Say hello to tomorrow’s smart electric meter, literally.
You can ask some next-generation home energy hubs questions, just like you do Alexa or Siri.
Some devices, arriving this year, will display real-time simulations — vibrant as a video game — to show how you can lower your energy bill or reduce your carbon footprint. They’ll help manage power flows for solar panels and electric vehicle chargers.
Like smartphones, they’ll run apps. And they’ll give utilities a clearer picture of the grid, including warnings if a component on a nearby powerline could fail soon.
A Tale of Two Innovators
Devices from startup Anuranet will switch on in hundreds, perhaps thousands, of homes starting this fall using the NVIDIA Jetson edge AI platform.
Utilidata, a Rhode Island-based company developing grid software for a decade, is working with NVIDIA to develop a Jetson-based smart grid chip. It will be used in meters to bring real-time AI applications to the edge.
The efforts are part of a vision for a smarter, cleaner, safer grid that NVIDIA and partners are helping make a reality.
“Today’s smart meters are not very smart, but with the edge-computing power of the Jetson GPU and our software platform we will create truly intelligent meters that will change the energy-control network one home and one building at a time,” said Diane Zuckerman (pictured above), co-founder of Austin, Texas-based Anuranet, the latest of a half-dozen startups she has launched worldwide.
Anuranet is developing the Bullfrog brand of smart meters and circuit panels powered by the NVIDIA Jetson edge AI platform. They can connect to the grid as well as smart appliances, home energy hubs, solar panels, electric-vehicle chargers and their batteries to help customers save on energy bills while decreasing their carbon footprint in real time.
“Their simplicity will engage consumers, creating value with high performance and security,” Zuckerman said.
Although it’s just eight months old, the startup has signed contracts with two companies that will start deploying its devices this year in new homes.
Richland Partners, in Nashville, will use the Anuranet Bullfrog energy ecosystem in homes and apartments it’s building. Neu Communities in Austin will use them to manage private microgrids in new residential neighborhoods it’s constructing.
Jetson Reshapes Edge Computing
Today’s smart meters throw away most of the energy-related data in the home because they lack the compute muscle to process it. It’s a treasure trove of real-time data that Anuranet, Utilidata and others will use to improve the grid.
For example, NVIDIA Jetson-powered smart meters can take measurements tens of thousands of times a second.
“Encoded in that data is information about the quality of the power flow and what might be interfering with it,” explained Marissa Hummon, CTO of Utilidata, an NVIDIA Inception partner that’s already built applications for some of the world’s largest metering companies.
“A tree branch might impact the wave form and if we can see that we can predict an outage — that kind of insight is something lots of apps can use, so it’s creating an environment for others to innovate,” she said.
Deploying the Best Algorithms
Backed by investments from Microsoft and NVIDIA, Utilidata aims to understand power flows in ways that benefit both consumers and utilities. That requires machine learning and the compute horsepower to run it.
“No human can look at all these patterns to find the key insights, but computers can, and our software running on the Jetson-based smart grid chip will let us deploy the best algorithms,” added Hummon.
Utilidata and the U.S. National Renewable Energy Lab are collaborating to run the lab’s real-time power flow software on Utilidata’s smart grid chip. And several utilities are in discussions with Utilidata to pilot its chip in smart meter deployments.
Why the Grid Needs AI
Smart meters are well positioned for tracking and responding to the two-way power flows rapidly coming to the edge of the grid.
Homes and businesses are installing batteries and solar panels that make them power generators. At the same time, they’re putting in electric-vehicle chargers that draw more than three times the power of the large air conditioners that used to drive grid demand. And with recent funding in the U.S. many more chargers and their grid demands will be coming online fast.
“You can’t write a good physics model to track all this, it requires a different approach to the problem that’s incredibly well geared to GPU computation,” said Hummon, who holds a Ph.D. in physics from Harvard.
“That’s why edge AI has gone from a nice-to-have to an urgent need for utilities preparing for what their customers are doing,” she added.
Startups Get Accelerated
Products from both startups will rely on NVIDIA full-stack technologies and an open platform for third-party app developers.
Hummon, of Utilidata, praised NVIDIA Inception, which nurtures technology startups, providing investor introductions, expertise and technology.
As part of the work developing the software-defined smart grid chip with NVIDIA, Inception gave her team access to GPUs in the cloud and training on how to make the best use of them. Inception also provided connections to other innovators across the power sector.
Smart Meters Get a Voice
For its part, Anuranet will make its Bullfrog devices come alive with conversational AI that combines its internal code for natural-language understanding with the automatic speech recognition in NVIDIA Riva, a GPU-accelerated software development kit for building speech AI applications.
“The accuracy of our code is already double that of rival offerings, and with Riva I expect we will outperform anything on the market by an order of magnitude,” said Anuranet CTO David L. Brock, a Ph.D. in robotics and AI from MIT, who developed a novel approach to natural language understanding.
“Riva is very exciting, and we will expect to expand our use of it in the future,” he added.
It’s one more way NVIDIA is fueling the transition to a more resilient grid and sustainable future.