market. This is often followed by an hour-ahead (HA) market and a realtime market also conducted by the system operator. All of these except the realtime market are financial markets in the sense that suppliers need not own a generator to sell power. The realtime market is a physical market, as all trades correspond to actual power flows. While the term spot market is often used to include the DA and HA markets, this book will use it to mean only the realtime market because it is the only physical market. A customer in the DA market does not purchase electricity but rather a promise to deliver electricity. If the promise is not kept, the supplier must buy the power it failed to deliver in the spot market. It is possible to sell power in the DA market without owning a generator and cover the sale with a spot market purchase. A clever speculator can make money on such purely financial transactions, but he cannot trade only in the spot market. Financial markets need a way for traders to unwind their position. If a supplier sells, in a financial market, 1600 MWh to be delivered evenly over the

4 4 PART 3: Market Architecture sixteen peak hours of July 1, this should impose only a financial commitment. Typically such a sale includes a clause for liquidated damages; if the power is not delivered, the supplier must pay the cost of replacement. This always allows the supplier the option of buying replacement power, and, except in the most extraordinary conditions, this can be done in a subsequent market. The most formal arrangement for purchasing replacement power occurs in the system operator s markets. Any power that is sold in the DA market but not delivered in real time is deemed to be purchased in real time at the spot price of energy. This is called a two-settlement system and has a number of useful economic properties. They are discussed in Chapter CONTROVERSIES Three main controversies regarding architecture have beset the design of many power markets. 1 Central vs. bilateral markets 2 Exchanges vs. pools 3 Nodal vs. zonal pricing The first two controversies both concern the amount of centralization. In theory all power trades could be handled by bilateral markets in which private traders trade directly with each other, or through a middleman (power marketer), but not with an exchange or pool. This approach is particularly cumbersome for balancing the system in real time. Once it is admitted that centralization is needed, an attenuated form of the same controversy questions the extent to which the system operator should provide coordination. Should it collect large amounts of data on generators and compute an optimal dispatch, or should it let generators signal these parameters indirectly through the energy prices they bid? Last, there is a controversy over how finely the system operator should compute locational energy prices. A nodal pricing approach would define more than a thousand distinct locations in California. When the market was first designed the advocates of zonal pricing suggested that two zones would be sufficient, though many more were added later around the edge. More were subsequently required in the interior. This is the least fundamental of the three controversies and is not discussed in Part 3. Central vs. bilateral markets. The first two controversies concern the role of the system operator (SO). Some wish to minimize its role at almost any cost. Chapter 3-2 takes up the question of whether completely bilateral markets are possible and concludes that the system operator must perform a centralized allocation of transmission rights, but with the use of sufficient penalties and

5 CHAPTER 3-1 The Key Questions of Market Architecture 5 curtailments the SO could be kept from making any trades. Of course, its influence on the market would still be pervasive. Chapter 3-3 considers the possibility of a centralized spot market in transmission only. This would allow private bilateral markets to provide the spot energy market, but because this arrangement makes the realtime balancing of the system difficult and expensive, it is rejected in favor of an energy spot market (realtime balancing market) run by the SO. Chapter 3-5 considers the same question for the DA market, but in this case the answer is less obvious as the time pressure is far less severe. Here the answer hinges on the unit commitment problem and the need for coordination. Although a private bilateral market would cause much less inefficiency, there appears to be a strong case for at least the minimal central coordination that can be provided by a pure-energy market run by the system operator. Exchanges vs. Pools. Unit commitment is the process of deciding which plants should operate. Integrated utilities have always done this using a centralized process that takes account of a great deal of information about all available generation. If this is done incorrectly the wrong set of plants may be started in advance which can lead to two problems: (1) inefficiency and (2) reduced reliability. As just noted, a bilateral market solves this problem poorly, so a centralized DA market is preferred. The second controversy concerns the extent of central coordination. There are two polar positions: let generators bid only energy prices (1-part bidding) or let generators bid all of their costs and limitations (multi-part bidding). One-part bidding allows the SO to select the amount of generation to commit in advance but gives it very little information about the generators costs and limitations. Consequently it can apply none of the usual optimization procedures, but it can provide some coordination by purchasing the correct quantity of power a day ahead. With multi-part bids, the SO can select bids on the basis of the traditional optimization procedure. Typically, this controversy focuses on comparing the existence, efficiency and reliability of the market equilibria for 1-part and multi-part auctions. Chapter 3-7 demonstrates that both types of markets have equilibria that exist and that are likely to be very efficient and reliable. If there is a problem, it is that markets have difficulty in arriving at the equilibrium of a 1-part auction market when costs are non-convex as they are in power markets. This difficulty arises from the extensive information requirements of 1-part bidding. In such an auction, competitive suppliers should not simply bid their marginal costs but must estimate the market price in advance in order to determine how to bid. This is a far more difficult task than simply bidding one s own costs as required by normal competitive markets or by a multi-part bid auction. Unfortunately, not enough is known about how such markets perform in practice so no conclusion can be drawn as to which is preferable, though it seems plausible that a two or three part auction could be designed to capture

