Doubts intensify over New England’s radical new market for electric capacity.
Bruce W. Radford is editor-in-chief of Public Utilities Fortnightly.
What began nearly two years ago as a simple request by power producers to boost their chances for recovering fixed costs for several power plants in Connecticut has mushroomed into the single most complicated case now pending before the Federal Energy Regulatory Energy Commission (FERC).
Starting with must-run agreements for plants essential for reliability in local load pockets, and then offering a safe-harbor rate (PUSH) for peaking units with low capacity factors, the New England ISO (the ISO) now finds itself on the extreme cutting edge of FERC's already-edgy Standard Market Design (SMD). In sheer complexity, the New England proposal exceeds just about anything tried before. (See FERC Docket No. ER03-563, testimony filed Aug. 31, 2004, and thereafter.)
The goal is to solve the most confounding problem of electric utility regulation: how to encourage developers to build enough electric generating plants to ensure reliability-but without the systemic overbuilding of the regulatory compact, or the irrational exuberance of the recent merchant power boom and bust.
The case thus far has attracted a dean's list of star expert witnesses from all sides of the electric utility industry, including such luminaries as Steven Fetter, Steven Stoft, Roy Shankar, Steven Corneli, and Peter Fox-Penner, speaking for power producers, consumers, investors, and the ISO (not in that order).
But as the experts are beginning to discover, it ain't gonna be easy to manage this new model. And why impose it when New England enjoys a huge surplus reserve margin above the level of capacity deemed adequate under longstanding reliability rules?
In fact, New England would go beyond the basic market designs employed elsewhere for installed (ICAP) or "unforced" (UCAP) electric capacity. And it would improve on the second-generation downward-sloping demand curve model adopted with success last Spring in New York. (UCAP prevails in PJM and New York. It's the same thing as ICAP, except that plant capacity ratings are marked down - usually about 10 percent - to reflect an expected rate of unforced outages.)
Effective January 2006, New England would impose a fully locational ICAP market, known as "LICAP." It would set minimum capacity thresholds for electric generation in five different sub-zones encompassing the entire ISO market. The five separate zonal price markets would serve much like the nodal locational marginal prices (LMPs) in the day-ahead energy market.
By contrast, New York also has attempted to set locational UCAP requirements, but so far, only for New York City and Long Island. But as in New York, New England also would rely on a downward-sloping demand curve set by administrative fiat, to ensure that capacity prices rise and fall in a gradual manner, from scarcity to surplus, without "falling off a cliff," as occurs so often when the demand curve stays vertical. (See "New York Throws a Curve," Public Utilities Fortnightly, May 15, 2003; and the letter to the editor from Dr. Thomas Paynter, published July 15, 2003, p. 10.)
Yet the ISO's LICAP design strikes out on its own with radical new definitions of plant availability and after-the-fact revenue offsets to make sure plant owners selling through the monthly LICAP auctions cannot double-recover costs by selling energy simultaneously through the day-ahead spot market.
And more than that, this LICAP regime comes "hard-wired" (this apt description comes from Peter Fox-Penner) to make it virtually impossible for power producers to withhold individual units from the market to create artificial shortages to boost prices. Even if a unit owner "de-lists" the plant, or takes it off line for maintenance, or simply chooses not to participate in the LICAP auction, the ISO employs a fiction that assumes that all such plants actually did submit a bid. The supply curve includes all installed "iron in the ground," short of actual retirement and dismantling. (But the ISO would subtract net exports.)
This strange rule allows the ISO to "oversee" and eliminate the withholding of capacity without having to examine or second-guess the behavior of suppliers. With its new approach, the ISO complies with FERC's directive in June that New England's LICAP market must be purely voluntary, with no coercion from the ISO. (See, Docket No. ER03-563, June 2, 2004, 107 FERC ¶61,240.)
