Load and Revenue Analysis and Forecasting for Restructured Electricity Markets B. F. ROBERTS Presented to the Electric Utility Forecasters Forum Santa Fe, New Mexico October 24, 1996 I. Introduction The rapid pace of electric industry restructuring has quickly rendered the load and revenue forecasts (projected beyond 2-3 years) of most utilities irrelevant for planning future operations. Moreover, many of the quantitative models used to help analysts reason about the future are now obsolete. This is because the functional unbundling of electric services (and the associated vertical disaggregation, divestiture, and reorganization of utility operations), deregulation of electricity commodity markets, and re-regulation of transmission and distribution will profoundly change the way electricity markets and market participants will operate. New electricity market models consistent with new corporate organizations and market structures are required to give management relevant perspective for developing business strategies for operating in uncharted and unobserved new markets. The challenge for utility analytic/forecasting support staff is to develop useful quantitative electricity market models without first observing the new markets in operation. This challenge can be met by combining information about the emerging structures of utility operations and new electricity markets with economic theory and existing electricity market data in the design of new simulation models that can be progressively refined from simulation experiments and future market observations. The purpose of this paper is to outline a new quantitative model development strategy for the technical staffs of electric utilities and other participants in restructured electricity markets. First, a description of the emerging structure of the electricity markets will set the foundation for defining management's new analytic support requirements, and for specifying the logical framework for the new electricity markets analysis and forecasting models. Second, the new support requirements for utility management will be outlined. Third, the logical structure of the new Regional Restructured Electricity Markets Simulator (RREMTM) will be presented. Fourth, the data sources and model implementation procedures will be discussed. Finally, some RREMS results will be offered. II. The Emerging Structure of Electricity Markets The process of restructuring electricity markets is in various stages of deliberation and implementation across the country. California has adopted legislation that begins the phase-in of direct access on January 1, 1998. Some states have direct access experiments underway. Most other states now have undertaken some form of industry restructuring proceedings. Federal legislation is also being debated that would force a time-table for direct access implementation on all states. With so much momentum for industry restructuring, it is prudent for utility forecasters to begin factoring new perspective into their analyses and forecasts. While final design of the new markets cannot be known in detail until all states and the federal government have adopted legislation, the basic structural features of the new markets are becoming clear enough to begin factoring them into our reasoning about electricity pricing, generation, loads, revenues, and the determination of market shares. These basic features combined with economic theory form the foundation for developing a quantitative system to begin simulating, analyzing, and forecasting the new electricity markets. Unbundling Electric Service The primary objective of the move to restructure the electric services industry is to achieve retail customer direct access to power producers.1 To accomplish direct access power purchasing, unbundling of the electricity commodity (generation) from the wires components of electric service (transmission and distribution/customer services) is necessary. Identification and separate pricing of the components of electric service is essential to allow customers to choose their preferred power producer. Unbundling is also necessary to allow deregulation of the commodity component and regulation of the wires component to coexist. Direct access also requires that the vertically integrated structure of electric utilities be modified, particularly through divestiture of generating plants. Re-regulated Transmission, Distribution, and Customer Services Transmission The use of transmission assets will be under the control of regional Independent System Operators (ISO's), although ownership will continue to be held by current owners. The ISO's operations and transmission pricing will be regulated by the FERC. While the FERC has required cost-based tariffed rates to be filed, there is substantial discussion about the possible use of nodal congestion prices that would vary with grid loading. Maintenance of transmission lines will continue to be the responsibility of the owners. Distribution and Customer Services Distribution and Customer Services will continue to be provided by regulated monopolies within defined geographic boundaries -- Local Distribution Companies (LDC's). The LDC's will continue to be protected from competition for distribution services. They will retain the obligation to provide distribution services and to provide bundled service to customers in their franchised service territories who elect it. The LDC's will offer competitive power supply options for their customers, but will be required to deliver the power of other producers to customers on a non-discriminatory basis at tariffed distribution service rates. The LDC's will also perform meter reading, accounting, billing, and collection functions (although these functions could eventually also become competitive). The LDC's will be regulated by state regulatory agencies and offer tariffed distribution rates determined by performance-based price cap formulas. To reduce the local area market dominance or "market power" of existing utilities, a substantial portion of the generating assets of these firms will be: assigned to unregulated business units, spun off to independent new unregulated generating companies, or sold to other unregulated generating companies. Some generating capacity may continue to be held