Within a decade, there may be more distributed energy resources (DER) coming onto distribution systems than any utility control room can manage.
An autonomous energy grid (AEG) could optimize those high levels of DER for the benefit of power system and DER owners, research under development by the National Renewable Energy Laboratory (NREL) shows. But if this groundbreaking system autonomy proves elusive, utilities could face voltage and frequency fluctuations, potential supply-demand imbalances or even outages, according to distribution system experts.
"Our Autonomous Energy Grid (AEG) concept is about controlling hundreds of millions of different kinds of devices in real time on a second-by-second basis," DOE Power Systems Engineering Center Director Benjamin Kroposki told Utility Dive. "Over time, we will demonstrate a new paradigm for how the grid of the future can work with the massive deployment of DER devices that is coming."
A successful AEG concept would require greater technical precision than autonomous driving and the Internet, the two most comparable examples in terms of data management, data analyst specialists and utility system authorities told Utility Dive.
But the expected massive growth of DER makes NREL's ambition necessary, according to Kroposki.
The autonomous grid
Residential solar installations are expected to grow approximately 8% annually through 2050, and behind-the-meter storage deployments are anticipated to hit almost 1.9 GW by 2024, and current forecasts project around 18.7 million EVs on U.S. roads in 2030.
It is not unreasonable to imagine electricity customers a decade from now having up to five devices at a time — a rooftop solar system, a home battery, a smart thermostat, a smart water heater and an EV charger, said Kroposki. By that math, the 4 million customers in the San Francisco Bay area could leave PG&E with 20 million devices to manage.
Utilities will also see rising penetrations of bulk system wind and solar generation that will create supply-demand imbalances that traditional control centers will not be able to manage simply by ramping supply up or down, he said. Instead, it will require managing demand, which could be done through DER technologies. But the sheer volume of DER could exceed a utility's ability to optimize.
A comparable challenge is managing the Internet's hundreds of millions of data points, but the power system is under higher pressure to maintain precise moment-by-moment supply-demand balance and avoid any delays, he added.
"We want to minimize the need for communication between the control center and DER devices."
![](https://d12v9rtnomnebu.cloudfront.net/diveimages/corporate_site/teampage/square_profiles/placeholder-200.png)
Colton Ching
Senior Vice President Planning & Technology, HECO
The basic element of NREL's theoretical AEG architecture is the optimization and control of a "cell," which can be a home or building energy management system and their controllable devices.
Kroposki describes the AEG as "distributed cells with a hierarchical, scalable, reconfigurable and self-organizing control structure on top of them."
The next level up may be the distribution circuit, and the level above that might be a substation, Kroposki said. "Each level's cells have parameters and constraints, like voltage, currents or system pricing, that they use to self-optimize. The point is to minimize the amount of information passed between levels and maximize self-optimizing at each level."
Researchers have constructed algorithms that can "optimize and control hundreds of millions of distributed technologies in real time," he said. Holy Cross Energy (HCE), a 60,000 customer rural cooperative in Colorado, will be the first field demonstration of those algorithms.
Early insights into the potential of automation came during NREL's work to integrate high levels of residential solar into the Hawaiian Electric Companies' (HECO) distribution system. In pursuit of better responsiveness to grid conditions from the inverters on customers' rooftop solar systems, HECO system specialists and their NREL consultants realized they were building in autonomous capabilities.
![](/user_media/diveimage/UD-AEG-02-08-21-2019.jpg)
Utilities leading the way
"In developing early smart inverters, we found they could be pre-programmed to certain responses, reducing the need for control room signals," Senior Vice President Planning & Technology Colton Ching told Utility Dive. "We didn't realize we were building autonomous capabilities."
As HECO's grid modernization planning advanced, "we realized the communications network for traditional central command and control of the level of anticipated DER would be cost-prohibitive," Ching said.
HECO's 350,000 customers have installed devices with "about 250,000 inverters," he said. The utility's plan to achieve Hawaii's mandated 100% renewables by 2045 includes tripling distributed solar. "That would be about 750,000 inverters and, with battery systems, electric vehicles, and other DER, the number quickly multiplies."
HECO set out to "push autonomous control down to the local level as much as possible," Ching said. "We want to minimize the need for communication between the control center and DER devices."
A completely autonomous grid is beyond reach today because of technology limitations, he said. But NREL's AEG allows "prioritizing controls and focusing on critical operations rather than trying to manage individual devices," he added. Narrowing supervision to "the critical subset of operations" is "the only way we can have an efficient, cost-effective distribution operation" in a high DER future.
[T]he utility control room "will be more like air traffic control, guiding the autonomous activities of cells to maximize their individual value."
![](https://d12v9rtnomnebu.cloudfront.net/diveimages/corporate_site/teampage/square_profiles/placeholder-200.png)
Bryan Hannegan
President and CEO, HCE
HCE has taken on the first field test of the AEG to better understand what is possible with automation, President and CEO Bryan Hannegan told Utility Dive. It is adding 10 to 15 rooftop solar systems per week, has added over 2 MW of solar annually for several years, and is committed to adding 2 MW per year to its 150 MW summer peaking system through 2030.
"We want to understand what values DER can provide to consumers and the grid," Hannegan said. The resilience value in cells' ability to island and continue serving their owners and the system became especially important after a 2018 wildfire in its territory came "one burning pole away from putting Aspen in the dark for a week."