This miracle market also would feature CTRs (Capacity Transfer Rights), a near-exact analog to the tradable financial transmission rights (FTRs) featured in the energy market. The CTRs would serve as financial instruments, so that utilities and other load-serving entities (LSEs) who participate in the LICAP market could hedge the risk of any inter-zonal imports and exports needed to satisfy cost of their minimum capacity obligations, as imposed by the ISO.
Nevertheless, over the past 21 years, New England, on average, has enjoyed a supply of generating capacity greater than what the industry has required. This period indicates an average capacity level of 105.4 percent of what is needed to match peak demand plus the required reserve of 12 percent.
Some parties urge the ISO to boost the LICAP incentives even higher. They want the ISO to move its demand curve to the right, so that it is designed to encourage even greater supply margins than the ISO's current proposal. But listen to James Daly, the director of electric and gas operations for Boston Edison (the NSTAR subsidiary), testifying on behalf of a coalition of consumers and state attorneys general on how much the LICAP market could cost.
Daly claims that if the ISO's LICAP market should succeed in maintaining the ISO's targeted equilibrium level of electric capacity (known as C-target), then electric rates for New England's 6.5 million retail customers would climb by nearly $2 billion per year. That would amount to an extra $303/yr. on average for every customer, or about 1.5 cents extra per kilowatt-hour.
Paradoxically, the cost is even greater if LICAP does a worse job on encouraging investment in generation.
Thus, according to Daly, if the LICAP market "incents" developers to build only enough plants to serve peak demand plus the mandated 12-percent reserve (the ISO defines this level as "Operational Capability," or OC), then the yearly cost of the LICAP auction would soar to about $5.4 billion. That means $838/yr. per customer, adding an incredible 4.2 cents/kWh to the average retail bill.
And consider what would happen if the plan should succeed too well.
Suppose the LICAP auction encourages developers to build plants until capacity levels reach 1.12 times OC. That's a reserve margin of a whopping 24 percent, with an implied LOLE (loss of load expectation) at a miniscule 0.0019. In that case, Daly says that customers could end up paying $27,600 per year per megawatt to any generator electing to bid. That could reward a 100-MW gas turbine with a capacity payment up to $2.76 million a year, just for keeping itself available to avoid that single outage that would be expected to occur on one day in526 years!
The ISO would define five different LICAP zones within New England, tied to electric system geography (interfaces and such) rather than to political boundaries: (1) Maine, (2) Southwest Connecticut, (3) the Rest of Connecticut, (4) the Boston and Northeast Massachusetts zone, and (5) Rest of Pool. It would set the auction price of capacity within each LICAP zone using a complex process.
The ISO would calculate the quantity of generating capacity installed within any single zone (how much surplus or shortage), in terms of a percentage of the OC quantity. The ISO then would figure how much (if any) that LSEs should have to pay to developers to maintain that level of supply, or to encourage them to build more. (Or, if the surplus is great enough, exceeding the C-max level, then LSEs pay nothing.)
In truth, the demand curve (like the fictitious supply curve already discussed) does not represent a summation of bids by buyers, but in fact represents the ISO's administrative determination. The ISO estimates how much money should be paid to capacity suppliers, at various levels of surplus or shortage, in order to encourage suppliers to construct or maintain a target level of electric generating capacity in each zone. In this regard, the New England LICAP design mimics (on a zonal basis) the same basic theory that the NY ISO employs to set its demand curve.
Perhaps the strongest objections to New England's LICAP market design concern two major wrinkles: the "critical hours" rule, and the credit adjustment for infra-marginal energy revenues.
The first wrinkle is the definition of unit availability. First, the LICAP auction counts a plant as available only if it is actually dispatched by the ISO (or self-scheduled by the owner), or is capable of a startup in 30 minutes or less. But more important, the ISO defines certain hours as "critical hours" (100 each year, with a minimum amount for each month), based on the unit's "z-score," as determined by an evaluation of shortages, price peaks, and fuel-cost trends. Then, the ISO tests to what extent the plant was dispatched or scheduled during all of the critical hours during a rolling 12-month period. And some hours are deemed more critical than others. The ISO refers to this effect as "criticalness." The hours are weighted accordingly. Thus, a unit may qualify as available for only a fraction (say 40 percent) of such hours. That means the ISO will subtract a corresponding amount from the nominal LICAP price otherwise paid to the unit.