6 6 PART 3: Market Architecture most of the advantages of both extremes. Fortunately, the unit commitment problem is small enough that it may not matter much which design is adopted. (While this controversy is often lumped with the nodal pricing controversy, it has only a little to do with locational prices.) SIMPLIFIED LOCATIONAL PRICING Energy prices differ by location for the simple reason that energy is cheaper to produce in some locations and transportation (transmission) is limited. When a transmission line reaches its limit, it is said to be congested, and it is this congestion that keeps energy prices from equilibrating between different locations. For this reason locational pricing of energy is equivalent to congestion pricing. The pricing of congestion is not explained until Part 5, but Part 2 makes use of some basic concepts of congestion pricing. These can be explained without delving into the underlying economics. The interested reader will find all of the following results explained in Chapters 5-3, 5-4 and 5-5. Locational prices of energy are just competitive prices, and these are unique. They are determined by supply and demand and have nothing to do with the architecture of the market, provided it is a competitive market. This means a purely bilateral market that is perfectly competitive will trade power at the same locational prices as a perfectly competitive, centralized nodal-pricing market. Of course, a bilateral market is likely to be a little sloppier with its pricing and not arrive precisely at the competitive equilibrium, but given enough time and small enough transaction costs it should arrive at the full set of nodal prices just as efficiently as a fully centralized market. Because there is a unique set of locational prices, there is also a unique set of congestion prices, which will also be called transmission prices. Again, these are determined be competition and supply and demand conditions and have nothing to do with the market architecture, provided the market is perfectly competitive. If the competitive energy price at X is $20/MWh and the price at Y is $30/MWh, then the price of transmission from X to Y is $10/MWh. Transmission prices are always equal to the difference between the corresponding locational prices. If this were not true, it would pay to buy energy at one location and ship it to the other. In that case arbitrage would change the energy prices until this simple relationship held. This relationship can be expressed as follows: P XY = P Y P X, which is read, the price of transmission from X to Y equals the price of energy at Y minus the price of energy at X.

7 CHAPTER 3-1 The Key Questions of Market Architecture 7 This relationship is all that is needed to understand Part 3, but it contains one surprise that deserves attention. Because transmission prices are not based on a cost of transporting the power (we ignore the cost of losses which is minor) but are based instead on the scarcity of transmission (line limits), transmission costs can be negative. In fact, if the cost of transmission from X to Y is positive, then the cost from Y to X is certain to be negative. This is a direct result of the above formula. This peculiarity can be understood by noting that when power flows from Y to X it exactly cancels (without a trace) an equal amount of power flowing from X to Y, thus making it possible to send that much more power from X to Y. A second consequence of the above formula is that the cost of transmitting power from X to Y does not depend on the path chosen All of the markets discussed in Part 3 are assumed to compute locational prices and to operate competitively. Consequently, they will produce the locational energy prices and transmission prices just described. In spite of their ubiquitous presence, the reader will not need to understand details of how locational prices are computed and nor rely on either of the properties just discussed. They are presented merely to provide context.

8 Chapter 3-2 The Pure Bilateral Approach THE MIN-ISO APPROACH TO POWER MARKET DESIGN, POSTULATES THAT THE LESS COORDINATION THE BETTER THE MARKET. 1 This philosophy underlies attempts to keep the system operator out of the energy market. In other markets this would be possible. For example the U.S. Department of Transportation does not buy or sell trucking services, it just provides highways and charges for their use. This chapter examines the inefficiencies that would result from keeping the system operator out of the real-time energy market. Section 1: No Central Coordination. Imagine an electricity market run as a system of highways. Every power injection by a generator could be measured and charged to pay for the cost of the system. Any generator could sell power to any customer and deliver that power by injecting it at the same time the customer used it. Unfortunately, there would be no way to prevent theft. With trading fully decentralized, no one would know who had paid and who had not. Section 2: Central Coordination without Price. The simplest actual proposal for a power market suggests that all trades be registered with the system operator who would accept only sets of trades that did not cause any reliability problem. This proposal takes the first step towards the enforcement of reliability and the prevention of power theft, but it does not take the second step of specifying what happens when traders violate their schedules. In this system the operator knows nothing of prices, and imposes no penalties. 1 Although, it does recognize the need to enforce transmission constraints it proposes to do this with arbitrary curtailments and without the use of any market mechanism..

