Logo E-Lab
January - March 2001 Issue

Buying and Selling Electricity: Getting the Best Deal

One of the most intriguing aspects of the new competitive industry is how electricity is bought and sold. Many deals between electricity generators--remnants of the traditional utilities and new profit-making companies--and their potential customers take place in a centralized "spot" market. Generators and customers submit bids specifying how much they will sell or buy at various price levels at a given hour on the following day. (In most regions, demand is simply estimated; at present, individual consumers generally cannot or do not respond to price.) The system operator then informs generators whether they can run in a given time slot (the lowest-priced generators are chosen first) and how much electricity they should produce. The system operator also tells customers how much electricity they will receive and--after the transaction is completed--the price they must pay for the electricity generated plus transmission service.

The financial risk to all participants in the spot electricity market is enormous. One cannot buy electricity now and use it later because practical methods of storing electricity are not yet available. As a result, generators' incomes and customers' expenditures are constantly at the mercy of fluctuating spot market prices. However, participants can "hedge" their risk by buying long-term contracts on the "forward" market. The generator agrees to provide a certain amount of electricity at a set price at a defined time in the future, and the customer pays a fee to retain the right to exercise that deal when the time comes.

Generators and customers can thus lock in a deal for the future. But there is a new risk. How will the contracted price compare with the spot price that prevails when delivery time comes? If the contracted price (plus the fee) is lower than the spot price, the contract makes the customer better off and the generator worse off. If the spot price is lower than the contracted price, the opposite is true.

Graduate students Petter Skantze, Andrej Gubina, and Michael Wagner, supervised by Dr. Ilic, have developed techniques that will enable individual customers and generators to determine whether a proposed long-term contract is likely to be a good deal. Making that determination requires predicting electricity prices--a difficult task. Techniques are available for forecasting prices of most commodities. But those techniques do not suit electricity because it is not like most commodities. Gold or coal is easy to store; electricity is not. And for most commodities, tomorrow's prices bear some relationship to today's. In contrast, electricity prices are highly variable. The value of electricity varies with conditions on the power system, so a spike in demand or a plant outage can bring an abrupt change in price.

The MIT team has therefore developed a new method of forecasting the spot price of electricity for a given regional market. The price of a commodity is determined by the combined forces of supply and demand. Therefore, the researchers look at the behavior of electricity supply and demand independently and then combine the outcomes to infer price. They have developed a "physical" model that simulates the operation of an electric power system. To determine how supply and demand will move on the system, they use physical "drivers" that are more easily predicted or have been characterized by abundant historical data. Drivers for supply are prices of fuels (oil, coal, natural gas), rainfall (which affects hydro capacity), generation outages (unexpected and planned), and the availability of emissions rights contracts. Demand is driven by temperature (estimated using historical data) and economic growth (a long-term effect). Finally, the researchers draw on models of other regional power systems to track the impact of imports and exports on supply and demand, hence on price.

The supply and demand estimates generated by the physical model run a pricing model that forecasts electricity price over the short and long term. Any price spike that occurs can be traced to one of the physical drivers. As a result, the model can determine how quickly price will revert to "normal" levels--a critical factor in assessing the value of a long-term contract.

Armed with those price forecasts, generators and customers need to know what long-term contracts--prices, quantities, and dates for delivery--will give them the best financial outcome. That dilemma is standard fare for traders of options for stocks and commodities; techniques exist for valuing traded options. But again the researchers had to adapt the traditional financial techniques to suit the idiosyncrasies of electricity. Taken together, the pricing model and the options-based decisionmaking tool enable a market participant--either generator or customer--to define optimal strategies for using long-term contracts to hedge the risk involved in electricity sales and purchases.

A case study demonstrates the power of the new MIT tools. The study focuses on a utility company that has sold off its generating units (as recommended by regulators) but still has an obligation to serve its customers. The company is in a precarious position. If it loses more than $100 million in the coming year, it will go bankrupt. Depending solely on the spot market is unacceptably risky, so the company is considering buying a set of contracts for the next twelve months. Company managers have decided on an acceptable level of risk: the company's "portfolio" of monthly contracts must have less than a 5% chance of incurring a loss of $100 million. What portfolios of monthly contracts will meet that criterion?

To perform the analysis, the MIT researchers calibrate the spot pricing model for this generator's electricity market using historical data for the relevant drivers. They then define a single set of possible monthly contracts for the coming year and, using the options-based tool, calculate the profit (or loss) for the year under that set of contracts. Since demand and spot price are uncertain, they allow those inputs to vary randomly and re-analyze the same set of contracts repeatedly. By similarly analyzing hundreds of sets of possible monthly contracts, they define a group of portfolios that have less than a 5% probability of losing the company more than $100 million. Thus, the MIT decisionmaking tools can help generators minimize their loss--or with luck maximize their profits--from their electricity generating equipment.

The researchers are now working with other MIT investigators to find "smarter," less computer-intensive methods of implementing their analytical techniques. In addition, they are adapting their tools for use in "dynamic hedging." In real life, forecasts of fuel prices, rainfall, economic growth, and other drivers of supply and demand continually change. The adapted tools are intended to help decisionmakers constantly adjust their portfolio of contracts in response to such changes, thereby maintaining an optimal portfolio over time and increasing the probability of investment recovery.

Marija Ilic is a senior research scientist in the Department of Electrical Engineering and Computer Science. Petter Skantze is a PhD candidate in the same department. Andrej Gubina is a Fulbright visiting scholar at the Energy Laboratory. Michael Wagner is a master's degree candidate in the Department of Electrical Engineering and Computer Science. This research was supported by the MIT Energy Laboratory's Consortium on New Concepts and Software for Competitive Power Systems: Operations and Management. Consortium members include ABB Power T&D Company, Inc.; Constellation Power Source, Inc.; Electricité de France (EdF); and TransÉnergie US Ltd. (a subsidiary of Hydro Québec). Futher information can be found in references.

[Back to Table of Contents]


[e-lab Home Page] [Energy Lab Home Page] [MIT Home Page] [Up]

Last updated: 26-Jul-2001
Copyright © Massachusetts Institute of Technology 2001. Material in this bulletin may be reproduced if credited to e-lab.