Can Auto DR Meet the Needs of Renewables?

LBNL lays out the needs and challenges facing the demand response industry.

About a year and a half ago, we wrote about the challenges of integrating renewables with demand response. It’s still tough stuff, but new research is digging in and providing a roadmap on how to get there.

States with high renewable portfolio standards -- and California in particular -- are looking at how to balance all of their new, intermittent wind and solar power. One study found that California would need up to 5,000 megawatts of regulation and other ancillary services to meet its goal of 33 percent renewables by 2020.

Additional ancillary services could be met by traditional generation -- but there will be a need for other options, especially cost-effective ones like demand response, the art of turning down power use to balance the grid. That's cheap -- about one-tenth the installed cost of grid-scale battery storage, for instance, and pretty competitive with its natural competitor, natural-gas-fired peaker plants.

But demand response comes with its own set of problems. A recent study by Lawrence Berkeley National Laboratory looks at some of the challenges -- and promises -- of integrating commercial demand response with grid operations. Although the cost is attractive, there are some major limitations today, including communications, latency and market design, according to Sila Kiliccote, head of LBNL's commercial-sector demand response projects.

Let’s start with the good news. The cost of telemetry is already dropping. When Berkeley Lab first looked at the technology and communications for demand response to play in ancillary services, it cost $70,000 per site. A few years later, that was down to $7,000. But really, that needs to be at about $500 per site, said Kiliccote. The move to 4G and LTE networks will help, but there also needs to be lower cost telemetry equipment, she said.

And then there’s a little issue of control. The two ancillary markets that the study looked at were two-hour shed and 20-minute shed, both of which are faster than the usual day-ahead calls in the industry, but will need increased resources with more renewables. Automated demand response, or Auto DR, could provide help here. The two-hour product has a ramp time of 15 minutes, while the 20-minute has a ramp time of 5 minutes. The study did not look at AutoDR providing spinning reserves, which have to respond in just seconds, although Kiliccote said it’s technically feasible.

But to make fast demand response work, grid operators have to know the load is responding, instead of doing measurements a month later to see what happened. Berkeley Lab is currently working on studies that prove this feedback loop can be completed. It’s still early days, but it can work, Kiliccote said.

The analysis found that for these sectors, Auto DR could provide between 0.18 and 0.9 gigawatts of capacity, and with “modest investments” to upgrade and expand Auto DR, that figure could double to nearly 2 GW.

Kiliccote said the study focused on commercial buildings and data centers that already had building management systems -- in other words, a reasonable subsection of the commercial building stock that could turn down lots of power loads without adding lots of expensive new controls. Residential DR was left out of the study altogether, although SoCal Edison has done pilots to aggregate residential load for ancillary services.

Most of the kilowatts being shed would come from HVAC and lighting, according to the study. In the case of HVAC, the shorter duration of the load shed means that buildings could be more prone to participate because there’s less of a chance of discomfort -- if your HVAC compressor only shuts off for 20 minutes, your office will probably stay the same temperature.

Some critics question Auto DR because of demand response fatigue, where participants don’t want to be called on again and again to shed load throughout the summer. This gives rise to the question, what if you’re being called on all year long for ancillary services?

“Automation is key for ancillary services,” said Kiliccote. “There’s also the value creation for customers and the role of education.” In other words, it has to pay enough to be worth it and it has to be seamless. Also, a 30-minute event might not be seen as the same inconvenience as multi-hour event.

That’s possible, according to Berkeley Lab’s research. Small and large offices, retail stores, grocery stores, schools and hospitals could all double their participation in the markets, based on the data about technical potential to receive and respond to signals.

Of course, the lights might already be off in an office building at night, when wind usually blows strongest. In that case, refrigerated warehouses and data centers could also provide ancillary services. Unlike most of the demand response today, which is mostly large loads from single sources, the patchwork of ancillary services would be aggregated loads from within one location (lights dimmed just a little, HVAC cycled off shortly), as well as aggregation across various sites.

Even with technical feasibility checked off the list, there are more challenges to fast demand response, including forecasting what’s available. Kiliccote described commercial building load as “erratic," though more sophisticated analytics could help on that front.

On the utility and ISO side, there is also an issue of localization. Ideally, when the clouds clear and solar arrays in one neighborhood suddenly start pumping power onto the grid, the balancing services would also come from the same area. Currently, the markets and the grid are not built for that sort of localized response.

“In PJM, there’s not a surgical removal of load on the grid,” said Kiliccote, referencing PJM because it is the one ISO who is further along than others in testing these markets. She noted that a lot of work is being done all over, however, and this could be worked out in the next two to three years.

Kiliccote’s outlook could come across as bullish given the market, technical and regulatory challenges. “We see a lot of customers that are being cautious about when they’re using energy and they’re installing energy information systems,” she countered. Many ESCOs are also helping businesses traverse the demand response landscape to raise money to fund retrofits. “A lot of chains are also paying attention,” said Kiliccote. “There is a growing awareness of how much energy is being used and when it’s being used.”

Another development that could help bring DR into ancillary markets is the OpenADR standard, which is a communications standard for automated demand response. Currently, it is just a fairly crude communications standard, but the OpenADR Alliance is working on a standard that would allow for DR to participate in these sorts of markets. 

Surgical demand response is already happening at a variety of places. Powerit Solutions provides cloud-based, automated demand response to industrial sites so that they can seamlessly drop load without completely shutting down.

Marriott hotel chains have teamed up with Constellation to leverage smart building systems to play in energy markets across the country. Startups Viridity Energy and Enbala Power Networks are part of a PJM program to let small-scale projects sell their power into frequency regulation markets. EnerNOC is using DR to manage mind in Bonneville Power Administration’s territory and has said the resource can respond in near-real time.

Even if those are one-off examples, every ISO is looking hard at this issue. On the other side, there is still a long way to go with customer engagement, said Kiliccote. Make no mistake about it: these are complicated, constantly evolving markets. But where there’s money to be made without too much downside, industrial and commercial companies will likely participate. “We wanted customers to be able to participate because it’s the right thing to do,” said Kiliccote, “but also because it creates value for them.”