9 CHAPTER 3-2 The Pure Bilateral Approach 9 Section 3: A Pure Transmission Market. The next step is to sell transmission rights. This improves on the previous system by removing the arbitrariness from the distribution of transmission rights and reduces the transaction costs by centralizing the transmission market. It does not solve the problem of balancing the real-time market. This could be accomplished with penalties and curtailments. These would induce the development of private system operators. Because system operation is a natural monopoly it would be necessary to limit their size. Although this could produce a workable power market it would have much higher transactions costs than necessary NO CENTRAL COORDINATION Most markets do not need any central coordination. To understand why a power market does, imagine one without coordination. As with highways, any supplier could use the wires and would be charged for their usage. The grid owner would meter each supplier s output and impose a per MWh or a annual peak-mw charge sufficient to pay for the cost of the wires. Without any central coordination a supplier could sign a contract with a customer for 100 MW all day on April 1 at $40/MWh. The generator would then inject 100 MW and the load would take 100 MW and pay the generator. There would need to be thousands of such contracts, but this would be no different from other markets. The most fundamental problem with this design is that loads would steal power. Why have a contract? Just turn on the lights. The grid operator does not care because it gets paid for every watt transmitted it could measure either all injections or all withdrawals of power. Other generators and customers care because their power is being stolen, but they have no way to prove this. Because electricity cannot be directed by a supplier to its intended destination, there is no way to prevent theft. Power from every generator flows to every load. A second problem, which would only occur if the first could be solved, is that certain transmission lines would be overused. There are only two approaches to protecting lines: (1) load and generation can be disconnected to reduce flow on overloaded lines, and (2) overloaded lines can be taken out of services. The system is already programmed to take lines out of service instantly and automatically when they reach their limits. While this protects lines, it also destabilizes the system which then requires extraordinary central coordination to restore its balance. Centrally controlled circuit breakers could be installed on all loads and generators, and these could be used to prevent

10 Chapter 3-3 Why Have A Spot Energy Exchange? FORWARD MARKETS SELL PROMISES OF POWER, THE SPOT MARKET SELLS ELECTRICITY. But the system operator need not be allowed to trade electricity and instead could sell transmission to bilateral traders of electricity. Although there is nothing inherently wrong with this process, it is slower than centralized energy trading. When balancing the power system, time is of the essence. In fact the price mechanism is orders of magnitude too slow to do the complete job. So to get the maximum benefit from markets, the fastest market must be used for balancing, and that is a centralized locational energy market. Section 1: A Pure Spot-Market For Energy. Every deviation from balance is handled by a sequence of procedures the first of which take place in less than a tenth of a second. Because great speed is required, the initial process, to the extent it is not automatic must be centrally directed. Current pricing mechanisms are not much use in time frames under ten minutes. But at some point, the job of equating supply and demand can be handed over to a market mechanism. This is the real-time or balancing market. System balance for a control area is determined by a combination of net inflow and system frequency. Even if every bilateral trade using a particular control area is in perfect balance, the system operator will be directed to either increase or decrease generation if the system frequency is off, which it almost always is. Consequently the operator must have some control over energy. Section 2: Conclusions about the Real-Time Market. The essence of the problem of system balancing is speed. Once the most time-critical part of balancing has been handled their comes a point where price can do the job. But

