Reliability & Markets

In 2005, the Federal Energy Regulatory Commission (FERC) began discussing regulatory policy for reactive-power procurement and pricing in competitive electricity markets. This paper summarizes findings from a unique, interdisciplinary program of public-interest research that lays a formal foundation for evaluating aspects of FERC staff recommendations and offers early insights that should be useful in guiding policy implementation, specifically by:

• clarifying the consumers and economic characteristics of reactive power as a basis for creating incentives to appropriately price it,
• defining specific challenges in creating a competitive market for reactive power as well as new tools needed to help ensure such a market functions efficiently, and
• demonstrating the importance of accounting for the physical characteristics of the transmission network in planning for reactive power and avoiding the exercise of market power by suppliers.

The FERC report on reactive power clearly and succinctly lays out the issues and raises important questions about market power, contingent-claim versus real-time markets, the need for an optimal power flow that incorporates reactive power, etc. Unfortunately, the economic/engineering models so far available in the literature fail to represent the true economic optimum. This optimum involves maximization of the expected net benefits of electricity production, transportation, and use under the constraint of a full alternating-current (AC) power flow where the expected net benefit is defined as the sum of the probability-weighted economic outcomes for all contingencies, including line and generator failures. This is the correct way, in terms of economics, to determine optimal reliability, levels of investment, and operation parameters under alternative contingencies, as well as efficient and optimal production and prices for real and reactive power.

The purpose of this paper is to take a broad look at how markets should be organized, not only for reactive power but for real power and reliability, since these markets are fundamentally interdependent and essential for efficient and reliable delivery of electric power.

To accomplish this end, the paper opens with specification of an economic/engineering model of optimal investment and operation that is then simulated so that principles and goals for optimal market design can be established. The paper then examines issues of market power through both simulation and experimental economics. Finally, a variety of possible market designs are presented and evaluated in light of the conclusions drawn from the conceptual model, simulations, and experiments. The paper concludes with specific recommendations.

The conclusions drawn from the research are:

1. Although network reliability has long been identified as a public good, voltage and frequency have not been so identified. The modeling and simulations presented here using AC power flows demonstrate the public-good nature of voltage and reliability of lines. Note that, except under special circumstances, private incentives for provision of public goods are insufficient and efficient provision requires some central authority.
2. Both real power and reactive power are technically private goods since they are excludable and rival. Thus, unlike voltage and reliability, well-designed private markets are theoretically efficient if the public goods needed to support the system are optimally provided.
3. Simulation of optimal operation under different contingencies demonstrates that nodal reactive-power prices are almost always equal to zero if optimal investment in reactive-power sources (e.g., generators and reactive-power compensators) occurs throughout the system. Nonzero reactive-power prices, which optimally are found only during contingencies, such as when failures occur, remain relatively low because of the low cost of investment in a reactivepower supply.
4. The average or expected revenue derived from sales of reactive power at optimal real-time prices during rare contingencies is sufficient to provide incentives for optimal private investment in reactive-power capacity. However, this is a highly volatile and uncertain source of revenue that depends on such rare contingencies actually occurring. Thus, private investments in reactive capacity are inherently risky investments. Note that when reactive capacity is in short supply the willingness-to-pay, or social value, of reactive power is many orders of magnitude greater than investment based on cost per unit during contingencies, so it is optimal to make large investments in reactive-power capacity to prevent shortages even during rare contingencies.
5. However, both simulations and economics experiments show that opportunities for the exercise of market power by private suppliers of reactive power in real-time markets are plentiful in a network environment in which transmission of reactive power is limited to short distances, as established by Kirchoff's laws. Thus, even with sufficient reactive-power capacity, suppliers are likely to submit offers that will produce positive real-time reactive-power prices in noncontingency states for which optimal prices are zero.
6. Whereas virtually all demand for real power comes from private buyers (losses in the transmission network are small, on the order of 2-3%), demand for reactive power comes from two sources: (1) demand from private buyers to meet the needs of motors, arc welders, and other equipment that require a magnetic or electric field to operate and (2) the often greater demand from a central authority, acting in the public interest, for reactive power as an input to provision of a public good—voltage. We include in this category the often large amount of reactive power consumed by transmission lines that are operating far from their surge impedance loading. Thus, since voltage is essential for reliability, reactive power plays a special role in the security of the power system.
7. Optimal real-power prices show greater upside volatility than optimal reactive-power prices, principally because investment in generation is so expensive that it cannot optimally cover all contingencies. Optimal generation investment does conform to the conventional wisdom of covering the worst single contingency by meeting load after the loss of the "largest" generator in the simulation. However, peak prices during some contingencies are much higher for real power because optimal investment in capacity cannot efficiently cover load with the loss of multiple generators. Note that optimal investment will occur in a free market only when prices are not capped since the next-to-worst-case scenario for generator failure provides the incentive for investment in the surviving generator to provide optimal reliability through prices on the order of $10,000 per megawatt hour.
8. Optimal investment in lines in the simulation is sufficient so that thermal line constraints are never binding, even during contingencies. Thus, if a lossless DC optimal power flow were used there would never be nodal price differences that could be used for, for example, transmission fees. Using a full AC power flow does, however, show substantially different nodal prices for both real and reactive power, but only during contingencies. There can be no transmission fees from nodal price differences for real and reactive power unless prices are computed from a full AC optimal power flow. These prices are needed to provide both revenue and charges for reactive power consumed by lines (to maintain voltage) that will generate optimal incentives for line investment.