by the regulated LDC's to supply customers that elect to continue bundled service from the LDC's. Unregulated Cash, Forward, and Future/Options Markets for Generation The generation or commodity business will be deregulated. Buyers and sellers will be free to conduct transactions in cash (spot), forward, and futures/options markets for electricity. These transactions will not be constrained by the geographic boundaries of LDC's (and/or the historic boundaries of previously vertically integrated utilities). Similarly, these transactions will not be constrained by the former distinction between wholesale and retail commerce. Direct access virtually eliminates the wholesale/retail distinction. The owners of generating plants will have access to broad new markets. They will be free to sell power to: retail customers, LDC utilities, aggregators, marketers, and spot markets, in any market area that can be physically reached through the transmission grid. However, they will face competition from virtually all producers that can physically reach the same buyers. Buyers will be free to contract for power supply from one or multiple sellers (generators, LDC utilities, aggregators, and marketers). For example, a retail buyer may enter a multi-year contract for base load power from one seller and purchase above base load requirements from another and/or make daily purchases when needed from the spot market. Sellers will offer differing forward contract price configurations to buyers across the various markets to achieve their marketing objectives recognizing: locational differences in demand and production costs, transmission losses, transmission charges, and transmission constraints; and the pricing behavior of other sellers. Sellers' offers will differ in terms of: expected delivered price levels, price volatility, dates of delivery, reliability of delivery, and duration of contract consistent with the buyer's preferences concerning price risk. The delivered price will include: the seller's commodity price, transmission/ISO and distribution/customer service costs, allocated competitive transition costs (stranded asset charges), and any public benefits charges. In contrast to the stable tariffed electricity prices that have been the rule in regulated markets, volatile prices will be a major characteristic of the restructured markets. The commodity component of electricity prices will be determined by supply and demand which will shift with the hour of the day, day of the week, season of the year, and with a multitude of extreme conditions, such as abnormal weather and power outages. Electricity spot price data will be available on a virtual real time hourly or sub-hourly basis for some specific locations. The spot price is the price that clears the local electricity market and is a measure of the market value of electricity for a specific time and location. Spot prices will vary across locations because of differing supply/demand conditions. Therefore, spot price measured at one location may not be a good proxy for the market clearing price at another location. Facing volatile prices, risk-averse buyers and sellers will seek to negotiate fixed price forward contracts and/or futures/options contracts to lock in prices for future delivery dates. Price forecasting for negotiating forward contracts will be a important task for profitable operations in the new markets. Sellers can be expected to accept fixed price long term contracts at prices lower than their forecast of spot prices to reduce their market risk. Buyers can be expected to accept fixed price long-term contracts at prices higher than their forecasts of spot prices to reduce their market risk. Since the electricity prices for forward contracts will be private, buyers and sellers will need to draw on a variety of other information sources to develop forward price forecasts. The electricity futures/options markets may provide useful information about forward prices. Futures prices represent the market participants' forecasts of what future spot prices will be. An essential feature of electricity futures contracts (for delivery at a specific location) is that as the delivery date of the futures contract approaches, the futures contract price and the spot price will converge. While the futures contract prices provide forecasts of forward spot prices, there is no assurance that the forecast will be correct, although the forecast error can be expected to diminish, the shorter the time remaining to futures contract maturity. Electricity futures contracts are currently available for power deliveries at only two locations, the California-Oregon Border (COB) and Palo Verde. Contracts for other locations will likely be developed as industry restructuring broadens. The currently available futures price data may provide imperfect information for estimating forward contracts prices because of locational and product differences.2 As mentioned above, locational differences in spot price may make the spot price at one location a poor proxy for the market clearing price for another location. Similarly, the price for a futures contract for delivery at one location may not be a good proxy for the price of a futures contract at another location. Product differences could also occur because the product specification for a tradeable futures contract may not adequately reflect the product (primarily in terms of delivery schedule) specifications for a forward contract. III. Management Requirements for Analysis and Forecasting in the New Electricity Markets Management's requirements for analysis and forecasting support in the new electricity markets will multiply in terms of: frequency of forecasts, the breadth of scope, and the complexity of needed details. Short Run Load Forecasts (Daily) All power producers that serve retail customers, as well as marketers, aggregators, and LDC's that buy power on behalf of their customers will be required to forecast hourly loads one day ahead for submission to the ISO for scheduling dispatch. The ISO will charge the costs of scheduling errors to the parties that submitted faulty forecasts.3 This places clear financial incentives on all producers, marketers, aggregators, and LDC's to develop accurate forecasting methods prior to any participation