The second wrinkle is the so-called PER adjustment (Peak Energy Rent). The ISO determines the hypothetical amount of infra-marginal revenues that the unit should have received for energy sales. (Infra-marginal means in comparison to revenues earned by a hypothetical unit that operates on the margin - with its marginal cost equal to the energy market price). It then counts those hypothetical infra-marginal revenues as an offset against the LICAP price otherwise paid to the unit in question.
But what really sets the dogs to barking is the fact that each adjustment comes after the fact. The auction occurs in June, let's say, to set the price of LICAP for July, but when the auction market settles at the end of July, the LICAP seller is docked for energy revenues or the unavailable critical hours, as determined by events occurring during July. These ex post adjustments add uncertainty to any anticipated revenue stream from LICAP payments, the detractors say, putting merchant plant financing at risk, and making it impossible to engage in forward contracting to hedge the LICAP value at risk.
Speaking for a coalition of capacity suppliers, consultant Robert Stoddard, a vice president at Charles River Associates, calls this regime "commercially intractable." As notes Harvey Reed, a consultant for Constellation Energy, "There is no single, market clearing price ... There are as many prices as there are assets."
Utility financial consultant Steven Fetter, who cut his teeth at the Michigan Public Service Commission and later at Fitch, doubts that merchant gens could obtain financing in such an environment. Thomas Boland, managing director at the Seneca Financial Group, working in tandem with Charles River Associates, concludes that a merchant gen project likely would acquire only a "B" rating - meaning that its ability to pay its credit obligations likely would be impaired in the future.
Peter Fox-Penner, from the Brattle Group, concludes along with many other experts that the ISO's restrictive definitions of "availability" (especially the 30-minute startup requirement) will encourage developers to favor baseload or peaking plants at the expense of mid-merit or intermediate resources that cycle on and off. They fear that plant owners will respond by self-scheduling, to increase the chance that their units will achieve higher availability. But that only reduces the dispatch flexibility for the ISO.
Michael Hachey, power marketing director from TransCanada, warns that the failure of power plants to run during critical hours may not stem from gaming strategies, but from simple engineering factors, such as unit trips, failed starts, system disturbances, or ramping problems. And many experts from New York (and Long Island in particular) warn that New England's ICAP product, with its incompatible definitions of availability, will prove difficult to export to New York. They would prefer that New England retain the traditional definition used in PJM and New York (and in New England under the prior ICAP regime), based on the factor known as equivalent forced outage rate (EFORd).
How Much Reliability?
The proposal also raises questions about what size of reserve margin really prevails in the region, and whether it is adequate for reliability.
For example, the conventional wisdom across the utility industry states that SW Connecticut (SWCT) is a classic load pocket that faces a crisis in electric supply. Transmission constraints make it difficult to import power into the area, meaning that local resources come under strain in order to meet demand.
But if that is true, then why is it that the ISO's plan would treat SWCT as having a capacity surplus?
According to Thomas Austin, a staffer at the Maine Public Utilities Commission, the LICAP plan assigns a capacity ratio of 1.083 to SWCT.
"In other words," says Austin, "SWCT has 8.3 percent more capacity than it needs to maintain reliability, and about 3 percent more capacity than the ISO's target capacity level."
Other expert witnesses remark on similar aspects of the same phenomenon.
Steve Corneli and Robert Stein, representing the NRG companies, explain that when the ISO examines individual resources to determine how much capacity they can contribute to OC (peak demand plus the 12-percent reserve), the ISO counts all resources as system resources or "pool-planned units" (PPUs), capable of delivering power anywhere on the regional grid. And it then allocates these system resources to the individual LICAP zones according to load, when measuring the installed capacity recourse level of each zone, using the GE MARS software program.