Based on these research conclusions, we recommend the following:

1. The first conclusion presented—that some central authority is needed to provide the public goods of reliability and voltage (as well as frequency)—implies that electric power does not lend itself to the degree of decentralized decision-making that is common in other markets. Thus, there must be some institution that has government-like authority to design, plan, and manage the system. This entity is referred to as the central authority because current independent system operators do not have authority for planning and design.
2. For the central authority to act in the public interest and be able to optimize the system, as well as provide necessary public goods, the central authority must possess a robust AC Optimal Power Flow (AC OPF) program that can resolve real- and reactive-power problems properly. The program should include unit commitment, ancillary services, and contingencies. A proper AC OPF is needed to give an accurate picture of the system for operations, as well as accurate price information for real power, reactive power, and transmission. Such prices are necessary as a basis for proper investment decisions in generation, reactive compensation, and transmission. A major research and development program should be undertaken to provide this capability.
3. The third conclusion states that reactive-power prices will be mostly zero with appropriate investment in reactive compensation. When a real-time market for a private commodity has financial transactions on rare occasions and, because markets are expensive to operate, natural economic forces will restructure the market to avoid transaction costs. The commodity used in these markets is called a contingent claim, which is a claim for services that can be made only if one or more specified events occur. Contingent-claim markets, which are appropriate for reactive power, operate well in advance of the contingencies that justify claims and are the normal replacement for real-time markets, which rarely have transactions. In a contingent-claim market, the central authority essentially rents major reactive-power sources from suppliers that submit the lowest-priced offers and instructs the suppliers in how to operate those sources in real time. For generators, the contingent-claim contract can provide fixed compensation for reductions in real-power output if reactive-power needs require such reductions.
4. Because the central authority responsible for reliability and operations needs reactive power on demand so that it can deal with contingencies and thereby assure reliability (conclusion 6) and because the substantial investment required to meet that demand must be assured in advance (conclusion 4), the reactive-power market must be run well in advance of any contingency to assure needed supply. Contingent-claim markets are appropriate for reactive power because they close well in advance of any claims made and, if run sufficiently far in advance, can provide a sure source of revenue, encouraging investment. Rather than obtaining revenue through unpredictable rare contingencies, sources that submit winning offers obtain steady revenue in the form of rent to compensate them for providing reactive power on demand. Note that, since this market is run far in advance, the central authority must project the amount of reactive power capacity needed for private buyers and to maintain voltage. Thus, based on the central authority's determination of how much reactive power will be needed and where (projected nodal demand), this market must be run locally to efficiently acquire reactive-power sources.
5. Market power is a serious problem for the supplying of reactive power (conclusion 5) in realtime markets, more critical than in the case of real power. The overall demand for reactive power comes in great part from the central authority responsible for system operations and providing reliability to meet public needs. Competitive prices will be assured in contingent-claim auctions only if some offered units are potentially excluded. We recommend that contingent-claim auctions for reactive power be run sufficiently far in advance to allow construction to occur (three to five years) so that existing suppliers are placed in competition with potential investors and new sources of reactive power, encouraging competitive prices.
6. Although real-time markets for real power are potentially feasible and can provide stable revenue for investment in generation through forward markets (as the Australian example suggests), capped real-power prices paid to generators, as is common in U.S. power markets, will not provide sufficient incentives for investment in generation to assure optimal reliability. Thus, we support measures to supplement generation investment if prices to generators are capped. Although a number of approaches are being tried to encourage investment in generation, it is not yet clear if any are successful or cost-effective.
7. The central authority responsible for reliability and operations must have legal authority to impose the stringent penalties necessary to enforce contracts purchased in contingent-claim markets.
8. To provide incentives for conservation of real- and reactive-power demand, large customers and marketers should pay real-time nodal prices for real and reactive power as derived from the system AC OPF. Contingent claim markets are appropriate for the supply side of the reactive power market, but real time nodal prices are appropriate for sales to marketers since they will then have incentives to install metering and then either pass on real-time prices or install automated controls on customer equipment in exchange for a lower fixed rate. This will also make distributed energy resources and load response much more economically viable.
9. Nodal prices from a DC power flow can provide incorrect price signals, including indicators for investment in lines. Proper incentives require that transmission fees be equal to the nodal price differences for real and reactive power derived from a full AC power flow and be applied to transmission of both real and reactive power. Transmission must also pay for reactive power consumed by lines. Note, however, that transmission fees typically may be near zero with optimal line investment but tend to be positive during contingencies. As in the case of reactivepower markets, real-time markets may be inappropriate. To assure efficiency and reliability, the central authority must plan and manage transmission.