in the new electricity markets. Short run hourly forecasts are also currently required by utilities in scheduling dispatch, maintenance, and in making other immediate operating decisions. Similar forecasts will now be required across all markets in which a supplier competes, reflecting differing weather conditions, volatile prices, and competition for customers in each market. Frequent adjustments for lost or gained customers will be required. Intermediate and Long Term Forecasts Consistent with the unbundling of electric service, unbundled forecasts for each component of electric service will be required. The forecast requirements for the electricity commodity business (generation) are different from those for the transmission functions which are also different from those for the distribution function. Generation Under the new market structure, electricity pricing, production, and consumption will be denominated by hours and location. To support marketing and investment decisions, price and demand forecasts will be required by both hour and location for years ahead, across multiple regional markets. Also, the competitive structure of the regional markets implies that pricing decisions and demand forecasts will be inextricably connected. In other words, a detailed market-specific hourly price plan must be developed as a part of load and revenue forecast development. Each seller's price plan must also take into account expectations about the pricing strategies of rivals in each market served by the seller. The load and revenue forecasts must estimate market shares in each market, consistent with the pricing behavior of all viable competitors. Distribution The forecasting requirements for LDC's will not be as dramatically changed after restructuring, except that electricity commodity prices will no longer be set by the LDC. The primary needs of LDC's are annual customer growth and peak demand per customer by location (within the franchised service territory). To accurately forecast LDC service area electricity demand, good estimates of the electricity commodity prices, determined by the competitors serving power to LDC customers will be required. Hourly price elasticities will be essential to capture time-of-use shifts in consumption in response to competitive pricing. Transmission Scheduling of grid loading will be the responsibility of the ISO. However, the owners of transmission assets will need to forecast the revenue from their transmission assets. This requires estimates of interregional power flows and transmission charges. One of the biggest unresolved issues in the restructuring deliberations is the pricing of transmission services during congested periods and the effects of pricing policy on system expansion. How this is resolved will impact the revenues earned from existing transmission assets and the investment plans of utilities. As a starting point, it is reasonable to utilize the FERC cost-based transmission rate methodology. At this point, utility management's primary forecast requirement is for estimates of power transportation over the utility's transmission lines and the associated revenue. IV. The Design of the Regional Restructured Electricity Markets Simulator (RREMS) Overview As described above, electric services industry restructuring will substantially change the way power production and delivery is organized, how buyers and sellers conduct these transactions, and how electric prices are determined. The new electricity markets will be more complex and more volatile, implying more comprehensive, detailed, and frequent information requirements for market participants. The challenges for electricity market analysts is to develop useful quantitative market simulation systems for analysis and forecasting without first observing the new markets in operation. The strategy proposed here, is to combine information about the characteristics of emerging market structures with economic theory and some of the existing electricity market data in the design of new simulation models that can be progressively refined from simulation experiments and future market observation. An overview of ESC's Regional Restructured Electricity Markets Simulator (RREMS) gives an example of how this strategy can be implemented. A description of the logical foundation of RREMS is presented in this section. The next section discusses data and implementation issues. The RREMS Model is a generic representation of multiple regional retail electricity markets that is implemented and calibrated uniquely for each application. RREMS was developed as an hourly system to accommodate both short term and long term forecasting requirements and to capture the implications of real-time pricing. The model is modular to facilitate substitution of various model component specifications for simulating and analyzing alternative scenarios. RREMS is integrated with the UQARTM utility database system for easy retrieval of utility operating statistics4 and regional economic5 and weather6 data. The RREMS Model also interfaces with the TRADELECTM Model of the national wholesale electricity market as a source for spot prices and national interregional power flows.7 Regional electricity markets are defined (for RREMS) by the LDC service area geographic boundaries. This provides easy reference to buyer groups and facilitates accounting for customer access tariffs on delivered power. Regional electricity markets are partitioned by customer class, where a class is any homogeneous group of customers. The set of regions included in any RREMS implementation are those regions within reasonable electricity transmission cost reach of the service territory targeted for analysis. This set is determined by ranking of relative production and transmission costs across nearby LDC service areas. The RREMS Model is organized into five interconnected modules: Economic/Demographic, Demand, Pricing, Customer Choice/Market Shares, and Cost/Supply, as shown on Figure 1. The Economic/Demographic and Demand Modules have conventional structures, not unique to RREMS. The Pricing and Customer Choice/Market Shares Modules are the distinguishing features of RREMS. Figure 1 Economic/Demographic The economic/demographic module quantifies multi-regional economic development and