July 21, 1997 Jonathan Jacobs Manager of Market Evaluation Pacific Gas and Electric 77 Beale Street San Francisco, CA 94177 Dear John: I am sending this to you via the fax because it may contain information that would require you to destroy it or to black out selected sections after you have read it. (I can edit it as you may request and then send an email version.) As is noted in the report I sent you, we have "verified" the inevitable dynamics of deregulation by using our models to correctly predict the evolution of events in the UK (where I spent the year in Cambridge developing the European energy analysis capabilities for the European Commission), South America, Australia, and the US (were we have not only compared the results to published events, we have modified the model to take out the computer's decision algorithms so that humans - utility executives and commissioners - can make the decisions and change the rules - only to show the resulting same set of events). This note indicates that we know how to take advantage of those events over the full spectrum of subtle to extreme gaming tactics. We have a methodology that directly considers the uncertainty in competitor and regulatory actions/responses. It determines any underlying patterns and can automatically produce an certain-to-win but continuously changing portfolio of tactics based on the actual market conditions it sees and analyses. To put this proposal in perspective, the transition from a regulated to a deregulated market presents a tumultuous time for all the companies that prospered under the "old" rules. Regulators attempt to provide an orderly transition from the past to the future that protects existing rights and commitments while simultaneously clearing the way for new free-market activities. These two demands are in economic opposition and new rules necessarily contain ambiguities and market distortions. Physical constraints to the system insure that any definition or rule will become ambiguous under some operating condition @unt 1996, p. 61) and thereby allow competitors to respond in way that dramatically changes the expected character of the future market place2 (Newbery 1996, p. 64). In ' Sally Hunt and Graham Shuttleworth, Competition and Choice in Electricity, John Wiley and Sons, 1996, p.61. 2 David M. Newbery, "Regulation, Public Ownership and Privatisation of the English Electricity Industry" in International Comparison of Electricity Regulation, Richard Gilbert, and Edward Kahn, ed., Cambridge University Press, 1996. California, significant efforts were made to ensure the market acted as the regulators deemed appropriate. This not only causes market distortions that can be advantageous to some and detrimental to others, it also means that the rules probably will have effects contrary to the desired effect. As Richard Tabors of MIT notes:".. .the market rules cannot be gotten right on the first try if the goal is a pre-designed supervised market. The regulatory interference has prevented the types of market mechanism ... normally seen in a commodity type market"3 (Tabors 1996, p. 47). From electricity deregulation experience in other parts of the world, "the message .. is clear. It was incorrectly assumed that the new commercial entities would continue to operate by the intent of the rules, even if not formally stated, when the new structure began. But commercial markets are commercial markets, profits are profits and any commercial advantage will be taken" (Tabors 1996, p. 49). The dynamics of deregulation as they relate to how competitors may act toward each other are documented further by Wolak4 and Backus.5 My work has focused on the games people play during the transition from the regulated to the deregulated environment. During this period, the optimization tools typically used by utilities cannot provide adequate answers. Not only do the uncertainty in rules and competitor responses invalidate the assumptions of the analysis, I can show that an optimal "player" is the one that is easiest to defeat in a readily-producible "gaming" situation. The comprehensive gaming we address necessarily includes strategic, tactical, operational, real time components. Wars are not won with a singe one-shot battle using a single tactic and a single weapon. We simulate (with the needed validation and confidence bounds): dynamically changing own and competitor financial strength, tactical sequencing, repetitive and non-repetitive tactics, portfolios of tactics, misinformation efforts, gambits, bidding, diversions, third-party tacit cooperation, counter-responses, counter-measures, over/under booking, availability strategies, protocol ambiguity/rule conflicts, congestion tactics, must-run/take tactics, market separation effects (e.g., day versus hour versus ancillary), market isolation (region and segment), combined customer/suppler distortions, signaling, financial instruments/hedging, financial restructuring/asset placement, divestiture re-direction, affiliated marketers, affiliated generators, exacerbating weather effects, new rule "guidance," legal restraining orders/law suits or any other categories that become recognized. If required, we can keep the gaming below regulatory thresholds or hide it in the noise of the market. Some "desirable" tactics could be blatant to cause a rule to be changed to hurt a competitor more than PG&E, or to help PG&E more than it helps the competitor. We also show the advantage of both the regulated and unregulated segments of the company 3 "Lessons from the UK and Norway," Richard D. Tabors, IEEE Spectrum, New York, pp. 45-49. 4Frank A Wolak and R.H. Patrick, "The Impacts of Market Rules and Market Structure on the Price Determination in the England and Wales Electricity Market, Stanford University Department of Economics Working Paper," April 1997. Available on www-path.eecs.berkeley.edu/UCENERGY. 