However, when the ISO reviews a request by a resource owner for authority to retire or mothball a unit, it apparently applies a different standard, known as SCED (security-constrained economic dispatch). The SCED model is more conservative in estimating real resources, as it discounts some capacity, based upon grid constraints, simultaneous transfer limits, and various local constraints related to reactive power, grid support, and so on.
Corneli would prefer to move the entire curve to the right. He would prefer higher values for C-target and C-max. (Recall that C-max is the x-intercept, or the highest level of capacity, as a function of OC, for which the ISO would allow a LICAP auction payment-see Figure 1, p. 19.)
Generally speaking, utilities, LSEs, and consumer groups would prefer to move the demand curve to the left, producing lower prices at various capacity levels, and producing a smaller value for C-max. They question the generous incentive payment of 2 times EBCC (double the estimated marginal cost of constructing a new benchmark plant), as the ISO has proposed on its demand curve, when capacity levels fall below OC. Instead, some suggest that the premium should never exceed 1.2 x EBCC. (Speaking for the ISO, consultant Steven Stoft predicts that suppliers might end up earning the double-weighted premium payment about 15 percent of the time.)
In short, the ISO designed the curve so as to produce a long-term equilibrium that would maintain the historical average level of capacity (C-target, or 1.054 x OC), as observed over the past 21 years. On the y-axis, the corresponding equilibrium is where price equals 1.0 times EBCC.
But on the ISO's demand curve, a price of 1.0 x EBCC occurs at a lower capacity level - 1.037 times OC - a point the ISO identifies as the "kink" in the curve, or Ck. In reality, the price at C-target is only about 0.85 x OC. All of this is because the ISO front-loads the curve at low capacity levels with premium payments, but makes the curve less steep (with smaller negative premiums) at the back end, with the lower capacity levels. That's why the equilibrium point (C-target) is placed slightly off-center from the kink point.
In a unique perspective, John Daly points out that, due to the very steep slope of the LICAP demand curve, the ISO can create huge changes in LICAP price simply by recalculating the value of installed capacity along the x-axis. He notes that this could be accomplished through an apparently neutral process such as load forecasting. In his mind, the ISO should be required to apply to the FERC for authorization before it is allowed to recalculate peak demand or Operational Capability:
"The reality of the situation is that this is a rate case, and any change to OC would amount to a significant change in rates."
When all is said and done, the ISO's LICAP plan may prove to be most vulnerable because of several perverse results-situations in which the market seems to reach the opposite result of what you might expect:
• Balkanization. First, the very idea of dividing the capacity market into zones with separate prices, on order to tailor the remedy to specific load-pockets and problem areas, may actually increase the prevalence of market power and manipulation.
• Perverse Incentives. Second, while the plan is designed to increase resources and capacity delivery in particular crisis areas, it may have the opposite result when savings are considered across the entire ISO region.
• Gaming by Load. Finally, for all its worries about market power, and its hard-wiring of auction format to make it impossible for suppliers to game the markets, the ISO may have overlooked the opposite problem -that the biggest threat for gaming the market may come from the customers who buy the capacity.
First, understand that the ISO proposes to use the same demand curve in every capacity zone, scaled down in proportion to the zonal ICAP requirement, but without regard to the risks created by this smaller scale. But as Robert Stoddard has noted, generation investment is necessarily lumpy. Faced with the task of designing strategies to maintain or boost capacity levels in smaller zones, suppliers face the risk of much larger price swings. That's because the addition of even a single, minimum-scale unit sharply depresses the LICAP price in the pocket, and it might takes years for normal load growth to absorb the addition.
Also, there's the problem of the big fish in the small pond. A larger number of smaller-sized zones implies higher degrees of concentration of power plant ownership, with higher HHI indices.