Robert Thomas has shown, using simulations of experimental results, that the power flow on any line in an electric network is linearly proportional to the total system load when that system is optimally dispatched using accurate generator cost data. By comparison, when offers from generators obtained in a wholesale market that is not perfectly competitive are used to dispatch the system, that relationship between line flow and system load becomes nearly random. These simulations were conducted in a single-sided market environment, however, that is typical of most wholesale market regimes around the world. Here the central dispatcher (ISO, RTO, etc.) accumulates the demand from various buyers and satisfies that load with a least-cost purchase schedule, regardless of price, subject to all of the physical and reliability constraints imposed on the system. If buyers were also able to submit a schedule of bids that are related to price, does the same random relationship between line-flows and system load prevail?

This experimental analysis demonstrates that letting the customers participate fully in the market re-establishes the predictability of line flows as a function of system load. In all of these experiments there are no restrictions on permissible offering behavior by suppliers (e.g. no price caps, prohibitions on withholding capacity or automated mitigation procedures). Two alternative forms of demand side participation are considered: 1) a demand response program (DRP) where customers are alerted to high prices in the subsequent period and are paid a pre-specified amount for each kWh less than their benchmark level of usage for that period, and 2) a real time pricing program (RTP) where customers are given forecasts of prices for each period over the subsequent day and they then pay the actual period-by-period market clearing price. As a benchmark, these experiments with six suppliers and seventeen buyers are also repeated where customers pay an average constant price in all periods (FP); although in all cases sellers receive the market-clearing price in each period.

R-squares were greater, variances were smaller and the t-tests on regression coefficients were stronger on the relationship between line-flow and system load for RTP, as compared to the FP system that is commonly used in most electricity markets. DRP was usually somewhere in between. Not only does inducing active customer participation in the market through RTP lead to better system predictability, it also reduces price spikes and leads to greater overall economic efficiency in these markets. It is a winner on both economic and operational grounds.