5 George A. Backus, The Dynamics of US. Electric Utility Deregulation, U.S. Department of Energy, Office of Utility Technology, Washington D.C., August 1996. having the tool so that efforts are coordinated - but without direct communication. "Perversities" in the market dynamics exist such that the unregulated marketing arm can, under an orchestrated game plan, be most useful helping adversaries take away UDC market share. Similarly, generation may want to over-sell to old grandfather contracts even if the power "goes" to an out-of state supplier. We can determine and coordinate the comprehensive set of moves, from real-time to long-term end-game, that maximize income or any other measure of company success. The highly convoluted rules of the California system, by nature, produce abundant but highly convoluted gaming moves. We call what we are proposing a "Real-time Competitive Response System." The system, however, does and must include the financials simulation of all California market players as well as the full simulation of all the North American gas and electric energy suppliers. This allows both short-term and the long-term analysis of merger, acquisition, take-over, bankruptcy and market entry activities. In additional to the financial simulation capabilities, the model contains the plant, electric transmission, and gas transmission physical detail. The detailed aspects of the PX and ISO will be included as they change AND are expected OR preferred to change. It also includes the all-fuel, end-use simulation of demands. Very realistic (and accurate) retail-wheeling and marketing dynamics/impacts are intrinsic to the model. Gaming may be a dirty word to FERC and the California commission, but the sooner the market clears out the distortions, the better it works for everyone. The "gaming" defeats the flaws in the system and ultimately removes the players or features that lead to market distortions. There may be ethical issues related to "the end justifying the means" but there is a large region of opportunities between what is ethically viable (profitable) and ethically dangerous (illegal). It is prudent to understand the full spectrum of possibilities, and through the understanding of market dynamics that it provides, to select that appropriate subset of strategies which best serve the long-term interests of PG&E. In 1986 (not 1996), Policy Assessment Corporation (PAC)and Systematic Solutions Incorporated (SSI) developed the Competitive Industry Gaming Model (CIGMOD) to analyze the dynamics of deregulation for the State of Illinois. The model was based on the ENERGY 2020 model used in 40 states and provinces in North America by the energy industry and regulators. Versions of ENERGY 2020 are the US National Energy Policy Model (FOSSIL2/IDEAS) and the Canadian government's National Energy Planning Model (ETPM). ENERGY 2020 was selected to analyze the energy and environmental evolution of Western Europe. The model is extensively used in Eastern and Central Europe to address the severe dynamics of both economic and energy deregulation there. We are currently initiating the invited efforts to assist Brazil in its deregulation process. The model is also being used by the US and Canadian governments to provide analytical support for the international greenhouse gas negotiations. When the conventional models used in the UK to analyze deregulation failed to reproduce the unfolding events, it was the re-parameterized US CIGMOD model that reproduced the gaming dynamics, the re-regulation activities and the merger processes within the UK. Most observers would argue that the UK experience has little to do with the US. The CIGMOD "experience" turns this view on its head and goes so far as to say a "market is a market" in all languages and countries. The work of Sally Hunt as referenced above, later corroborated that conclusion. Our combined efforts with the Perot System's (PSC) staff working on the ISO (primarily those familiar to PG&E: Paul Gribik, Ali Vojdani and Dariush Shirmohammadi) have shown a large number of additional gaming opportunities that their unique experience with the PX, ISO and the PX/ISO interface allow. These opportunities are not only in the design of the PX and ISO itself, but also in the data transfer, settlement, and physical response issues. Thus, my associated organizations (PAC and SSI) along with PSC provide a capability that is not even remotely available elsewhere. What we would propose to PG&E would be specific tasks that adds the California ISO/PX simulation to and modifies the structure of the deregulation dynamics model developed by Systematic Solutions, Inc., to faithfully reflect the specific rules and protocols associated with the California market as they change from now until the end of the transition. The Project The proposed effort could tentatively be divided into three or more phases to verify that each current phase indicates the success of the next phase. This hypothetical proposal is provided for exemplary purposes only. An entirely different effort, more or less ambitious, could be readily accommodate. Careful consideration of the system's design strongly indicates minimal technical risk under all hypothesized conditions. We have all the tested parts to the system except the real-time PX/ISO data acquisition software. The parts do need to be modified to reflect the specific and continuously changing, California regulatory conditions, PG&E priorities, and the calculation times needed to provide real-time response. The changing ground of the California deregulation process, however, also indicates that certain tasks may not be needed, that some tasks may need to be redefined and that other tasks may need to be added in Phases II and I. For example, recent (June) modifications to the PX protocols would require that our PX simulation model be modified accordingly for the interim PX operations. This example proposal provides for the testing of the complete system in Phase I, followed by two other phases contingent on the expectations of the PX/ISO release of specifications, an exhaustive model testing regimen, and PG&E staff training needs. This phasing also allows PG&E to gain confidence in the system and approach prior to using it in the actual market place to help facilitate up to $12 million in transactions per day. Phase I The first phase, called the "functional model phase," reviews all the information of the California ISO and PX as well as any other relevant commission rulings to determine any areas where PG&E or competitors could find opportunities. These are translated to strategies and incorporated into the existing strategy options within the current deregulation model framework. The simulation of the way that the actual PX/ISO operates is also incorporated into the model. Perot Systems Consulting (PSC) has help designed these tools for the actual PX/lSO and are thus the most capable to make sure this project provides valid