the growth of the customer base. The structure of this module is consistent with conventional methodology for utility service area models that have been successfully developed and used by utilities over the past two decades. Included for each region is a cohort survival submodel augmented with an interregional migration submodel to forecast population growth and household formation (residential customers). The economic submodel develops multi-regional construction activity, employment, income, and the growth of the commercial, industrial, and other customer bases. The Economic/Demographic Module has input linkage to the Cost/Supply Module (to transmit the effects of generating plant construction and operation) and the Pricing Module (to transmit the effects of electricity costs on location decisions and local economic activity). It has output linkage to the Demand Module (to transmit customer base, income, and employment). Demand The Demand Module quantifies the hourly electricity use per customer and total electricity use in each class and market. The equations are conventional econometric demand functions that predict electricity use as a function of: the delivered price of electricity; hourly, daily (weekday, weekend, and holiday) and seasonal load patterns; extreme weather conditions; economic factors (e.g., income and employment); and DSM adjustments to demand. The delivered price is the minimum price among all competitive offers and includes: the electricity commodity price, the transmission/ISO charge, the distribution/customer service charge, the competitive transition charge, and any public benefits charge. The demand equations exhibit price elasticities that vary by time of day, season, etc. The Demand Module has input linkage from the Economic/Demographic Module (to transmit customer base, income, and employment) and the Pricing Module (to transmit competitive prices). It has output linkage to the Customer Choice/Market Shares Module (to transmit demand by market and class). Pricing The Pricing Module quantifies the elements of the delivered price and contract offers of all sellers, for each customer class, in each market. The Module includes: Performance Based Ratemaking price cap equations for transmission/ISO charges (regulated by the FERC) and distribution/customer service charges (regulated by state PUC's); tariffed competitive transition charges and public benefit charges (set by state PUC's); and electricity commodity prices and related contract terms. Determination of the electricity commodity price offers is the core of the Module. The Pricing Module specifies the commodity pricing/offer behavior of electricity sellers in pursuing their marketing objectives (across all feasible markets and customer classes), while recognizing that their market rivals are also pursuing their own objectives. Essentially, the Module is designed on the assumption that sellers will exploit, to whatever extent they can, any locational cost and market information advantages that they might have. Seller pricing behavior is represented by a set of price reaction functions that capture interdependence among: locational production costs, transmission losses, and rival sellers' prices (including spot prices). Seller pricing strategies also include bids to sell excess supply to the spot market. Since little is known at this time about the rivalous pricing behavior of electricity sellers, a variety of interchangeable theoretical price reaction functions are being developed to support a broad range of scenario analyses. The Pricing Module has input linkage from the Cost/Supply Module (to transmit production cost information). The Pricing Module utilizes production costs combined with theoretical specifications of rational pricing strategies to forecast forward electricity prices for each power producer. This is a key feature of the RREMS Model. The Pricing Module has output linkage to the Economic/Demographic, Demand, and Customer Choice/Market Shares Modules (to transmit seller prices). Customer Choice/Market Shares The Customer Choice/Market Shares Module decomposes the aggregate electricity demand functions (by class in each regional market) into sets of market share demand functions facing each individual seller (for each class in each market). The market share demand functions reflect customer switching among sellers as a function of relative seller prices (including spot prices) and other contract terms. The market shares demand functions are kinked and display changing elasticity with respect to an individual sellers' price, as that seller's price varies relative to other sellers' prices as shown on Figure 2. The line A-C-D represents the demand curve for one market (a particular customer class in a particular market area). Let PL represent the lowest price among the sellers in the market. If seller K's price is the lowest PL, then his sales would be Q the total demand in the market. Should he lower his price further, he would move along the market demand curve increasing the amount demanded in the market. If K sets his price above PL, say at PH, then he will move along the upper elastic portion of his demand curve to QL and only capture the market share (QL/Q). He may also lose (or gain) market share because of changes in the position and shape of the demand curve segment B-C due to price changes by other rival sellers. Each seller, facing differing demand curves in each of the markets he serves, will attempt to set prices in each of the markets that maximize his profits, given the pricing behavior of his rival sellers. This description assumes instant market reaction to price changes which may not actually occur because of fixed contracts and other factors that would slow adjustments. This diagram simply demonstrates the direction of adjustments that can be expected in the new electricity markets. Figure 2 This discussion also assumed that all prices and price changes would be instantly known. As was discussed above, forward price information will not be readily available. Price forecasts, however, are produced by the Pricing Module. The Customer Choice/Market Shares Module determines the market share and total hourly end-user load and revenue of each electricity producer, given estimated prices and other