Yes, We Have No Negawatts


When you sell demand response back to the grid, how much capacity are you now not buying?

Fortnightly Magazine - October 2011

The first hint of trouble came in December 2009, when PJM revealed in a load management performance report that customers who bid demand response (DR) resources into the RPM capacity market for the 2009/2010 delivery year had actually over-performed—they had backed off 18 percent more capacity (1,299 MW) than promised.

Trouble, that is, because when PJM’s independent market monitor began to parse the data, that surplus started looking more like a shortage.

Led by its president, Joseph Bowring, the IMM Monitoring Analytics LLC had found that by taking advantage of a loophole in PJM rules, certain aggregators of retail customers (known in PJM as curtailment service providers, or CSPs) had figured out clever ways to assemble portfolios of demand-side resources so as to earn twice the credit for capacity relief that PJM ordinarily accorded to DR offers.

So in February of this year, PJM and the market monitor issued a joint statement declaring that any CSP choosing to exploit market rules by “double counting” the capacity credit earned by DR resources would seen as engaging in market manipulation. Two months later, PJM filed tariff amendments with the Federal Energy Regulatory Commission, seeking to plug the gap. (FERC Docket ER11-3322, filed April 7, 2011.)

What the IMM found, as later documented in its 2010 State of the Market Report, released in March 2011, was that for any given emergency event, when PJM on a peak day called for demand response to ease stress on capacity and reserves, at least 47 percent of participating customers had failed to meet even half of their committed customer-specific reductions in demand, and that approximately 31 percent of customers showed little or no demand reduction at all. Yet the CSP portfolio as a whole was ahead in the game.

The reason was obvious: aggregators were arbitraging by matching over- with under-performers—and doing so with PJM’s blessing.

On one hand, PJM tariffs said that no customer could receive credit for capacity reductions any greater than its past peak load contribution (PLC)—as determined by its actual power demand recorded during the five highest daily coincident-peak hours in the year prior to the delivery year. In PJM’s eyes, that PLC metric represents the amount of capacity that the RPM market already would have procured for each customer. And so even if a customer should offer to back off all its demand, down to zero, it could never free up a measure of capacity any greater than its PLC—the baseline level of capacity presumably reserved for its benefit.

Yet at the same time, PJM rules allowed CSPs to measure and verify the magnitude of DR performance—the value of backed-off capacity—using a guaranteed load drop (GLD) method applied across the entire aggregated portfolio, with all customer PLCs added together. Thus, with the right portfolio, an aggregator could match many “under-performing” customers carrying small loads with a few large customers having abnormally low PLCs (the “over-performers”). And such customers might well be found. A new and growing company, for example, might post a strong increase in consumption from year one to year two, driving its current demand far above its PLC. Another might have cut its year-one demand far below its usual level, simply by trimming its peak consumption during the hottest days of the prior summer—the coincident peak days that determine PLC under the PJM construct—under the common strategy known as peak shaving, employed when possible by large industrial customers, especially those who run batch processes that can be rescheduled, and often for no other reason than to reduce the retail capacity charge on monthly utility bills.

Hess Corp. explained how easy it all was:

“A CSP may contract with a Peak Load Shaver [the over-performing customer] for 20 MW. The PLS has a 10 MW PLC. That CSP then goes out and enrolls 10 1-MW customers … (that is, their PLC is greater than their expected performance). PJM calls an event. The PLS provides a 20-MW load drop. The remaining 1-MW customers do nothing. The CSP gets credit for 20 MW because the Peak Load Shaver covered the remaining customers’ obligations.” (See, Post-Technical Conference Comments of Hess Corp., p.5, FERC Docket EL11-3322, filed August 15, 2011.)

“The result,” wrote PJM and the IMM in their joint statement, “is that the CSP is paid twice for a single load reduction, thus double counting the value of the compliance.”