simulations of actual PX/ISO operations. Further, PSC's intimate knowledge of PX/ISO protocols has a particular advantage for this project. The deregulation framework contains detailed demand, energy, financial and physical representations of all the known potential players in the California market. (Unanticipated entities can be added as needed.) The short-term forecasts for California and local competitors demands are determined endogenously so that the generation available to serve the California market can be correctly ascertained. Phase I provides the integration and testing of all critical components as well as determines what protocols and rules have relevancy to future market conditions. PG&E staff will use this framework to learn the modeling system plus gain confidence in its usage as an operational tool. The human interface will also be modified as needed to maximize the efficacy of using the model for bidding and other strategies. The real-time data available from the PX/ISO will not have been determined during Phase but will be the focus when it becomes available for Phase II. Nonetheless, the ability of the model to simulate basic competitive strategies against human players will be tested in Phase I to ensure that the model performs as expected. Phase II The second phase is the "operational model phase." In this phase the project team determines the available and actual real-time data from the ISO/PX and adds the analysis capabilities to take advantage of that data stream within the model. Advanced statistical methods (primarily associated with the technical analysis of commodity time series and cointegration) will be added to the model. These sub-systems will determine the statistical qualities of the prices, supplies and demands in order to understand the underlying direction of their variance and averages. Depending on the intelligence data available PG&E pursues, the software will be able to detect any pattern in a competitor's actions that may be detrimental to PG&E operations or whether competitors are somehow directly tracing any PG&E efforts. In Phase I below, neural networks may be advantageous for the short-term forecasting of price and demand. Neural nets, however, do require specific data structures that would not yet be understood in the early days of the deregulation. Both Phase II and Phase III use a sophisticated confidence/validity package called HYPERSENS. It can search essentially all the possible options to find those real-time strategies that ensure PG&E is not disadvantaged by the behaviors of competitors. HYPERSENS cannot only accommodate the uncertainty in competitor behavior, generator availability, and hourly demands, it can determine the set of options available that best meet PG&E goals. Because the modeling system of Phase II will have the capabilities to analyze hourly data, test data sets will be generated for PG&E staff to use as a means of learning and gaining confidence in the system. Any uncovered limitations will be mitigated as appropriate. Model testing with PG&E Staff is a significant part of the project to insure that the model does well under anticipated and unanticipated conditions. Synthetically produced data streams that approximate what is actually expected during the first quarter of 1997 will be used in the testing. Phase III The third phase is the "real-time implementation phase." Only after January 1, 1998 will real data be available for both analysis and real-time decision-making purposes. The modeling system described here will need to be monitored to insure that it produces the. information PG&E expects. The system will contain software that notifies staff of any anomalies that may jeopardize the result's validity. Further, fail-safe" positions will be included that limit any financial exposure during the time interval before adequate data have become available to fully validate the system's operation. The algorithms will be "tuned" as appropriate and enhanced features are added as required. The model will also be benchmarked against "reference" assumption to determine its success rate compared to equilibrium conditions. The model methodology will, by default, certainly have the ability to determine the best bidding strategy given market information. The only area where added care and testing must be taken is in those strategies that respond directly to the aggressive actions of a competitor. Optional Follow-on Phases The data flows during the transition may be overwhelming to PG&E operations. We also have the ability to provide the data-visualization (GIS) that clearly delineates for PG&E staff the options and recommended actions. This activity could constitute a Phase IV As an intrinsic part of the system, the CIGMOD deregulation dynamics model will allow PG&E to test the impacts of others entering the California market (generation or marketing) or the impacts of (gas and electric) mergers, acquisitions, and takeovers among competitors or by PG&E. This could represent a Phase V consulting effort. The consolidation (or dilution) of the market changes the strategies that a company must use to insure profitability. This effort would extend the short term (hour-to-hour to monthly) analyses to longer time frames, thereby providing company-wide analyses of the impacts of such longer-term strategies on day-to-day as well annual operations and profitability. This Phase can also test competitors' financial strategies which could impact PG&E's operations/profitability. The ISO/PX system in California is premised on the assumption that DC analysis is adequate to determine system requirements and behaviors. Efforts of the project staff indicate that the AC considerations lead to significant additional opportunities for competitors to disadvantage PG&E operations. Some of the advanced technical efforts to increase the transfer capabilities within WSCC may make this area the most crucial to insuring the real-time availability of PG&E generating units. Adding AC considerations to the proposed modeling system could represent a Phase VI. The specifics of these potential Phases IV through VI are not pursued further here. Attached is an example of what the tasks of a "real" proposal might look like. I want to make this as concrete for you as possible so that your evaluation is as easy as possible. Please contact me with any questions or concerns you may have. Sincerely, Dr. George Backus President