variables. Total end-user load plus losses imply production levels which are linked to the Cost/Supply Module. Cost/Supply The Cost/Supply Module includes marginal cost (supply), average variable cost, and average total cost curves for each producer operating in the set of markets under study. These curves serve to constrain the pricing strategies and load volumes that are feasible for each producer. The cost curves are constructed from the plant operating statistics in the UQAR System. In addition, efficiency indices are calculated for key plant performance measures: heat rate, average fuel cost, and non-fuel O&M. These indices can be used for scenarios analyzing the competitive implications of efficiency improvements. The Cost/Supply Module provides marginal cost inputs to the Pricing Module. The Integrated RREMS System The RREMS System provides a logical structure for reasoning about the implications of the new electricity markets. Through simulation experiments, users can begin to evaluate alternative strategies for successfully participating in these markets. As actual experience operating in the new markets accrues, the System will be progressively refined. The key components of the RREMS System are the Pricing and Customer Choice/Market Shares Modules, which attempt to characterize how pricing strategy will be used by rival sellers, and how their pricing strategies will resolve market shares. This approach is a departure from other recent attempts to simulate the impact of restructured electric markets. For example, the California Energy Commission reported an extensive study by LCG Consulting that purported to simulate the new California market.8 What the study appears to have done was simulate the WEPEX power exchange auction, implicitly assuming that all California power transactions would be through the WEPEX. While the study showed a chart acknowledging bilateral power transactions outside the WEPEX, such transactions were not simulated. Such an approach is surprising, since direct access and bilateral forward contracts are the most important aspects of the new restructured markets. Both buyers and sellers will seek firm forward contracts to avoid the risks associated with the volatile and unpredictable spot prices (including the WEPEX prices). The prices in those contracts will only by chance equal the WEPEX or other spot market clearing prices. The process of rivalous price competition is already evident in California. Power producers, brokers, aggregators, and other sellers are aggressively offering power supply contracts to large power consumers outside the WEPEX. Power consumers are also sending out RFP's for power supply contracts outside the WEPEX. Each and every forward contract reached outside the WEPEX will have an impact on the WEPEX market clearing price, not the other way around. The most important task in simulating the new markets is therefore, the appropriate characterization of how the bilateral forward contract prices and supply commitments will be set. That is the primary aim of the RREMS Model. The RREMS Model operates in both evaluation and optimization modes. The evaluation mode solves for the implications of any specified set of seller prices. The optimization mode solves for the profit-maximizing pricing strategy for a specific seller. V. Data and RREMS Implementation As stated above, the challenge to electricity market forecasters and analysts is to develop useful market simulation tools without first observing the new electricity markets in operation. The RREMS System was designed to be initially implemented with relevant existing data, economic theory, and professional judgment, and to be progressively refined as actual market observations become available. The data exist to implement the Economic/Demographic, Demand, and Cost/Supply Modules. The regional data are available from commercial databases such as ESC's UQAR System and directly from the public sources, including: the FERC, EIA, RUS, Census Bureau, Labor Department, Weather Bureau, and Department of Commerce. The data do not currently exist to empirically implement the Pricing and Customer Choice/Market Shares Modules. The functional forms used are suggested by economic theory and practice and could be estimated using conventional econometric methods if the data were available. The values for the coefficients, however, must be initially based on professional judgment. As data become available, the coefficients can be progressively updated. VI. Some RREMS Results Simulation experiments with a partially implemented version of RREMS, for some Midwest markets, have yielded common-sense results, such as: Low cost producers will have extraordinary opportunities when restructuring allows them to pick their markets and design their prices. By contrast, high cost producers will find rough going. Typically, their plants will run fewer hours and produce less revenue than under current regulation. Passive producers, those who do not develop multi-market pricing strategies, will lose market share and revenue. Market shares will be very sensitive to rival price changes -- e.g., a cost reduction by one producer can affect market shares in multiple markets. The closer together relative seller costs are, the more volatile will be shifts in market shares. Endnotes: Direct access to non-utility providers of various customer services such as meter reading, billing, and collections is also being promoted in the restructuring debate. See Electricity Futures, New York Mercantile Exchange. See Jaske, Michael R. Forecasting Load Schedules, CPUC DAWG Working Group Background Paper, May 23, 1996, for a detailed discussion. Utility, Query, Analysis and Reporting System (UQAR), FERC Form 1, EIA 826, EIA 412, EIA 861, RUS Forms 7 and 12, and other data modules. UQAR ECON module. UQAR Weather Track module. TRADELECTM is a product of OnLocation, Inc., an ESC affiliate. Modeling Competitive Energy Markets in California: Analysis of Restructuring October 3, 1996; California Energy Commission, prepared by Rajat K. Deb, Richard S. Albert, and Lie-Long Hsue of LCG Consulting. ESC ELECTRIC UTILITY ANALYSIS REPORT 96-3 Copyright ©1996 Economic Sciences Corporation To comment on this paper, send e-mail to: comments@econsci.com Economic Sciences Corporation