'The future grid': How one DOE program is pushing the boundaries of aggregated DERs

ARPA-E is funding research aimed at fuller coordination of distributed resources on the grid

Since it was first funded by the Obama administration's stimulus act in 2009, the Department of Energy's advanced technology research arm has doled out millions to support pathbreaking research in renewable energy, storage and other distributed resources. 

Now, as utilities become increasingly comfortable with new distributed technologies, the agency is turning its sights to research on how to make all the new technologies on the grid operate as one, providing new aggregated resource techniques to utilities and grid operators. 

The DOE's branch, Advanced Research Projects Agency-Energy (ARPA-E), announced grants in December totaling $33 million for 12 projects aimed at creating automated systems to operate DERs through its Network Optimized Distributed Energy Systems (NODES) program.  

“ARPA-E is trying to do things off the beaten path that others aren’t looking at,” said Eric Rohlfing, the deputy director for technology. “The U.S. electric grid is complicated and uncertain and we are looking at technologies that will enable flexibility to reduce some of the uncertainty.”

NODES specifically looks to fund teams of experts in power systems, control systems, computer science and distributed systems aimed at improving grid efficiency, slash carbon emissions and lower energy-delivery line losses. 

Ultimately, NODES is about "the future grid,” Rohlfing said. The program is seeking algorithms and advanced control technologies with dynamic capabilities to help the grid evolve into a more automated system, similar to the analogy of an auto-pilot for the grid. 

“The pilot only controls the control system," Rohlfing said. "The auto-pilot flies the plane.” 

From the grid operator’s point of view, aggregated, automated DERs will look like fast response reserves, ramping reserves, or day-ahead planning reserves and provide a specific and reliable response.​

But like many technological endeavors, only one or two projects will be homeruns that develop into commercial successes. With the rest, “we may learn a lot about what doesn’t work," Rohlfing said. 

Here's a rundown of three of the NODES project ARPA-E funded in its last round. A full list is available here.



Packets of innovation

One interesting DER project funded as part of NODES is a $1.54 million grant to the University of Vermont (UVM) that will test Packetized Energy Management (PEM).  

The project “builds on approaches used to manage data in communication networks without centralized control,” ARPA-E explained, and aims to aggregate “millions of small end-use devices” to balance grid supply-demand fluctuations from the increased penetration of renewables.

In Vermont, Green Mountain Power continues to grow solar on its distribution grid, said Mads Almassalkhi, an electrical engineering professor at the University of Vermont in Burlington.

“Our proposal borrows from the way the internet operates to address the increasing variability," he said. 

A utility or system operator needs the grid to respond to supply-demand fluctuations, and DERs provide very useful loads for that purpose.

“We want to, in a fair, private, autonomous, and scalable way, leverage resources like water heaters, to provide flexibility to the grid without inconveniencing consumers,” Almassalkhi said. “This requires load coordination or, as Green Mountain Power calls it, load choreography.”

For example, a packetized energy technology device can be added to a consumer’s water heater and would “ping the grid operator or an aggregator or coordinator, asking every five minutes if it can heat the tank’s water,” Almassalkhi said. “The answer will be ‘yes’ or ‘no.’ The consumer does not need to do anything or even be home.”

If hundreds or thousands of water heaters are asking and if heating is aligned with the grid’s supply-demand needs, the water heaters become “a virtual power plant,” he said.

To ramp load down, the grid operator will increase ‘no’ answers to the water heater pings. And conversly, the grid operator would boost "yes" answers to use excess supply from a sudden spike in generation.

PEM will also be used with other DERs, including air conditioners and electric vehicle chargers. It will also be designed so that water heating, air conditioning, or EV charging would be varied only within parameters acceptable to consumers. There will only be pings when the water or room temperature or the EV is ready for use, Almassalkhi said. 

Traditional demand response providers like EnerNOC use “direct control” and work “top-down” from a group of contracted loads, he said. “We use a bottom up approach. Each device uniquely requests access to the grid, given its own needs.”

The PEM plan is being tested in collaboration with Green Mountain Power (GMP), Vermont’s dominant electricity provider, and Vermont Electric Company (VELCO), the state’s transmission operator.