We analyze welfare and distributional properties of a two-settlement system consisting of a spot market over a two-node network and a single energy forward contract. We formulate and analyze several models which simulate joint dispatch of energy and transmission resources coordinated by a system operator. The spot market is subject to network uncertainty, which we model as a random capacity derating of an important transmission line. Using a duoploy model, we show that even for small probabilities of congestion (derating), forward trading may be substantially reduced, and the market power mitigating effect of forward markets (as shown in Allaz and Vila 1993) may be nullified to a great extent. There is a spot transmission charge reflecting transportation costs from location of generation to a designated hub whose price is the underlying for the forward contract. This alleviates some of the incentive problems associated with the forward market in which spot-market trading is residual. We find that the reduction in forward trading is due to the segregation of the markets in the constrained state, and the absence of natural incentives for generators to commit to more aggressive behavior in the spot market (the "strategic substitutes" effect). In our analysis, we find that the standard assumption of "no-arbitrage" across forward and spot markets leads to very little contract coverage, even for the cast with no congestion. We present an alternative view of the market where limited intertemporal arbitrage enables temporal price discrimination by competing duopolists. In this framework, we assume that all of the demand shows up in the forward market (or that the market is cleared against an accurate forecast of the demand), and the forward price is determined using a "market clearing" condition.

System operators in the electricity industry are required to procure reserve capacity to deal with unanticipated outages, demand shocks, and transmission constraints. One traditional method of procuring reserves is through a separate capacity auction with two-part bids. We analyze an alternative scheme whereby reserves are procured through the energy market using only energy bids, and capacity payments are made based on a generator's implied opportunity cost. By using the revelation principle, we are able to derive the equilibrium bidding function in this market and show that generators have a clear incentive to understate their costs in order to capture higher capacity rents. We then show that in spite of making energy payments based on the marginally procured unit, the expected energy costs under our scheme are bounded by that of a disjoint auction. We then give a numerical example for a special case of uniform demand distributions.

This paper explores a mathematical method for detecting groups of generators in an electric power system that have the potential to benefit from exercising market power. Applications of this method include metrics for measuring or detecting the possibility of market power. This paper focuses on the properties of revenue and dispatch to bid sensitivity matrices, and develops methods of identifying load pockets from the sensitivity matrices, and how the matrices can provide metrics for market power.

An experimental structure is demonstrated that represents end-use customers in electricity markets who can substitute part of their usage between day and night. Individuals' demand relationships are represented by a two-step value function for each period that are disaggregated from observed market demand relationships. Demand varies between day and night and during heat waves. Three alternative demandside market structures are evaluated: 1) customers pay the same fixed price (FP) in all periods ñ the base case, 2) a demand response feature (DRP) is added in periods of supply shortages, wherein buyers receive a prespecified credit for reduced purchases, and 3) a real time pricing (RTP) case where prices are forecast for the upcoming day/night pair, then buyers select their quantity purchases sequentially and are charged the actual marketclearing prices.

Initial experiments were conducted with active demand-participants, but with a predetermined typical 'hockey-stick' supply structure that was varied randomly, over eleven day-night pairs that included heat wave and supply shortages. The RTP structure resulted in the greatest market efficiency, despite the more difficult cognitive problem it poses for buyers. Furthermore, a preference poll comparing DRP and RTP was conducted after each trial; and while 64% of the participants said they preferred DRP before RTP experiments, 76% selected the RTP structure afterwards, a statistically significant reversal of preferences.

Policymakers have increasingly recognized the structural impediments to effective competition in electricity markets, which has resulted in a renewed emphasis on the need for careful market design and market monitoring in wholesale and retail electricity markets. In this study, we review the market monitoring activities of four Independent System Operators in the United States, focusing on such topics as the organization of an independent market monitoring unit (MMU), the role and value of external market monitors, performance metrics and indices to aid in market analysis, issues associated with access to confidential market data, and market mitigation and investigation authority. There is consensus across the four ISOs that market monitoring must be organizationally independent from market participants and that ISOs should have authority to apply some degree of corrective actions on the market, though scope and implementation differ across the ISOs. Likewise, current practices regarding access to confidential market data by state energy regulators varies somewhat by ISO. Drawing on our interviews and research, we present five examples that illustrate the impact and potential contribution of ISO market monitoring activities to enhance functioning of wholesale electricity markets. We also discuss several key policy and implementation issues that Western state policymakers and regulators should consider as market monitoring activities evolve in the West.