11 CHAPTER 3-3 Why Have A Spot Energy Exchange? 15 in order to maximize the usefulness of the market, the fastest market should be used, and that is a centralized energy market. A transmission market can only sell transmission when two equal but opposite energy trades have been found. Thus a transmission market is just an energy market with restrictions, and it is inherently more expensive when great speed is needed. Of course by spending more on market infrastructure any market can be made faster. The usual proposal is to impose penalties on bilateral trades that get out of balance. This appears to keep them in balance cheaply, or even at a profit. (To stay in balance they must adjust very quickly, so this is method of speeding up the market.) But penalties simply hide the costs, which must be born by those with bilateral contracts. It is wiser to use a locational energy market as the real-time market and leave bilateral trading to forward markets that can proceed at a more leisurely pace A PURE SPOT-MARKET FOR ENERGY The previous chapter considered three approaches to energy trading that might have eliminated the need for a centralized real-time energy market, but all were needlessly expensive. The real-time market needs fully centralized coordination because electrical energy is not stored and so supply must equal demand second by second. In fact, this requirement is so severe that even a centralized energy market requires several types of reserves. Regulation operates most quickly because it is automatically controlled. Next comes 10- minute spinning reserves, 10-minute non-spinning reserves and finally 30- minute non-spinning reserves. Each of these requires many generators to be under the direct control of the system operator. The balancing market overlaps with the 10-minute reserve markets, which, by providing a safety net for emergencies, allow the more sluggish and less reliable mechanism of a market-clearing price to be utilized in this time frame. Sluggish as this process is by engineering standards, real-time electricity prices are probably the most nimble and effective prices to be found anywhere. On a daily basis they balance supply and demand to within a few percent as these change at rates of up to 20% per hour. Rarely do any other markets see price changes of this speed and then only during panics. More typically prices adjust 1000 times more slowly. Those who demand that the real-time energy market be taken out of the hands of the SO to be replaced by a transmission market and uncoordinated bilateral energy trades base their demands on ideology and not a study of the capabilities of present-day markets. The main substantive objections to energy

12 Chapter 3-4 Real-Time Pricing and Settlement CUSTOMERS AND GENERATORS SHOULD RESPOND TO THE SPOT PRICE AS IF THEY HAD BOUGHT AND SOLD ALL THEIR POWER IN THE SPOT MARKET. Typically, however, they buy and sell almost all of their power in forward markets. Fortunately, the correct settlement system insulates the real-time markets and preserves their incentives. Contracts for differences (CFDs) insulate forward contracts from the realtime (spot) price even though loads and generators trade all of the power in their spot market after having already traded it in the forward markets. From a trader s point of view, the main problem with spot markets is timing of transmission costs. These are not posted in advance but are determined along with the price of energy in the spot market. This is often called ex-post pricing. Although determining the real-time price is simple in theory (just set price so supply equals demand), it is complex in practice. Most real time markets have a large number of rules, and there is little consistency in these rules between systems. Their purposes vary. Some are designed to limit market power, some to protect the system from sudden shifts in supply and demand that might result from or cause price instability. Others are the result of software anomalies or various superstitions, but this chapter will ignore such complexities and stick to basics. Section 1: The Two-Settlement System. If the system operator runs a dayahead (DA) and a real-time (RT) market, generators should be paid for power sold in the DA market at the DA price regardless of whether or not they produce the power. In addition, any real-time deviation from the quantity sold a day ahead should be paid for at the real-time price. This system allows an

13 CHAPTER 3-4 Real-Time Pricing and Settlement 23 almost complete separation between the markets. Even if a generator sells essentially all of its power in the day-ahead market, it will still have the correct incentive to deviate from that contract in response to realtime prices. In real time the generator has the same incentives as if it were selling all of its power in real time. Loads are treated analogously with the same effect. If bilateral traders use contracts for differences (CFDs), and if the spot price does not vary with location, bilateral trades will be unaffected by the spot price even though the generators and loads sell all of their power in the spot market, provided they generate according to their contract. In spite of this, CFDs leave them with the proper incentive to deviate in ways that benefit the deviating party and leave the other parties unaffected. Section 2: Ex-Post Prices: The Trader s Complaint. Spot prices that differ by location impose transmission costs on traders. These cannot be avoided by the use of CFDs, and they make trade risky. Time-of-use transmission charges could be posted in advance as an approximation of congestion charges, but their inaccuracy would cause an inefficient dispatch. A reservation system would be required to avoid the most serious inefficiencies. A market in transmission rights would be preferable to reservations sold at regulated prices. Such markets exist, but are limited and illiquid. Technical and practical difficulties have prevented the development of more robust markets, but these problems are receiving considerable attention. Transmission rights can be financial or physical, and financial rights can be used effectively as a reservation to assure complete protection from realtime transmission charges. Section 3: Setting the Real-Time Price. The real-time price should be set to clear the market. As not all response to price changes is reflected by supply and demand bids, if price is set strictly on the basis of these bids, it will overshoot. This problem will grow as real-time pricing becomes more prevalent, which will make it necessary for the system operator to improve its understanding of the dynamic affects of price changes.