After extensive mathematical modeling and multiple, increasingly larger-scale computer simulations, “we will implement the project on a GMP feeder with about 500 buses to demonstrate the concept’s viability.”

The goal is to be able to aggregate enough DERs to offer the grid operator a virtual power plant that balances the variability of renewables without the use of large thermal generators, Almassalkhi said.

“Officially and technically, we are providing regulating reserves,” he added. “The goal is to be able within five minutes to supply at least 5% of the load as a reserve.”

The limiting factor for the virtual power plant could be the bandwidth of the communication network, Almassalkhi said. 

“Ideally, the pinging would be in real time, but in reality we are looking at ten minutes to fifteen minutes today," he said. "As communications systems improve, we hope to push toward one minute.”

How PEM programs will be valued in the market remains unknown, though Almassalkhi expects they would be treated like any other ancillary resource. 

“We expect markets would allow the system operator and aggregators to earn back their costs because this is basically an ancillary regulating reserve service,” Almassalkhi speculated.

But the immediate ARPA-E goals are technology driven: Can it work on a small scale and then on a real scale? “Once it is working, we will do simulations to see what the benefits are to the grid,” he said.

Synthetic Regulating Reserves

A $2.34 million grant was awarded to the University of California at San Diego (UCSD), in partnership with the University of Illinois and others, that will attempt to create synthetic regulating reserves for the grid operator. Here again, algorithms and software will be key to how energy aggregators can leverage customer-sited resources to respond to incoming requests from grid operators.

Distributed coordination algorithms will provide a platform so that aggregators can “quantify reserves, system objectives and constraints, customer usage patterns, and generation forecasts,” ARPA-E explains. With this information, they will be able to “rapidly respond to operators while considering network constraints and quality of services for customers.”

“We would like to harness the power and flexibility of DERs,” said Sonia Martinez, a UCSD Professor of Engineering.

There are many different approaches to meeting the challenge of reliably answering an energy services request issued by a system operator, she acknowledged. “The systems are complex, but we want reductions that are structurally similar so they can be worked with easily.”

“At the heart of the undertaking is a ConText engine software tool being developed by UCSD Professor Tajana Rosing and University of Illinois Professor Alejandro Dominguez-Garcia and IBM is collaborating,” Martinez said. “It will formulate the abstractions to determine what a specific building or DER aggregation is capable of delivering.”

Three levels of communication take place in the process of using DERs to meet system needs, Martinez said.

First, grid operators ask aggregators if there are sufficient reserves to regulate frequency. Aggregators must then respond, but but they must first communicate with consumers who own DERs to know in real time what resources are available.

“Direct control allows the consumer no freedom,” Martinez said. But dialogue between the aggregator and each customer requires a large volume of communications. That can cause traffic problems and bottlenecks.

“Our concept is to manage those communications in a simplified way that reduces bottlenecks but still gives consumers control,” she said.

There is another level of communication that could provide more resources for grid operators, she added. “We envision a group of aggregators that interact in response to the grid operator’s request. If one aggregator does not have currently available resources and another does, the resources could be shifted between aggregators.”

To meet ARPA-E-set standards, UCSD’s solution also has to be fast, Martinez said.

“The first response has to happen within five seconds, the delivery has to ramp up to full capacity within five minutes, and the aggregator has to provide the full capacity need within plus-or-minus 5% for at least 30 minutes.”

The researchers have not determined the DER mix for the project but it has to be "a heterogeneous collection," Martinez said. It will likely include resources from buildings on the UCSD campus, electric vehicles, rooftop solar, and batteries.

The project will begin with mathematical analyses applied to increasingly larger-scale simulations that model larger loads and more DERs. “The ConText engine software tool will start with a specific building. Other DERs will be added one at a time and eventually simulations will deal with the heterogeneous portfolio,” Martinez said.

The project will conclude with a demonstration, probably on UCSD’s microgrid, which is part of the San Diego Gas and Electric (SDG&E) system. The demonstration will not include the utility, but the researchers will have access to data from project partners Ameren and SDG&E for the simulation phase.

