California's 1993 qualifying facility (QF) auction dramatically illustrates problems that can be encountered in structuring auctions for electric utility solicitations of supply-side resources from qualifying cogeneration and small power production facilities.
In the 1993 California QF auction, three California utilities were to select QFs that would be awarded long-term purchased-power contracts. The auction produced some unexpected outcomes that could potentially cost the utilities and their customers tens of millions of dollars per year. A year after the auction was held, the parties were still attempting to revise contracts, or even rebid a portion of the auction. And all of this effort may come to naught, depending upon what happens now that the Federal Energy Regulatory Commission has struck down the auction process under California's Biennial Resource Plan Update, claiming that it violates the Public Utility Regulatory Policies Act (PURPA) by using an improper method to calculate avoided costs.1
A study of the California QF auction illustrates yet again that "the devil is in the details." And the stakes will only rise farther as regulators in other states rely more on auctions to open up the electric industry to further competition.
THE SECOND-PRICE DESIGN
The California Public Utilities Commission (CPUC), the utilities, and the QFs used a very simple second-price auction as a design model:
s Each bidder supplies one item, identical to the item supplied by any other bidder.
s Each bidder submits a single sealed bid specifying the price at which he is willing to supply the item.
s A set number of lowest bids win the auction.
s Each winning bidder receives the price bid by the first losing bidder (em that is, the lowest supply price that failed to win.
In this auction, each bidder will maximize his expected profit if he bids his actual cost of providing the item, including any opportunity costs and risk premiums. A winning bidder cannot increase the price he receives, even by bidding a price above his actual cost, since the auction price is set at the level of the first losing bid. A bid far above cost only increases the bidder's chances of losing. By bidding less than his actual cost, he increases his chances of winning, but at a price below his cost. In this simple second-price auction, theory and experiment show that a bidder will bid his true cost. This effect is often referred to as the "true-cost-revealing property" of the second-price auction.
Another feature is also noteworthy. In the simple second-price auction, all winning bidders are paid the same price for the items they sell. That is, the second-price auction sets a uniform price that is paid to all winners.
The CPUC mandated a second-price auction because of these two features. It wanted each QF to bid its true costs, and such an auction would naturally select generation resources with the lowest cost to society as a whole.2 It also wanted an auction in which the winners were paid a uniform price for their capacity and energy. In selecting a second-price format, the CPUC noted that "paying QFs at full avoided cost . . . is also a uniform price system . . . [T]he second-price auction is the logical extension of full avoided cost."3
THE QF SETTING
The CPUC, QFs, and utilities worked to extend the steps in the simple second-price auction to the QF setting. This task was accomplished through a series of design steps and CPUC decisions.
The QFs who bid differed in terms of unit size, capacity cost, energy cost, availability, transmission losses and costs, expected dispatch by the utility, and contract length. The winning QFs would supply two items of value to the utility: energy (in all periods of the year) and capacity (in peak-demand periods). But the California procedure combined all QF operating characteristics and any related benefits accruing to the utility into a single cents per kilowatt-hour (›/Kwh) figure. In this way, the QF auction treated all kilowatt-hours as equal, even though the bidders were trying to win two different things:
(a) the right to displace capacity of designated resources from the utility's resource plan (the Identified Deferrable Resources or IDRs)
(b) the right to use the utility's transmission capacity.
This duality led to a sequential bid scoring and selection procedure. In each step, only one winning bidder was selected. The IDR capacity and transmission capacity awarded to that winning bidder were removed from the auction. At the start of each step, the remaining bidders were rescored using the remaining unassigned transmission capacity. The CPUC specified the blocks of capacity (the IDRs) that would be auctioned by each utility. It compared the QFs on a cost-per-Kwh basis and assigned each QF a single ›/Kwh score that reflected the costs and capacity factors bid by the QF. The CPUC selected winners until it filled the effective capacities of the IDRs. The auction price (per kilowatt-year) was pegged at the "second price," giving a premium to the winning bidders.
Each winning QF would receive a contract to sell energy and capacity over several years. Thus, total payments to a winning QF during a given year would depend upon factors not known at the time of the auction. These factors would include the QF's actual availability and energy production, plus external factors such as inflation rates.
DISTORTIONS AND GAMING
The QF auction looked like a true second-price auction. Most parties simply assumed that the QFs would bid their true costs of supplying capacity and energy. But some parties testified that QFs would not want to reveal their true costs. For example, after winning the auction, QFs would have to negotiate with their suppliers and regulatory agencies. They would naturally want to conceal their precise amount of profit. The CPUC discounted these arguments.4
In addition to these factors, several features internal to the QF auction could have caused it to lose its true-cost-revealing property. Here are two examples. The first shows how QFs could increase payments by cutting capacity factor; the second shows how a large QF can increase its payments by bidding above its true costs.
Example 1 (em
Earning More With Less
A 15-megawatt (MW) wind-powered QF submitted a bid with the following operating data:
Peak Capacity Factor: 100%
Overall Capacity Factor: 90.39%
Cost: ($/Kw-year) 5550
Peak Shortage Cost:
Energy Cost: (›/Kwh) -69
This QF submitted a bid with a negative energy cost. The auction allowed a QF to bid a negative energy cost to account for tax credits for energy production from renewable resources. With a negative energy cost, the QF would pay the utility to take energy. The utility would pay the QF for capacity. Nevertheless, this bid is unusual for several reasons. First, the capacity factors total about five times higher than expected for a wind project. Second, the energy-related capacity cost runs 10 to 50 times higher than expected for a wind project. Third, the energy price of -69›/Kwh seems much lower than tax credits could support.
