by Jeff St. John
June 29, 2018

This month, California’s investor-owned utilities announced the winners of the fourth and final Demand Response Auction Mechanism (DRAM). Since 2015, the pilot program has brought in 717 megawatts of demand response, behind-the-meter batteries, smart electric vehicle chargers, and other forms of distributed energy resource (DER) capacity. 

That’s not a huge amount, compared to the state’s overall demand response portfolio, but it could become a much bigger share in years to come. That’s because DRAM is also meant to serve the role of a test market for demand-side resources — a system where DER owners, operators and aggregators could decide what it was worth to commit to grid services or energy market opportunities, and compete against one another to provide the megawatts to fill those needs. 

Still, it remains up to the California Public Utilities Commission to decide whether DRAM becomes a permanent part of the state’s new customer and DER-focused demand response regime, or ends just like most other pilots. And according to the California Public Utilities Commission, the research on whether DRAM is on track to meet the threshold for permanence still has too many uncertainties to answer that question — at least, not without more study. 

A “mixed and in some cases inconclusive” report on DRAM so far

Last month, CPUC Commissioner Martha Guzman Aceves decided (PDF) to extend the deadline for the CPUC’s broader demand response overhaul, from next month to July 2019, for two main reasons. First, the CPUC still has to untangle some complicated issues within that docket, such as rules for dual participation in different programs, incentives for automated DR and serving locally grid-constrained or economically disadvantaged communities, and other nitty-gritty program design questions.

Second, and more cryptically, Guzman Aceves noted that she has read a preliminary report from CPUC staff on DRAM program performance, one that “indicates issues that are too complex to be addressed in the informal resolution process” that would have otherwise concluded its findings this month. 

The “mixed and in some cases inconclusive” results of this report need instead a formal review, since their findings “may raise matters related to the next steps of the pilot, which may include demand response budgetary implications,” she wrote. 

Specifically, there’s the huge question, currently held by a “$0” placeholder in the three IOUs’ 2019-2022 demand response budgets (PDF), of how many tens or hundreds of millions of dollars might go to DRAM if it’s made permanent. 

The preliminary review referenced by Guzman Aceves hasn’t yet been made public, leaving unclear which parts of have yielded “mixed” or “inconclusive” results. 

As with any program trying to fundamentally alter the way that utilities, customers, grid operators and third-party DER providers do business, there is a long list of potential sources of delay. 

The CPUC staff review in question was launched last year (PDF) to measure DRAM’s performance from 2015 through 2017 against “six specific criteria that shall serve as the objectives that the DRAM must meet in order to expand its role in the resource adequacy market” — in other words, to be made a permanent program.

These six criteria are relatively simple questions, and some have simple answers. The first two — “Were new, viable third-party providers engaged?” and “Were new customers engaged?” — can be answered with a simple yes, given that utilities have announced new contract winners and counted new participating customers in each round. 

But the remaining criteria, though posing simple questions, have very complicated answers, as the eight full pages of metrics and equations laid out in the CPUC’s review plan indicate. 

The challenges of comparing costs of old-school demand response to DRAM’s new DER portfolios

Take the issue of the third and fourth questions: “Were bid prices competitive?” and “Were offer prices competitive in the wholesale markets?” To start, these answers are virtually impossible for outsiders to calculate, since the CPUC has kept the dollar figures for competing and winning bids strictly confidential. 

The only broad trend to be gathered is that prices for DRAM offers are coming down, since utilities have increased the number of megawatts procured in each auction, while keeping their spending capped between minimums and maximums set by their budgets

But even parties with the data have had trouble making the appropriate comparisons to existing programs to determine which are more cost-effective. 

For example, Southern California Edison’s filing from last month — by far the most detailed of the three IOUs — noted that its 2019 DRAM offers “seem on average more cost-effective” compared to existing demand response programs. 

But it also threw in many caveats to making “definitive benchmarking conclusions,” such as the fact that most DR is year-round while DRAM bidders choose the months and amounts per month they’re bidding; that DRAM’s non-performance penalties aren’t as strict as many other DR programs; and that DRAM bids don’t include utility overhead costs included in DR costs. 

There are plenty of other differences between old-school demand response and what DRAM is doing that make direct comparisons difficult, GTM Research Grid Edge Analyst Elta Kolo noted. 

"Through DRAM, you have behind-the-meter energy storage, EV charging, behavioral DR, and combinations of resources optimized behind the meter,” she said. “There is a learning curve associated with this for the CPUC, the utilities and vendors that have traditionally partnered in the utility programs.” 