And not only that: this arbitrage would create the appearance of a greater supply of capacity in the RPM market than would actually exist, threatening reliability, because the over-performer would be earning credit for backing off demand and supposedly freeing up capacity for others—phantom capacity that had never been reserved for it in the first place.

Yet only some CSPs had double-counted, according to PJM’s tariff proposal filed in April.

“Most CSPs, do not engage in this practice,” PJM wrote, “because they believe it is inconsistent” with the rules.

Nevertheless, PJM feared that more aggregators soon would begin using the loophole: “They must do so to compete with the CSPs that already measure compliance in this way.”

Marie Pienizaek, CEO of Energy Curtailment Specialists, testified at the FERC’s July conference that diversity of customer performance was “fundamental” to aggregation. But Viridity Energy, a CSP founded by former PJM exec Audrey Zibelman, complained in comments of “performance alchemy” and favored closing the loophole. It warned that “assembly of a portfolio cannot change the measurement of any customer’s capacity performance.” Comverge, another aggregator, denied recruiting any “ghost customers.”

Just before the conference PJM said it had observed that the offending CSPs “appear to target or knowingly engage in recruiting customers with little or no capacity curtail capability.” (Response of PJM to Notice of Topics for Technical Conference, filed July 11, 2011.)

PJM added later that if all CSPs would have engaged in this double-counting for the 2014/15 delivery year, the region would have fallen 4,320 MW short of capacity, forcing PJM to boost the installed capacity requirement in the capacity auction, upping the capacity price by about $35 per MW-day for most of the PJM footprint, and forcing the region’s ratepayers to pony up an added $1.8 billion each year.

“The maximum capacity that an end user can commit,” PJM asserted, “is the customer’s contribution to the system peak load for which capacity was procured.”

An Old Song

If that was all there was to the story, we could stop now. FERC would OK PJM’s new tariff, closing the loophole and removing the threat to reliability.

But there’s more.

EnerNOC, the demand response aggregator believed to be the primary “double counter” among CSPs, has mounted an aggressive and remarkable fight against PJM’s proposed fix. Aided with intellectual capital from Maine attorney Donald Sipe (Preti Flaherty), EnerNOC has advanced a new and highly controversial theory of how to measure the capacity value of demand response—a theory so enamored of the potential of demand response that that it just might win over such a prominent champion of DR as Chairman Jon Wellinghoff.

Readers might recall that this column last year gave credit to Sipe for turning Wellinghoff’s head in the case that eventually led to FERC Order 745, approving payment of the full locational marginal price for DR sold in energy markets, a decision seen by many as theoretically unsuportable. (See, “Two Hands Clapping,” Fortnightly April 2010)

And in June, in fact, the commission suspended PJM’s tariff proposal, finding that it likely was unjust and unreasonable, and called for a staff-led technical conference on July 29 to delve more deeply. (See, 135 FERC ¶ 61,212, June 3, 2011.)

Sipe’s theory goes against conventional wisdom, treating electric capacity and energy as very nearly the same thing.

At the technical conference in July Sipe sparred with market monitor Joseph Bowring, but first warmed up the crowd with a folksy analogy:

“Energy is kind of like dating, right? We hope you show up at the bar and we hope you’re attractive, and you know, we’ll have a good time when you’re there.

“Capacity is like marriage; you’ve made a commitment. You’ve got to be there the whole time, but essentially, you’ve still got to deliver a charge or you’re not doing your job.”

With this ruse Sipe caught the PJM’s market monitor off guard, with Bowring himself tripping the snare:

“I agree with Don that energy and capacity are clearly distinct and different products. Well, actually I don’t agree because I say they are different products and Don says they’re all the same thing, I guess.”

In PJM’s world, and throughout the power industry for that matter, capacity is thought of generally as a static resource matched up with customer demand, such that the load can assert ownership of that resource, or at least lay a claim that the resource is dedicated to serving that customer. That claim is evidenced by the customer’s obligation to pay for the capacity—or, in the case of demand response, by a promise to curtail consumption traded to gain release from that obligation. The obligation is cemented by PLC, the customer’s prior usage recorded at peak, occurring at the very moment when resources are stretched to the breaking point. If the customer consumes power when capacity is most needed, that proves that the customer “caused” the need and, having caused the need, can now presume to remove that need by backing off.