Congestion management is an important component of electricity supply that is, in the U.S., typically achieved by operation of a transmission rights market, often purely financial. In principle, financial transmission rights serve market participants attempting to hedge against uncertain, and often sizable, congestion charges. In addition, effective congestion management can make primary energy markets more efficient and can identify areas where transmission investment is needed. The Wholesale Power Market Platform white paper circulated by the Federal Energy Regulatory Commission (FERC) in April 2003 proposes the establishment of receipt point-to-delivery point (PTP) obligations called Firm Transmission Rights (FTRs), if locational pricing is employed in the energy markets. These rights would allow the holder either to collect or pay the congestion rent between the specified point of injection (POI) and point of withdrawal (POW) for each right. This proposal system is similar to the Transmission Congestion Contract (TCC) system employed by the New York Independent System Operator (NYISO), which has been operating since the spring of 2000. NYISO TCCs are financial derivatives that can be freely traded both by market participants and by speculators.

The physical nature of electricity generation and delivery creates special problems for the design of efficient markets, notably the need to manage delivery in real time and the volatile congestion and associated costs that result. Proposals for the operation of the deregulated electricity industry tend towards one of two paradigms: centralized and decentralized. Transmission congestion management can be implemented in the more centralized point-to-point approach, as in New York state, where derivative transmission congestion contracts (TCCs) are traded, or in the more decentralized flowgate-based approach. While it is widely accepted that theoretically TCCs have attractive properties as hedging instruments against congestion cost uncertainty, whether efficient markets for them can be established in practice has been questioned. Based on an empirical analysis of publicly available data from years 2000 and 2001, it appears that New York TCCs provided market participants with a potentially effective hedge against volatile congestion rents. However, the prices paid for TCCs systematically diverged from the resulting congestion rents for distant locations and at high prices. The price paid for the hedge not being in line with the congestion rents, i.e. unreasonably high risk premiums are being paid, suggests an inefficient market. The low liquidity of TCC markets and the deviation of TCC feasibility requirements from actual energy flows are possible explanations.

It is well known that given a network that can become constrained on voltage or real power flows, reserves must also be spatially located in order to handle all credible contingencies. However, to date, there is no credible science-based method for assigning and pricing reserves in this way. Presented in this paper is a new scheduling algorithm incorporating constraints imposed by grid security considerations, which include one base case (intact system) and a list of possible contingencies (line-out, unit-lost, and load-growth) of the system. By following a cost-minimizing co-optimization procedure, both power and reserve are allocated spatially for the combined energy and reserve markets. With the Lagrange multipliers (dual variables) obtained, the scheduling algorithm also reveals the locational shadow prices for the reserve and energy requirements. Unlike other pricing and scheduling methods in use, which are usually ad-hoc and are based on engineering judgment and experience, this proposed formulation is likely to perform better in restructured markets when market power is a potential problem. An illustrative example of a modified IEEE 30-bus system is used to introduce concepts and present results.

The underlying structure of why and how consumers value reliability of electric service is explored, together with the technological options and cost characteristics for the provision of reliability and the conditions under which market mechanisms can be used to match these values and costs efficiently. This analysis shows that the level of reliability of electricity provided through a network is a public good within a neighborhood, and unless planned demand reductions by customers have the identical negative value as an unexpected service interruption, market mechanisms will not reveal the true value of reliability. A public agency must determine that value and enforce the reliability criteria. Furthermore, in order to get an efficient level of demand response by customers in periods of system stress, they must see real time energy prices plus they must be paid an amount equal to the suppliers' cost of adding reliability to the system, if that amount is not included in real time prices.

An illustration is provided of how VARs might be scheduled and priced in contributing to system reliability, and a co-optimization procedure is required to determine energy and reserves simultaneously, similar to the method proposed by Chen, Thorp, Thomas, and Mount [1] for locational reserves. The optimization can be decomposed into a two step process—first, both required capacity and energy are selected based upon suppliers' offers over both dimensions through the minimization of expected costs over the list of contingencies necessary to satisfy the reliability criteria. This first step commits the reserves, but energy supplies are allocated in real time based upon the previous offer prices but the actual realized state of the electric system. This procedure which satisfies physical realities has a natural parallel in financial markets that have a forward option market with a strike price, followed by real time market clearing.