14 24 PART 3: Market Architecture THE TWO-SETTLEMENT SYSTEM If a supplier sells most or all of its power in the forward markets, the realtime price would appear to have little effect on that producer s behavior. In a properly implemented two-settlement system the opposite is true. The supplier will behave as if it were selling its entire output in the realtime market and will still behave, when selling in the forward market, as if that were its final sale. In this way, if the markets are competitive, suppliers (and also consumers) will behave optimally in both markets. Separation from Forward Transactions Say a supplier sells Q 1 to the system operator (SO) in the day-ahead market for a price of P 1. If this amount of power is delivered to the real-time market, the settlement in the day-ahead (DA) market will hold without modification. But what if none is delivered, or more than Q 1 is delivered? In either case the DA settlement should still hold, but there should be an additional settlement in the real-time market. If no power is delivered to the real-time market, the supplier is treated as if it had delivered the amount promised in the DA market, Q 1, and purchased that amount in the real-time market instead of generating it. Consequently the supplier is still paid P 1 for Q 1, but is also charged P 0, the realtime price, for the purchase of Q 1. In general, if a supplier sells Q 1 in the DA market and then delivers Q 0 to the real-time market, it will be paid: Supplier paid: Q 1 P 1 + (Q 0 Q 1 ) P 0 This is called a two-settlement system. If a customer contracts for Q 1 and then takes only Q 0 in real time, it is charged exactly the amount that the supplier is paid. Result A Two-Settlement System Preserves Real-Time Incentives When the real-time market is settled by pricing deviations from forward contracts at the real-time price, supplier and customers both have the same performance incentives in real time as if they traded all of their power in the real-time market. The incentive of this settlement rule can be revealed by rearranging the terms as follows: Supplier paid: Q 1 (P 1 P 0 ) + Q 0 P 0 When real-time arrives, Q 1 has been determined in the day-ahead (DA) market. Assuming the market is competitive, the generator has no control over either price, and by real time the first term will be taken as given. The first term will be viewed as a sunk cost or an assured revenue. This leaves the second term

15 Chapter 3-5 Why Have a Day-ahead Market? POWER MARKETS ARE DIFFICULT TO COORDINATE BECAUSE THEY DO NOT SATISFY THE ASSUMPTIONS OF A CLASSICALLY COMPETITIVE MARKET. Under classical assumptions, suppliers need to know only their own costs, and no central coordinator is needed. For a power market to perform efficiently, either it must be centrally coordinated or suppliers must know a great deal about the market equilibrium price in advance. The root of the problem is generation costs that fail to satisfy a key economic assumption used to prove the efficiency of competitive markets. 6 Because the proof of efficiency fails, uncoordinated power markets are often believed to have no equilibrium or only a very inefficient one. In fact they have equilibria that are extremely efficient but difficult to discover. This chapter argues that at least a small amount of central coordination is well worth while and should take the form of a centralized day-ahead market. The question of whether this market should perform a full centralized unit commitment is discussed in Chapter 3-7. In a classic competitive market, suppliers can offer to supply (in a bilateral market) or to bid (in an auction market) according to their marginal cost curve. When all do so, the market discovers a perfectly efficient competitive equilibrium. But with non-convex costs of generation, it becomes necessary for generators to bid in a more complex manner. 6 Part 2 focused on the consequences of the far more serious demand-side flaws in contemporary power markets. Part 3 ignores these and focuses on problems with generation costs that are very small but unavoidable.

16 CHAPTER 3-5 Why Have a Day-Ahead Market? 35 One market design allows suppliers to continue bidding their marginal costs but include other costs and limitations in their multi-part bids. This has the advantage of allowing suppliers to base their bids on easily obtained information: their own costs. Another approach can take the form of a decentralized bilateral market or a centralized market with one-part energy bids. In both cases, suppliers must account for all of their costs and limitations in their energy price bid so they do not bid their true marginal costs. 7 With this approach, suppliers must utilized considerable information about the external market. This chapter argues that the second approach, with its formidable information requirements, causes coordination problems that are more severe in bilateral markets than in a centralized one-part-bid energy auction. It concludes that the coordination problems in a bilateral market will be substantial enough that this approach should not be adopted for the day-ahead market. If bilateral markets promised some important advantage, their reduction of efficiency and reliability might be justified. But bilateral markets have higher transaction costs and are less transparent than a public auction. They are also impossible to use for settling futures contracts. Finally, adopting a centralized day-ahead market does not preclude the operation of a bilateral day-ahead market. Section 1: When Marginal-Cost Bidding Fails. A cost function is nonconvex if costs increase less than proportionally with output. Startup costs, no-load costs, and several other components of generation costs contribute to making them non-convex. Consequently generation costs fail to satisfy the conditions necessary to guarantee a competitive equilibrium. This does not necessarily prevent the market from being very efficient, but will cause competitive suppliers to bid above marginal cost if they cannot bid their startup and no-load costs directly. The amount they should bid above marginal costs depends on the outcome of the market which can only be estimated at the time of bidding. Section 2: Reliability and Unit Commitment. In a bilateral market, generators must commit (start running) without knowing which other 7 Day-ahead bilateral markets could allow very complex contracts but do not because it would make contracting too expensive.