Bowring elaborated at the conference:

“A customer cannot offer to not pay for a level of capacity for which it has no level of obligation to pay. If you’re not obligated to pay for it, you can’t very well give up the obligation to pay for it. Such an offer would be meaningless and without value.”

That leads directly to PJM’s policy that DR capacity credit can’t exceed the customer’s previously recorded peak load contribution, as Bowring explained:

“PLC is in fact used to define your obligation to pay for capacity …

“So clearly, the only metric by which one can judge how much you’re not paying is the price of that capacity and the amount you otherwise had to pay for. That’s what the PLC is and does.”

Thus, even if a customer engages in peak shaving in year one, causing its PLC to fall far below its normal contemporaneous level of demand and energy consumption (known as the customer baseline or CBL), that customer under PJM rules still could not cut its energy usage to zero, for example, and then claim capacity credit for doing so, because the customer, by peak shaving, has already saved those dollars, and can’t ask to be paid a second time, as Bowring related at the conference:

“What Mr. Sipe would have us believe is that, from a 100-MW unrestricted customer who has historically shaved peak load [to] 20 MW, that represents an opportunity to provide new savings to the system of that 80 MW. Well, it does not. Those 80 MW have already been saved.”

But haven’t we heard that song before?

The student will sense right away that Bowring here is reprising the old argument of “LMP minus G.”

Bowring is echoing the words of Bill Hogan, Robert Borlick and all the others who argued sensibly but unsuccessfully that when a customer sells demand response into the energy market—as distinguished from the capacity market—that customer will have avoided having to pay the cost of the curtailed generation, and so the proper level of compensation for energy DR can’t be LMP, the wholesale market price, but rather, the wholesale market price minus “G,” the generation component of the retail utility rate.

But as we know, and as Sipe reminds us, FERC rejected that advice, and ruled in Order 745 that energy DR can receive full LMP, after the commission found forcing DR providers to forgo the commodity savings (via LMP – G) created barriers and tended in PJM to kill off any meaningful level of DR participation in energy markets. (See, Order 745, March 15, 2011, 134 FERC ¶61,187; see also Post-Technical Conference Comments of EnerNOC, p.35, FERC Docket ER11-3322, filed Aug. 15, 2011.)

And if anyone doubts that the commission would reprise its landmark ruling on energy DR, consider that in calling the technical conference, the FERC staff asked the parties to comment on the following question:

In other words, FERC might give full capacity credit for demand response by a peak-shaving customer, even if, as Bowring put it, the costs of those shaved megawatts “have already been saved.”

Meanwhile, Sipe and EnerNOC propose a dynamic view of capacity DR, not circumscribed by the customer’s prior consumption history, but limited only by present capability.

Citing definitions from FERC, DOE and even NAESB for support, Sipe and EnerNOC assert that capacity is simply the ability to deliver energy upon demand, and so demand response as a capacity resource must be defined in the same dynamic manner:

“The essence of demand response as a [capacity] resource is the ability to move in response to a dispatch instruction.” (See, Protest of EnerNOC, p.17, FERC Docket ER11-3322, filed April 28, 2011.)

It’s the “ability of demand response to affect supply and demand at the margin and is not a function of having purchased a particular amount of entire energy or capacity in advance.”

That means if an industrial customer engages in peak shaving in year one, causing its PLC to fall far below its typical power consumption, then the customer should remain free in year two to earn full DR capacity credit on a promise to back off its entire (and higher) year-two demand, even though that would imply a capacity credit above what the customer might then be paying for capacity in its retail rate.

The customer earns capacity credit on any promised curtailment that it can physically achieve, regardless of how much power it might have used during a peak hour in the prior year.