It is well known that given a network that can become constrained on voltage or real power flows, reserves must also be spatially located in order to handle all credible contingencies. However, to date, there is no credible science-based method for assigning and pricing reserves in this way. Presented in this work is a new scheduling algorithm incorporating constraints imposed by grid security considerations, which include one base case (intact system) and a list of possible contingencies (line-out, unit-lost, and load-growth) of the system. By following a cost-minimizing co-optimization procedure, both power and reserve are allocated spatially for the combined energy and reserve markets. With the Lagrange multipliers (dual variables) obtained, the scheduling algorithm also reveals the locational shadow prices for the reserve and energy requirements. Unlike other pricing and scheduling methods for reserves in use, which are usually ad-hoc and are based on engineering judgment and experience, this proposed formulation is likely to perform better in restructured markets when market power is a potential problem. The modified IEEE 30-bus system is used to introduce concepts and present results.

We analyze welfare properties of two-settlement systems for electricity in the presence of network uncertainty and market power. We formulate and analyze several models which simulate the different market designs adopted or proposed for many electricity markets around the world. In particular, we examine the extent to which a two-settlement system with zonal aggregation in the forward market facilitates forward trading, as well as the welfare and distributional implications of having such zonal aggregation in the presence of network uncertainty. Using a duopoly model over simple two- and three-node networks, we show that for even small probabilities of congestion, forward trading may be substantially reduced, and the market power mitigating effect of forward markets (as shown in Allaz and Vila, 1993) may be nullified to a great extent. We find that the imposition of a delivery requirement on the forward contract in the form of a spot transmission charge alleviates some of the incentive problems associated with zonal aggregation. Even with the imposition of the spot transmission charge, we find that some reduction in forward trading persists due to the segregation of the markets in the constrained state, and the absence of natural incentives for generators to commit to more aggressive behavior in the spot market. In our analysis, we find that the standard assumption of 'no-arbitrage' across forward and spot markets leads to very little contract coverage even in the no congestion case. We provide an alternative view of the market where we assume that all of the demand shows up in the forward market, and is aggregated to determine the forward price using a 'market clearing' condition.

The objective of this report is to investigate methodologies for zone creation suitable for zonal pricing by the California ISO. It is recognized that in meshed networks, zones are only approximations to individual node pricing. The objective, however, is to create zones that closely approximate the "correct" nodal prices under most conditions and where the majority of the "commercially significant" value of the locational pricing is captured. The report starts with a review of congestion concepts, then proceeds with a review of nodal and zonal pricing, it then reviews the CAISO's own criteria for zone partitioning, it then evaluates the notion of "nodal price patterns," and finally, using the notion of nodal price pattern it defines a methodology for zone creation and partitioning and finally demonstrates its used by means of an example. It concludes with some suggestions and recommendations.

We examine efficiency properties and incentive compatibility of alternative auction formats that an electricity network system operator may use for the procurement of ancillary services required for real-time operations. We model the procurement auction as a hierarchical multiproduct auction and study several designs such as a uniform price auction minimizing social cost, a uniform price auction minimizing the system operator's cost and a pay as bid auction. We take into account that rational bidders will respond to any market design so as to maximize their expected benefit from participating in that market. Under the assumptions of our model, we show that the uniform price auction minimizing social cost is the only one that guarantees productive efficiency. We also find that expected revenue (payment in our case) equivalence between pay as bid and uniform price auction does not extend to the hierarchical products case and the ranking of these auction is ambiguous and depends on the data. For the procurement auction minimizing the system operator's cost we show that misrepresentation of capability may result in capacity shortages if there are capacity constraints. For the case where only higher capability resources are constrained this will result in random price spikes decreasing in frequency with the price cap (this is the amount paid to capacity in demand states with shortages). When lower type resources are capacity constrained as well, price spikes will be seen for both type of resources. Artificial shortages result in reduced reliability in real-time operations.