17 36 PART 3: Market Architecture generators have decided to commit. Because starting up is costly, they will not start unless they expect to cover this cost, an outcome which depends on how many other generators have started and will compete against them the next day. The uncertainties of this problem cause a random level of commitment in a bilateral market, and this decreases reliability. Section 3: Efficiency and Unit Commitment. The randomness in the level of commitment, causes inefficiency. Although this randomness is caused by information problems, a similar phenomenon can occur because of the lack of a market clearing price. While this second phenomenon has received more attention, it is probably of less practical importance. Section 4: The Congestion Problem. Transmission bottlenecks (congestion) cause prices to differ by location and make the price more difficult to estimate in advance. The congestion problem significantly exacerbates the information problem of bilateral and one-part bid markets, because these need to know the market price in advance. In bilateral markets, this leads to a significant increase in randomness and inefficiency. A centralized one-part bid auction provides much of the coordination needed to take account of congestion efficiently WHEN MARGINAL-COST BIDDING FAILS The normal description of a competitive market, found in earlier chapters, requires bids that reflect marginal costs. In spite of all suppliers bidding their marginal costs, they were able to recover their fixed costs through inframarginal (scarcity) rents. This situation obtains in markets that satisfy the assumptions needed to prove the existence of a competitive equilibrium. These assumptions are well approximated by many markets, but certain aspects of power markets fail to satisfy these classic assumptions. Without these assumptions, economics cannot prove a market has a competitive equilibrium. This is a less devastating critique of a market than is often supposed. The concept of a competitive equilibrium is quite narrow, and a market does not cease to function without one. Instead it produces some other type of equilibrium which may involve some market power or some randomness. These flaws may reduce its efficiency very little. Economists, aware of this fact, depend on it when arguing that their results apply to the real world, which never quite conforms to their assumptions.

18 Chapter 3-6 Day-Ahead Market Designs THERE ARE MANY POSSIBLE DESIGNS FOR A CENTRAL DAY-AHEAD MARKET, BUT ALL CAN BE DESCRIBED AS AUCTIONS. The most obvious design just prices energy like the real-time market. A different approach turns the system operator (SO) into a transportation service provider who knows nothing about the price of energy but instead sells point-to-point transmission services to energy traders. Either of these approaches presents generators with a difficult question. Some generators must engage in a costly startup (commitment) process in order to produce at all. Consequently, when offering to sell power a day in advance, a generator needs to know if it will sell enough power at a price high enough to make commitment worthwhile. Some day-ahead (DA) auctions require complex bids that describe the generators startup costs and other costs and constraints and solve this problem for the generators. If the SO determines that a unit should commit, it insures all its cost will be covered provided the unit does commit and produces according to the accepted bid. Such insurance payments are called side payments, and their effect on long-run investment decisions is considered in Section The three approaches just named, energy, transmission and unit commitment can also be combined into a single auction that allows all three forms of bid; this is how PJM s day-ahead market works. Generators can offer complex bids and receive startup-cost insurance if they are selected to run. Anyone can offer to buy or sell energy with simple energy bids, and traders can request to buy transmission from point X to point Y without mentioning a price for energy. PJM considers all of these bids simultaneously and clears the