DRAM itself has changed from year to year, making comparisons between years difficult, Kolo noted. Last month’s auction was only added to the schedule by the CPUC in October.

One of DRAM’s most problematic design elements is its pay-as-bid structure, she added. DRAM pays each bidder exactly the price they bid, instead of paying all of them a uniform price based on the highest bid to clear the market, as many energy markets do. 

Pay-as-bid structures do mitigate against the threat that vendors will underprice their resources just to win bids, in hopes that another higher, yet still market-clearing, bid will allow them to earn their money back. This was a potential problem that some DRAM critics warned could undermine the entire price-discovery purpose of the pilot.

But pay-as-bid structures also leave bidders very little opportunity to earn anything for being less expensive — that is, more efficient — than their market-clearing competitors. And as markets mature, they end up becoming a problem for price discovery, given that their nature requires bids to remain secret even after auctions are cleared. 

“For the third-party community to gain confidence in DRAM, there needs to be more price transparency,” she said. “An evaluation of the pay-as-bid approach will be revealing of profit margins and opportunity.”

Measuring the reliability of an emerging DER asset class 

SCE’s benchmarking section also noted that the most recent offers are “generally more competitive in terms of pricing” than offers in the previous auction; however, “it is still unknown whether the Sellers can fully deliver the products at the contracted prices.” 

That brings up the fifth and sixth criteria on the CPUC’s review, which are focused on how reliably DRAM’s bids have been converted into real-world action. 

The first of these asks, “Did demand response providers aggregate the capacity they contracted, or replace it with demand response from another source in a timely manner?” The dual nature of this question underscores the fact that some larger demand response providers such as EnerNOC and CPower shifted existing assets into DRAM. But still, it’s aimed at measuring the first bar for entry for any DRAM participant, which is getting the customers they’ve promised. 

The second asks if these resources were “reliable when dispatched, i.e., did customers perform appropriately?” In simple terms, it’s asking for data on how well DRAM participants have delivered on their pledges to reduce load for up to four hours per day, up to three days in a row during the months they’ve contracted, both to help utilities meet their resource adequacy requirements, and to bid into the energy markets of state grid operator CAISO. 

Utilities and the CPUC have not shared this kind of performance or reliability data for DRAM participants in the 2016 and 2017 seasons, making it difficult to know if any have failed to perform on either of these two measures. 

The recent DRAM filings do mention some clues as to how utilities are addressing companies that fail to perform. For example, PG&E described how its most recent auction included “negative factors” it added to the price of bids from certain companies. 

A 15 percent cost increase was tacked on to companies that had “willfully terminated or defaulted on a past DRAM purchase agreement or submitted offers that demonstrated bidding behavior providing clear evidence of market manipulation or collusion” — two categories of behavior that could be detrimental to a functioning market. Meanwhile, a 5 percent increase was assessed to those that had either failed to deliver at least 50 percent of the capacity contracted for in previous awards, or for those that had rejected shortlist offers in the past. 

The list of companies involved in the DRAM pilot has changed over the four rounds of auctions, with only a handful participating in all of them. It’s possible that some of the companies that won early rounds, then did not continue their participation because they failed to perform on some of these measures. But it’s equally as possible that companies chose not to bid at all into later rounds, or placed legitimate bids that failed to clear against the increasingly lower-cost bids of their competitors. 

Most of the companies winning contracts in the auction have an established track record in demand response, such as EnerNOC and CPower. Others are behind-the-meter energy storage providers that have already won significant utility contracts in California, such as Tesla, Stem, Advanced Microgrid Solutions, and Engie Storage (formerly Green Charge Networks.)

The DRAM pilot faces a challenge in ensuring reliable delivery while also allowing novel forms of DER capacity and business models to participate, however. It’s noteworthy that over the course of the past two DRAM auctions, the largest shares of megawatts have gone to two companies with no prior track record in utility demand response or DER programs: OhmConnect and Leap. 

OhmConnect has built its residential, behavioral-based demand response portfolio by paying homeowners for reducing energy use independent of any market mechanism to recoup that payment, then bidding that capacity into programs like DRAM to pay for it. This model requires some faith in the company’s ability to deliver on its promises, but it has also helped bootstrap a business that requires no investment from utilities or traditional aggregators to deliver valuable capacity.  

Fellow San Francisco-based startup Leap has even less of a track record. It plans to help existing DR aggregators tap the latent ability of devices inside commercial-industrial buildings to bid their available load reduction capacity in response to automated energy pricing signals. It won a combined 90 megawatts of capacity in the just-completed final DRAM auction, and will be on the hook financially to deliver them to market next year.