“Traditional central generation has provided regulating reserves with fossil fuel-driven rotating mass that is costly because it is not efficient, produces pollution and greenhouse gas emissions, and has transmission losses,” Martinez said. If DERs can be harnessed, they can eliminate these costs.

But the challenge is how to satisfy all participants, she said. System reliability must be sustained by meeting the grid operator’s request for reserves. Aggregators must be able to fulfill their contracts. And consumers must be able to market their DERs without having their lives disrupted

Incentive-based aggregation and control

A $2.7 million grant to the Pacific Northwest National Laboratory (PNNL) will develop a plan to increase grid flexibility with aggregated DERs and attract participants with incentives.

The Multi-Scale Incentive-Based Control of Distributed Assets aims to supply reserves to the grid through “incentive-based control strategies.” By providing incentives, which have not yet been conclusively identified, it will “acquire flexible assets that provide services” and use “faster device-level controls that use minimal communication to provide desired responses to the grid.”

Like the other projects, PNNL will test a bottom-up approach. A distribution reliability coordinator — typically the utility — would be the interface between the aggregated DERs and the grid operator’s need for frequency response, frequency regulation, and ramping services

PNNL will begin by quantifying the flexibility available from the various DERs, said Karanjit Kalsi, PNNL senior research engineer. The aim is metrics that describe a DER mix as “a virtual battery with simple parameters like charge-discharge rates, energy capacity, and state-of-charge.”

With such metrics, it becomes possible to provide a specified amount of flexibility to the system operator, he said. The next step is setting up control strategies on the DER level and at the utility level.

“Once we can quantify the flexibility, we want to know how to control and coordinate it,” Kalsi said. “We acquire or engage the assets on a slower time scale and then control them on a much faster time scale and with minimal communications.”

Controlling and coordinating the assets will be done with a combination of sensors and control devices at the DER level. Some will be dynamic, and some will have set points.

“Frequency control has to be purely autonomous through local controls because it requires an instant response,” Kalsi said.

Frequency regulation will require a hierarchical approach, with both device level autonomy and a supervisory capability at the utility supervisor level.

“We don’t want all these devices responding in an uncoordinated way," Kalsi said. "We need to maintain some level of autonomy but we also need to be sure we get a smooth, stable, and predictable system response.”

Step three in the project will simulate the system at scale and combine transmission, distribution, markets, and communication systems to evaluate conclusions reached in the quantification and design stages.

In this step, PNNL will establish communications protocols and “look at communication network impacts like time delays and packet grabs,” Kalsi said. “We promised ARPA-E 100,000 simulated devices of different kinds.”

The simulation will be built resource by resource. “When we know how to control 10,000 HVACs and 10,000 water heaters, we will add a commercial building and then we will add a battery system.”

The challenge with controlling and coordinating that many devices is getting them to respond quickly enough to provide frequency response while managing the same set of DERs at the slower response times used in providing frequency regulation and ramping.

“We have a control and coordination approach for each individually and we have done that in demonstration projects but putting them together is not an easy thing,” Kalsi said.

Finally, the researchers will test actual devices in a demonstration across the PNNL campus and the campuses of project partners United Technologies Research Center and Southern California Edison (SCE).

“ARPA-E requires a demonstration including at least 100 real devices,” Kalsi said. “We will integrate the software we design into the GE/Alstom DMS and use it to operate SCE’s distribution feeders to test the control strategies we develop.”

By mid-2017, Kalsi expects this work to produce a very simple way for utilities and grid operators to understand the flexibility DERs offer along with an app that will integrate into their systems allowing them to monitor and visualize DERs.

“They will have control strategies for aggregated DERs, implemented as software applications, and they will have is a clear demonstration of how this can be done with 100 devices.”

Addressing uncertainty

For all the talk about the grid of the future, no one knows what the U.S. electricity delivery system will eventually look like, ARPA-E's Rohlfing said. It could be a series of interconnected microgrids, a hybrid of today's mostly top-down grid structure, or something else entirely. 

That uncertainty surrounding the future design of the grid is difficult for utilities, but whatever it is, the demand for services from aggregated resources is likely to grow.

“Utilities are not happy with that uncertainty because it is hard to build a business model with it,” he said. “These NODES programs are to address that uncertainty.”