19 CHAPTER 3-6 Day-Ahead Market Designs 45 market at a set of locational energy prices together with startup-cost insurance. The differences in energy prices from location to location determine the prices for transmission. Section 1: Defining Day-Ahead Auctions. All day-ahead markets organized by system operators are auctions. Market participants submit bids, and the auctioneer (the SO) arranges trades according to a simple principle: maximize the net benefit as defined by the bids. If a customer offers to pay $40 and manages to buy for $30, the net benefit is $10. The calculation for suppliers is similar. The auction accepts the set of bids that maximize the sum of these net benefits and sets prices so that all trades are voluntary. The four day-ahead markets discussed here follow these simple principles and differ only in the type of bids they allow. Some allow bids for energy, some for transmission, and some allow complex bids that specify many costs and limitations for each generator. Section 2: Four Day-Ahead Market Designs. Each auction is specified by three sets of conditions: bidding, determining which bids are accepted, and determining the payments associated with the accepted bids. Market 1, a pure energy market, determines nodal prices. Market 2, trades only transmission and involves no prices for energy. Market 3 adds unit commitment to Market 1. Market 4 combines the features of the other three and is modeled on the current PJM market. Section 3: Overview of the Day-Ahead Design Controversy. Forward markets are bilateral and realtime markets are centralized. The day-ahead market can be designed either way and this causes a great deal of controversy. The nodal pricing approach specifies an energy market with potentially different prices at every node (bus), and, almost always, specifies that the auction should solve the unit commitment problem as well. This requires a great increase in complexity of bids. The bilateral approach specifies that energy trades take place between two private parties and not between the exchange and individual private parties. To trade energy, the private parties require the use of the transmission system, so the system operator is asked to sell transmission. Market 2, takes a purely bilateral approach, while market 3, takes the full nodal pricing approach. Market 1, the simplest market, implements nodal pricing, but not unit commitment. Market 4, the most complex, implements all the features from both the bilateral approach and the nodal pricing approach.

20 46 PART 3: Market Architecture DEFINING DAY-AHEAD AUCTIONS This chapter concerns day-ahead markets run by system operators. These take the form of either exchanges or pools and are operated as auctions in which the process of selecting the winning bids is often complicated by transmission and generation constraints. Consequently, the selection process often requires the use of enormously complex calculations and sophisticated mathematics. Unfortunately, the outlines of the mathematics are often presented as a way of explaining the auction. This is unnecessary, often confusing, and generally less precise than an approach that focuses on the intent of the calculation instead of on the mechanics of the calculation. A Simplified Description of Auctions A bid acceptance procedure is often presented as a linear programming problem represented by several large sets of inequalities, a dozen sets of variables, and an objective function. This representation is generally an approximation to the actual program and does not account for such powersystem procedures as contingency analysis. For the purpose of defining the market and understanding its behavior, it is more useful and accurate simply to specify that production cost is to be minimized subject to transmission and generation constraints. This is the problem that the accepted bids must solve; linear (or nonlinear) programming is one possible technique for finding the solution. Ideally, before the market is implemented, the actual calculation technique should be tested to see if it is accurate enough to produce a reasonably efficient market. This is, however, no excuse for presenting the auction economics as a linear programming problem. Avoiding the details of the computation makes it easier to focus on more important economic considerations such as restrictions on the form of bids, how the winning bids are paid or charged, and penalties for non-performance. The following section presents such fundamental information for four types of dayahead markets and follows certain conventions to facilitate the comparison of these markets. Determining Quantities Auctions must determine the quantities sold and purchased and the price. Although the two are closely related they are separate problems, and the same set of bids can yield the same quantities but different prices under different auction rules. From an economic perspective, it is quantities that determine efficiency, and prices are important mainly to help induce the right trades.

21 CHAPTER 3-6 Day-Ahead Market Designs 47 In all four auctions described here, quantities of accepted bids are selected to maximize total net benefit. This assumes the bids reflect the bidders true costs and benefits. Although they may not, assuming that they do generally encourages truthful bidding. Total net benefit is the sum of customer and supplier net benefit, but it is also the benefit to customers minus the cost to suppliers. This simplification helps explain the role of price as well as the economist s attitude towards price, as an example will make clear. If a customer bids 100 MWh at up to $5,000/MWh, and the bid is accepted, the benefit to the customer is $500,000. If the market price is $50/MWh, the customers cost is $5,000 and net benefit is $495,000. Similarly, if a generator bids 100 MW at $20/MWh, its cost is presumed to be $2000. If the market price is again $50/MWh, its net benefit will be 100x($50 $20), or $3000. Writing this calculation more generally reveals that the price played no role in determining total net benefit. Total Net Benefit = Qx(V P) + Qx(P C) = Qx(V C), where Q is the quantity traded, V the customers value, C the supplier s production cost, and P is the market price. Thus the problem of maximizing net benefit can be solved independently of any price determination. In an unconstrained system, net benefit can be maximized by turning the demand bids into a demand curve and the supply bids into a supply curve and finding the point of intersection. This gives both the market price and a complete list of the accepted supply and demand bids. Unfortunately transmission constraints and constraints on generator output (e.g. ramp-rate limits) can make this selection of bids infeasible. In this case it is necessary to try other selections until a set of bids is found that maximizes net benefit and is feasible. This arduous process is handled by advanced mathematics and quick computers, but all that matters is finding the set of bids that maximizes net benefit, and they can almost always be found. Determining the Market Price In an unconstrained auction, the market price is given by the intersection of the supply and demand curves. The price determined by supply and demand is the highest of all accepted supply bids or the lowest price of an accepted demand bid. It depends on whether the intersection of the two curves occurs at the end of a supply bid, and in the middle of a demand bid, or vice versa. When the demand curve is vertical, the intersection is always in the middle of a supply bid, and the price is set to the supply bid price. Whichever curve is vertical at the point of intersection has an ambiguous marginal cost or value (See Chapter 1-5). If the demand curve has a horizontal segment at $200 that intersects a vertical part of the supply curve that changes from $180 to $220, then the marginal cost of supply is undefined but is in between the left-hand marginal cost of $180 and the right-hand marginal cost