To further support this dynamic view, Sipe notes that “PJM operators are fully aware” of the discrepancy between a prior commitment and current capability:

“EnerNOC regularly experiences PJM staff calling in advance of anticipated emergency events… The typical exchange requests information regarding how much response PJM is ‘really’ going to get…

“This regularly occurring vignette,” Sipe continues, “illustrates precisely the problem with PJM relying upon static measures of performance: PJM operators cannot dispatch ‘PLC’ during an emergency, and PLC performance measures do not give system operators a means to ascertain resource availability during an emergency.” (See, Post-Technical Conference Comments of EnerNOC, pp.7-8, FERC Docket ER11-3322, filed Aug. 15, 2011.)

In this way Sipe slays the traditional duopoly of capacity and energy:

“There is no way to separate capacity and energy,” he explains.

“As with electrons flowing now or later, ‘capacity service’ and ‘energy service’ differ only in their timing.”

Ski Resorts and Water Parks

Some, including Pieniazek at ECS, and Hess, have suggested solving the problem by calculating PLC on a portfolio-wide basis, eliminating any notion of individual over- or under-performers, but that might take much of the fun out of the aggregation business. Others, such as EPSA’s vice president for regulatory affairs, Nancy Bagot, offered that overperforming customers who wouldn’t qualify for capacity credit under the PLC test should be treated as having sold their demand response into the energy market, and should receive the energy price.

In agreement was Andrew Ott, PJM’s senior vice president for markets, who appeared at the conference and likened the situation to a generator who bids and clears the market to supply 90 MW of capacity, but then runs at 100:

“I’ve got to pay him 100 MW for the energy, but [the] capacity payment remains unchanged.”

So, too, with demand response:

“All I’m going to make sure is they gave me at least what they committed to give me in the capacity market, just like with the generator. If they gave me more, great, they get a payment in energy and everybody’s happy, including us.”

Well, perhaps not everybody.

Constellation and Energy Connect explained in joint written comments why an aggregator would want to design a portfolio to ensure capacity credit, rather than simply accept the energy price. Since you don’t know in advance when a peak emergency will occur, a DR capacity bid gets paid every day, 365 days a year, for having committed to stand ready, whereas a DR customer supplying energy gets paid only for the exact hours of curtailment.

Consider the DR product known as ILR (interruptible load for reliability). Constellation and Energy Connect showed that as the 2011/12 price for ILR came in at $110.04 per MW-day, 10 MW of cleared ILR would earn the customer over $400,000 over the course of a year. (Comments of Demand Response Aggregator Coalition, p.9, n.15, FERC Docket ER11-3322, filed April 28, 2011.)

Selling that same demand response for a few hours into the energy market—even at a typical peak price of $250/MWh—probably wouldn’t return such a high income stream.

In the end, however, the story belongs to EnerNOC and its outspoken advocate.

At the July technical conference, Sipe presented several graphs to illustrate how PJM’s policy could produce absurd results, such as for a ski resort, or amusement park (see Figures 1 and 2).

The ski resort might well likely carry a peak load contribution of zero, as it likely wouldn’t consume energy during the heat of the summer, when the year’s five highest daily peak hours would occur, and so likely wouldn’t be eligible for any DR capacity credit, even if it curtailed snowmaking operations during the winter.

The amusement park, by contrast, would sport a high PLC, but likely would remain too busy during a summer heat emergency to consider selling demand back to the grid.

Another example involved a business with a high energy consumption and PLC in a year one, but which was closed, shuttered and bulldozed in year two, yet remained eligible for capacity credit even as it no longer operated, as its demand fell below its PLC, thus appearing to free up capacity for use by other businesses.

Responding in early September, PJM termed these two examples “extreme,” and “not representative” of the “vast majority” of DR participants in the region, but suggested that the two facilities could submit a “composite bid,” with each in effect acting as the other’s retail aggregator.