We examine the design of bid selection protocols and settlement rules in ancillary service markets. Such markets are typically operated by an independent system operator (ISO) for competitive procurement of reserves that are needed to ensure the secure operation of a competitive electric power system. Reserve types are characterized in terms of response time and they are downward substitutable (faster responding reserves can replace slower ones). We explore how this substitutability is accounted for in alternative market protocols and we analyze the efficiency, distributional aspects and incentive compatibility of such protocols.

In deregulated electricity markets such as California's, the introduction of competition in the generation sector has reduced the role of vertically integrated utilities in maintaining system reliability in their own service areas. Consequently, the Independent System Operator (ISO) is now charged with ensuring system reliability by purchasing enough ancillary services (AS) through competitive markets. In order to determine how such AS would be priced in a competitive environment, we model wholesale markets for both electricity and AS using the market-equilibrium approach. We determine the equilibrium wholesale prices for both electricity and AS, as well as the optimal quantities traded in the markets by generators, retailers, and the ISO. Specifically, we seek to obtain a market-based pricing methodology for both electricity and AS that captures the physical and financial links between the markets for these two products. We show that the wholesale price of AS depends upon the opportunity costs of foregone electricity sales for the generator. Empirical analysis using California market data provides preliminary corroboration of our theoretical results.

To facilitate research in restructured electricity markets, the Consortium for Electric Reliability Technology Solutions (CERTS) has been collecting and maintaining a database of California electricity market information that covers the entire period that the California Power Exchange (CalPX) and California Independent System Operator (CAISO) have been operating, that is, since 1 April 1998. All data currently in place are in the public domain. To the best of our knowledge, this is the most complete data set freely available with up-to-date data from each California market.1 In June 2000, with funding from the Department of Energy (DOE), Berkeley Lab initiated the process of building a database on a more robust and readily accessible platform, that is, an Oracle 8i database running on Solaris with web-enabled query capability. The maintenance of such a database enhances the ability of CERTS to conduct research on reliability of power systems in restructured markets, electricity market operation, and efficient design of competitive electricity markets. This document describes the database as it currently exists and plans for its improvement.

The settlement rule used to determine payments in electricity market auctions can have a profound impact on the amount paid for a particular service. This paper evaluates two auction settlement rules: the uniform price auction in which the last offer accepted determines the price paid to all participants and the discriminatory price auction in which each participant is paid the amount bid by that party. Using an electricity market simulation tool to model the markets for energy and various ancillary services in a large control area, it is demonstrated that under conditions of market power substantial revenues with commensurately high profits can be commanded under a uniform price auction. By employing a discriminatory price rule, much of the impact of market power can be ameliorated. In addition, a discriminatory price auction, by virtue of requiring each bidder to explicitly state their desired revenue rate, provides greater visibility of attempts to make use of strategic pricing and market power. By discouraging the use of market power through greater price visibility, discriminatory price auctions also have the potential to reduce instances of strategic capacity withholding, which in turn, should enhance overall system reliability.

On 31 March 1998, restructuring of the California electricity industry brought the forces of the free market into a sector of the economy that had long been regulated. While the new market structure is similar to that of many other restructured markets in several respects, the competitive procurement of ancillary services by the California Independent System Operator (CAISO) is an almost unique feature. These markets are shown to exhibit price volatility that has resulted in high procurement costs for California electricity consumers. An analysis of the amount of generation offered in the AS markets relative to the known purchase requirement of CAISO shows that withholding of capacity in certain hours has probably occurred and this provides evidence that the generators have exercised market power. The preferred response to these problems on the part of the CAISO Board, with the approval of the Federal Energy Regulatory Commission (FERC), has been the imposition of a fixed price cap, the level of which has been adjusted between $250/MW and $750/MW. Mechanisms used in other restructured electricity markets could have been considered, such as threats of regulatory action and use of hedge contracts.

The original objective of this work was to use experimental economic and other methods to simulate the performance of both current and proposed market (including ancillary services) designs and to explore, test, and demonstrate both theoretical and experimental economic approaches that simulate market performance in ways that accurately reflect the physical capabilities and limitations of the electric power system and the risks inherent in linking it to volatile markets. We have explored the rules for the PJM market, the New England market and have devised a market of our own that has the property that, at least in terms of simulation results, mitigates the non-market-power price spike behavior seen in real markets.