HCE and NREL are partnering with other local groups to build affordable, zero-net-energy housing for local teachers. Four of the houses on an HCE feeder are now occupied and equipped with "solar PV, a battery, a level 2 EV charger, a heat pump water heater and an air source heat pump," Hannegan said. Each DER has a controller with autonomous capabilities.
The newly initiated field test models DER performance under utility control and when acting autonomously, he added. NREL will review the performance of the four homes functioning together as a cell, or fundamental autonomous unit, and of each individual home functioning individually as an autonomous cell.
Each cell will be tested when it is connected and when it is not connected to the HCE control room, and be reviewed to examine just how autonomously a single home as well as a set of homes can function.
The AEG does not eliminate the need for the distribution utility "to maintain the poles, wires, transformers and switches, because they make the DER more valuable," Hannegan said. But the utility control room "will be more like air traffic control, guiding the autonomous activities of cells to maximize their individual value."
"Given the needed standards and regulations, this is easily 10 years away. Over a third of utilities have still not deployed advanced metering and AEG is far beyond that."
![](https://d12v9rtnomnebu.cloudfront.net/diveimages/corporate_site/teampage/square_profiles/placeholder-200.png)
Tom Bialek
Chief Engineer, SDG&E
No results or conclusions from the field test are publicly available, Hannegan said. "The net impact of the project so far is to give us confidence that we can manage and control DER in a way that gives value to the grid and the consumer from our control room system or in a quasi-autonomous way."
Automation will be critical to the future grid, San Diego Gas and Electric Chief Engineer Tom Bialek, who chairs NREL's electricity systems integration technical review panel, told Utility Dive. But "sub-second control of a system with a high DER penetration from a control center is almost impossible. Only so much can be done that fast."
NREL faces two barriers, Bialek said. "The biggest challenge is technology. Utilities do not have the computing power to run the software that makes this work, and most devices are not sophisticated enough to work with that much computing power."
The second barrier is "the absence of technical standards," he said. "Utilities operate in a highly regulated environment. Given the needed standards and regulations, this is easily 10 years away. Over a third of utilities have still not deployed advanced metering, and AEG is far beyond that."
How far can autonomy go?
The more data autonomous operations have, "the more fine-tuned the algorithm is," Internet of Things Analyst Michael Kanellos of data and analytics specialist OSISoft told Utility Dive.
"A fossil fuel plant generates about 10,000 data streams per MW, but a wind project's turbines generate about 51,000 data streams per MW and a solar project's panels and electronics generate almost 436,000 data streams per MW," Kanellos said.
Wind and solar project operators are taking advantage of new operating capabilities from more sophisticated algorithms based on that increased data to improve performance and lower costs, he added.
Arizona Public Service requires no more than 10 technicians to service over 170 MW of solar because data feeds autonomously prioritize and schedule maintenance. Xcel Energy Colorado's sensor network does 15-minute interval forecasting that informs autonomous predictive maintenance, which saved it $46 million between 2012 and 2017.
"Full automation will never happen because humans make better judgment calls than robots," Kanellos said. "'Artificial intelligence' is morphing into 'decision support,' in which the algorithm lays out possibilities, but a human decides. And the coming flood of data will make decision support algorithms much better."
California is also working on a groundbreaking layered system architecture that is somewhat comparable to NREL's architecture.
NREL acknowledges its AEG "has technical areas that need a lot more research," electricity system consultant Lorenzo Kristov, who is working on the California project, told Utility Dive. Layered architecture is within the present capabilities of a system operator and addresses DER growth until the AEG technical challenges are solved, he said.
Kristov's layered architecture relies on "some of the same optimization mathematics as the AEG" to allocate transmission operations to the ISO and distribution operations to a distribution system operator (DSO), he said. "But the AEG is probably 10 years to 15 years in the future."
In Kristov's plan, DER are managed at the house or building level by DER owners and home energy management systems. The DSO manages only the cumulative system impacts of the total DER "delivered to the interface where the house or building interconnects," he said. "The DER devices do not need control by the DSO, whether they are consuming or delivering generation."
"A utility system is a lot more than driving a car, it is like organizing every part of every car on the road."
![](https://d12v9rtnomnebu.cloudfront.net/diveimages/corporate_site/teampage/square_profiles/placeholder-200.png)
Michael Kanellos
Internet of Things Analyst, OSISoft
At the transmission-distribution interface, "the ISO similarly sees each DSO as one resource," he added. "Each layer does not need to see what's inside the lower layers, it just needs to coordinate what is at those interfaces."
There is little disagreement about the present limits of autonomy, but there are differing opinions over the level of autonomy that will be possible in a future grid.
"Decision support" is "a good way to describe NREL's hierarchy," HECO's Ching said. "Devices can operate autonomously, but they may require a central signal for things that affect the entire system. Balancing supply and demand locally may first require seeing system needs."
NREL has compared the AEG to autonomous driving, but "it is not," Kanellos said. "A utility system is a lot more than driving a car, it is like organizing every part of every car on the road. An algorithm complex enough to ensure safety across that many parameters is beyond reach right now."
Reliability remains an important question, but the AEG "is like autonomous vehicles because both are on the street and more are coming," Kroposki said. "We will develop the smarts and work out the kinks to improve autonomous decision-making and there will be more automation built into the grid in secure and optimal ways."