Second, consider the perverse case of the 345-kV transmission expansion project now underway by NSTAR in the NEMA/Boston LICAP zone, and the Southwest Connecticut Reliability Project planned for the SWCT zone. John Daly, from Boston Edison, predicts that while the two projects likely will produce annual LICAP savings in the NEMA and SWCT zones ($8 million and $13 million, respectively) the projects could well add $42 million in annual LICAP costs across the entire ISO region, taken together.
This perverse result occurs because the projects are designed to produce benefits in smaller zones that, on a load-weighted basis, don't really represent the greater good for the entire region:
"The decrease in NEMA and SWCT clearing prices," Daly explains, "would impact fewer megawatts of capacity than the increase to [the] Rest-of-Pool and Maine clearing prices."
Finally, private consultant Roy Shanker worries about utilities gaming the market.
He notes that the ISO plan allows LSE's or other parties who represent load to offer up "phantom" load management projects in the LICAP auction process. That would depress the load requirement in the zone, and reduce the value of OC. Thus, as measured as a function of OC, the capacity level in the zone would increase, moving the zone to the right along the demand curve, and reducing the price.
But, as Shanker also notes, these load management programs have no obligation to perform, so the LSE's could then withdraw them without penalty after the LICAP auction is completed, but before the closing of the day-ahead market and the finalization of real-time dispatch:
Shanker sees this "gaming by load" as a perverse consequence and a serious flaw in the market design:
"Basically, load gets a no-risk ability to depress prices by artificially inflating the supply."
New England's LICAP Regime
1. Calculate LICAP Obligation.
• Region-wide Peak Demand. Fix peak demand for ISO region.
• Region-wide Reserve Margin. Add 12 percent margin to reach OC.
• Zonal Allocation. For each month (LICAP is a monthly obligation and the market clears each month), allocate OC by load among the five LICAP zones, and among LSEs by load within each zone. This OC allocation sets the LICAP requirement for each zone and for LSEs in the zone. (A portion of this requirement - the LSR, or "Local Source Requirement" - must be supplied from local resources because of local grid constraints and operational requirements.)
• Zonal Obligation. Require LSEs to obtain the required LICAP, either through bilateral contracting or through the monthly short-term LICAP auctions (two each month, held the month prior to the obligation month).
2. Calculate the estimated cost of constructing new generating capacity.
• Hypothetical Plant. This benchmark unit represents a "frame-style" gas turbine of utility-level quality, as opposed to an "aero-derivative" unit that is basically just an airplane jet engine turned on end. Characteristics include size (85-100 MW), heat rate (9,000 - 10,400 Btu/kWh), exhaust temp (990 - 1,120 degrees F), compression ratio (16:1), start time (30 min.), startup cost ($8000 plus fuel), minimum run time (4 hours), total installed cost ($550-650/kW).
• Zone-Specific Benchmark. For each LICAP zone, calculate the estimated benchmark cost of capacity (EBCC), expressed as $/kW-month. EBCC represents the estimated cost of building new generation on the margin. EBCC for each zone reflects the many separate, zonal-specific cost parameters that come into play, such as cost of land, labor costs, income and property taxes, etc.
• Benchmark Adjustment for Outages. In practice, EBCC for each zone will be 10 percent less than the nominal EBCC total, reflecting a benchmark standard forced outage rate.
3. Calculate the LICAP prices as set through the monthly auction
• Graph. Construct an x-y coordinate graph, with price on the y-axis expressed as a function of EBCC, and capacity quantity on the x-axis, as a function of OC.
• Demand Curve. Draw a sloping demand curve (see figure). This curve represents an administrative determination of how much that LSE should be willing to pay to capacity suppliers, rather than a summation of actual bids submitted by LSEs.
• Capacity Quantity. At each monthly auction, calculate the quantity of existing installed capacity for each zone on the x-axis. (This amount is a different value from the actual number and quantity of capacity supply bids submitted by suppliers - more on that later).
• Corresponding Price. At that value of x (quantity), move up in vertical direction to find intersection with the demand curve. The corresponding value for y (a function of EBCC) will represent the nominal price of LICAP.-BWR