22 48 PART 3: Market Architecture Figure Either marginal cost or marginal value is ambiguous. of $220. Consequently it causes no problem to say that the market price equals both the marginal cost of supply and the marginal value of demand. [fig] Consider how net benefit changes when an extra kw is added to the total supply of power at zero cost. This will shift the supply curve to the right and will have one of two consequences. Assuming that both curves are step functions, it will either increase the amount consumed by 1 kw, or not increase it at all. If consumption is increased, the benefit of that consumption will be the market price, and the cost of supply (the added kwh) will be zero. The net benefit per kw is the market price. If consumption is not increased, some supply with a cost equal to the market price will be displaced by the new zerocost kwh. This leaves benefit unchanged and reduces cost, so again the net benefit per kw is the market price. If the supply and demand curves were smooth, the result would have been the same except there would have been a contribution from both increasing benefit and decreasing cost. Similarly the reduction in net benefit from extracting a kw from the system is also given by the market price. Thus, no matter how you compute it, the marginal value of power to the system sets the market price. Contrary to popular belief, auctions are not designed to determine who sells and who buys by comparing bids to the price determined by marginal-cost. Marginal cost pricing is not a goal, it is a byproduct. Auctions determine which set of trades is the most valuable possible (feasible) set of trades, and selects this set. Once they have been selected, the market price at each location is set to the marginal value or marginal cost of supply to the system at that location. 12 The market price, MP, determined in this way has two properties. First, at every location, the MP falls on the dividing line between bids that are accepted and those that are not. If some bids are partially accepted then MP is equal to their price. Second, given the first property, the difference between the total 12 The net benefit should be in $/h.. A kw, rather than a MW, is used to indicate that only a marginal change is being made. Technically one should use calculus, but this is of no practical significance.

23 CHAPTER 3-6 Day-Ahead Market Designs 49 amount paid by customers and the amount paid to suppliers is as small as possible. No other price would have these two properties. Conventions and Notation for Describing Auctions The use of supply- and demand-curve bids is common in DA auctions. Typically these curves are represented by either piece-wise linear functions (connect the dots with straight lines) or step functions. Typically these allow the bidder to specify about ten sloped lines or horizontal steps, but all that matters is that bidders can submit a fairly accurate approximation to their actual supply and demand curves. This will be assumed, and the details will be ignored. An auction market has three distinct sets of rules: one set for bidding, a second for bid acceptance and rejection, and a third for settlement. The description of each of the four DA markets is broken into these three categories THE FOUR DAY-AHEAD MARKETS Four subsequent pages give summaries of the economics of four types of dayahead markets. Each is a locational market and these locations may be either single buses or zones containing several buses. If zones are used, the transmission constraints will represent the market less accurately, and so a more conservative representation of constraints may be required. This affects only the details of the constraint specification and not the specification of the markets. Market 1: Pure Energy The bids in a pure energy market are sometimes called one-part bids because, for a given quantity of energy offered, the only a single price is specified. In some respects this is the simplest DA market. Participants do not search for trading partners and do not have to consider many prices in many locations. Each trader simply trades with the exchange at the traders location. The SO s job is simple because it ignores the unit commitment problem. The one difficulty, discussed in Chapter 3-7, is that suppliers cannot always bid their marginal cost. Market 2: Transmission The transmission auction is equally simple for the system operator but requires a complex pre-market step for market participants. Buyers and sellers must find each other and make provisional energy trades that depend on whether or