The strong relationship between a properly designed ancillary services market and price spikes emanating from other than market power behavior is an important discovery produced as a result of this work. The fact that price spike behavior in markets such as New England may be a result of a poorly designed ancillary services market is significant.

This report first discusses the unique experimental platform POWERWEB and the LEEDR lab, and their roles in this project. PJM and New England market data were analyzed to determine their efficiency as well to try and quantify the effects of similarities and differences between market rules. Finally, experiments were performed to try and determine if market characteristics were indeed captured. We are especially interested in ancillary service market design. An important result is that we are able to design autonomous agents to operate the markets, where these agents capture essential market characteristics. This means we can use these agents to examine the operation of systems.

In competitive electricity markets, the costs for each ancillary service should be charged to those who cause the costs to be incurred with charges based on the factors that contribute to these costs. For example, the amount of generating capacity assigned to the regulation service is a function of the short-term volatility of system load. Therefore, the charges for regulation should be related to the volatility of each load, not to its average demand.

This report discusses the economic efficiency and equity benefits of assessing charges on the basis of customer-specific costs (rather than the traditional billing determinants, MWh or MW), focusing on two key real-power ancillary services, regulation and load following. We determine the extent to which individual customers and groups of customers contribute to the system's generation requirements for these two services. In particular, we analyze load data to determine whether some customers account for shares of these two services that differ substantially from their shares of total electricity consumption.

We defined and applied metrics for regulation and load following. For regulation, we chose the standard deviation (MW) of the thirty 2-minute values in each hour. For load following (MW), we chose the difference between the maximum and minimum values of the 30-minute rolling-average load during each hour.

We also developed and applied methods to allocate these system-level metrics to individual customers and to groups of customers. The regulation allocation method uses a trigonometric relationship to correlate an individual customer's regulation burden with the total burden. The load-following allocation method calculates each customer's share of the total requirement on the basis of its coincident load-following requirement.

Application of these allocation methods shows that charging customers for these ancillary services on the basis of average loads can be inequitable. For one control area, a few large industrial customers account for 34% of system load, compared with 93% of the regulation and 58% of the load-following requirements (Fig. S-1). These customers disproportionately use these services but, in general, are not paying their fair share under typical utility tariffs. The subsidies inherent in today's ancillary-service pricing methods cannot, and should not, be sustained. Indeed, industrial customers with near-time-invariant loads, such as aluminum smelters and paper mills, will justifiably claim they require none of these services and, therefore, should not have to pay for them.

The deregulated electricity markets of Norway, Spain, and California all have direct retail access and wholesale electricity markets. All three have also implemented elements of marginal pricing in their transmission tariffs. Transmission losses as well as congestion costs are key components of system marginal costs and these provide short-term signals to the grid. Fixed costs are covered by an energy component in California, while Norway and Spain have implemented both an energy component and a demand component to recover fixed costs. None of the three systems provide strong incentives in the "fixed cost element" of the transmission tariffs. To some extent the California access charge is time differentiated, and each area of the three large utility distribution companies has somewhat different access charges. However, the signals provided by this charge are limited. In Norway, the tariff elements that cover fixed costs are supposed to be neutral. In Spain, fixed costs are recovered through both an energy element and a demand element. In California, each grid owner suggests future investments, and is responsible of proposing ways of financing them, and firm transmission rights have been introduced as means to promote long-term signals for investments in constraint-reducing investments.

California has one of the few pioneering restructured electricity markets that has implemented open market purchases of most of its ancillary service, that is, reliability requirements. The performance of these markets has been characterized by serious deficiencies, especially during the summer of 1998, when generators were able to exercise extreme market power and force prices to hastily imposed ceilings. While electricity generation and provision of ancillary services are alternative products of generators, the prices of electricity and ancillary services have been weakly correlated in many cases. For example, the correlation coefficient between the PX day-ahead and regulation prices is only 0.12 in the first year of operation. Also, the lower quality services were often more expensive than the higher quality ones.