Otherwise, this QF achieved a score of 1.43›/Kwh in 1998 dollars, or about two and one-quarter cents under the score for the first losing bid for the IDR block, which came in at 3.68›/Kwh. This differential would produce a second-price premium of $179.67 per kilowatt-year. But capacity factors and performance factors would be treated differently in the bid-scoring procedure than in setting actual payments under a purchased-power contract, giving rise to possible distortions.
In scoring the bid and setting the second-price premium, the auction assumed that the QF's operating hours per se could be the minimum of 8,760 times the all-period capacity factor (90.4 percent), 8,760 minus allowed maintenance hours (840), or the economic dispatch hours (7,920). If the QF operated accordingly, it would generate 118.8 gigawatt-hours (Gwh) per year, receiving a net payment of $4.139 million in 1988 dollars.
But now suppose that the QF operates at more realistic peak-period and all-period capacity factors of 20 percent. In this case, the QF would generate only 26.3 Gwh per year. Assuming that the bid capacity factor remains constant for all periods and the QF uses all its allotted 840 hours of maintenance per year, its performance factor would be (0.2 x 8760) , (0.904 x [8760 - 840]) = 0.2447. The QF would receive yearly payments in 1998 dollars of $5.390 million (em up $1.25 million from the smaller payment with a larger capacity factor.
Why the increase?
Remember, because of offsetting tax credits for renewable energy, our wind-powered QF actually carries a negative energy payment. For each kilowatt-hour less of energy, the QF cuts its "payment" to the utility by $0.69. Of course, its performance factor would fall as energy production moves down to the lower capacity factor. However, its performance factor is divided by 0.9 (and capped at 1.0) when calculating the capacity payment that the utility makes to the QF. Consequently, the capacity payment that the QF receives falls at a slower rate than the rate of decrease in the negative energy payment.
One utility received 23 bids similar to Example 1. The problem with such bids comes from the fact that the CPUC's bid-scoring system diverged from the method used to calculate actual payments.
Example 2 (em
Both a Winner and Loser
A QF bid a 274-MW project against an IDR with the same effective capacity:
Energy-related Capacity Cost:
Peak Shortage Cost:
Energy Cost: (›/Kwh) 0
The QF was tentatively selected as a winner for its IDR. However, some smaller bidders won a share of the IDR capacity by posting lower scores. As a result, the capacity block won by the large QF was downsized to 271.6 MW. A small part of its bid (2.4 MW) lost in the auction. Consequently, it was also the first losing bid for the IDR under the California QF auction rules. That means that its winning bid inadvertently set the auction price. Thus, for the 271.6 MW that won, the QF would be paid exactly what it bid and would receive no second-price premium. It would receive payments of approximately $59.44 million per year in 1998 dollars.
If the QF's bid has actually represented its true costs, these yearly payments will cover its costs but no more.
Suppose instead that this QF had increased the energy-related capacity cost (ERCC) component of its bid. By so doing, the QF would have also increased its own winning bid and, coincidentally, also the first losing bid score. Consequently, the payments to all winning QFs would have risen, including the payment for the 271.6-MW block that this QF won. For each $1 per kilowatt-year increase in the ERCC, this QF would have raised its capacity payments by $271,600 per year. That's pure profit. This strategy would risk losing the bid, but might prove worth it in the long run.
FIXING THE HOLES
If a QF auction is meant to elicit a particular high-level behavior from the participants, the designer may have to compromise in choosing a model.
Several parties recommended developing the QF auction by extending a first-price auction. They argued that a first-price auction format was easier to extend to complex products such as the capacity and energy being offered by the QFs. But the CPUC rejected this approach, assuming that the resulting auction would not exhibit the desired uniform pricing and true-cost-revealing characteristics. While the CPUC succeeded in achieving an auction with a strong uniform-pricing characteristic, the results show less success in achieving the true-cost attribute. Of course, an auction with both these properties could have been designed, but only by using an alternate view of a simple second-price auction as our model.
Here is an alternate definition of a simple second-price auction:
s Select N winning bids to minimize the sum of the costs of the winning bids.
s Pay each winning bidder the most that he could have bid and remain a winner.
In this alternate approach, the utility selects winners from among the QFs and IDRs, consistent with regulatory constraints, to maximize the capacity and energy benefits provided by the winning QFs and IDRs (after subtracting the capacity and energy costs of the winning QFs and IDRs, and the transmission costs borne by the utility). The data required under this approach includes the same type of data used to score bids in the California QF auction. But the CPUC would achieve the desired results from the way it defined the payment structure.
One would calculate the premium due to a winning QF by removing that QF from the auction and resolving the optimization problem. The second-price premium for the QF is the resulting increase in the objective function. As in the California QF auction, this premium is allocated over the QF's capacity costs. The payment mechanism in the contract can be made compatible with the auction structure through a similar analysis defining the payment penalties for a QF that does not meet the capacity factors that it bid.
In this auction, a QF would maximize its profit by bidding its true costs and capacity factors. This bidding strategy would maximize a QF's profits no matter how the other QFs bid. And this definition of a second-price QF auction offers an added benefit. Even though the calculations involved are fairly complex, it is easy to explain how a winning QF's prices will be set: Each kilowatt of capacity that a QF wins in the auction will be paid the highest cost the QF could have bid for it and still have won the right to provide that increment. That's easy to understand for bidders. t
Paul Gribik is a vice president of Mykytyn Consulting Group, Inc., a firm specializing in business analysis and information technology for energy and telecommunications firms. Dr. Gribik's recent work includes structuring auctions for electric and gas utilities, and developing models for evaluating the business potential of new energy technologies. He has a BS in electrical engineering and an MS and PhD in industrial administration from Carnegie-Mellon University.
Articles found on this page are available to Internet subscribers only. For more information about obtaining a username and password, please call our Customer Service Department at 1-800-368-5001.