Brunswick Electric Membership Corporation (EMC) serves all or part of four counties around the town of Shallotte in coastal North Carolina. It’s an area known for fishing, golf, and white-sand beaches.

It’s also hurricane country.

Over the years, Brunswick EMC has developed a good sense of where it needs to position its crews and other resources when a storm is coming, along with an idea of where the most troublesome outages are likely to occur, says Lewis Shaw, the co-op’s vice president of engineering and operations.

But what if co-ops like Brunswick EMC could get an even bigger leg up when preparing for bad weather?

The co-op is in the early stages of a pilot project testing Storm Readiness, a new machine-learning analytics software by General Electric that uses the MultiSpeak® interoperability standard to integrate data from multiple co-op systems and make outage predictions before severe weather hits.

“Our experience has taught us a lot of lessons over the years. We do a pretty good job gauging the impact based on the weather data we get and knowing our system,” Shaw says. “But you’re never 100 percent sure. This would give us a tool to back up what we know and make us feel more comfortable when we go out on a limb and allocate resources upfront and make commitments.”

Storm Readiness integrates high-resolution weather forecasts, outage history, crew response, and geographic information system (GIS) data and uses artificial intelligence (AI) to put together predictions and suggestions on a rolling, 72-hour look-ahead. MultiSpeak allows it to take data from a variety of sources, including GIS, outage management systems, and standard weather services, which is fed into an algorithm that is capable of learning over time.

Being MultiSpeak-compliant is key to the system’s ability to incorporate information from various sources, says Tony Thomas, NRECA’s senior principal engineer. The MultiSpeak standard is used by more than 800 electric cooperatives, investor-owned utilities, municipals, public power districts, and water and gas companies worldwide.

Thomas says MultiSpeak integration with AI analytics software like Storm Readiness or PwrMetrix, a program that uses data analysis to cut outage times, is the culmination of many years of smart grid development.

“These analytics are the holy grail of the smart grid. It’s what we’ve been working toward forever,” he says. “We’ve been collecting loads of data, but we’ve not been sure what to do with it. Now we can start merging all this information and really start using it.”

Training the AI

The pilot participants—Brunswick EMC; Great Lakes Energy in Boyne City, Michigan; Central Georgia EMC in Jackson; and Berkeley Electric in Charleston, South Carolina—are providing five years of outage and other data to GE as a first step in the project.

Ryan Lacy, a GE grid software representative, says the data will be used to train the AI algorithm at the heart of the system.

“We take that data, along with some boundaries data, and then we have standard weather service information we pump in over the top of that, and the analytic software crunches through it, and it learns,” Lacy explains.

The boundaries data is a geographic mapping overlay which divides a co-op’s territory into a grid of 7-mile-by-7-mile squares that it labels “townships.” Through its graphical user interface, the software will highlight areas within townships likely to be affected by outages as a storm approaches.

By looking at what has happened in the past—how and where outages occurred under different weather conditions—the software is able to predict with high accuracy where system problems are most likely to occur, Lacy says.

The next step in the project will be a demo for the participant cooperatives sometime this fall that shows how well the algorithm would have predicted weather-related outages that occurred in the past based on the data it has ingested.

The initial results will rely on past events, but the software will continue to learn in real time as it goes, refining its predictions about outages and where a utility might best position equipment and crews to deal with storm damage.

‘Fast and nimble’

To work effectively, data integration will have to be “fast and nimble,” Lacy says, especially because co-ops often don’t have large teams to handle the effort. Thomas notes that one of the benefits of MultiSpeak is it incorporates instruction sets known as WSDLs (Web Services Description Language) that make it easier to implement than some other interfaces.

The potential benefits to cooperatives from analytics software like Storm Readiness are significant.

Last September, when Hurricane Florence roared ashore in the Carolinas, it left a wide swath of damage across much of the state as strong winds and record flooding disrupted service for several co-ops. Brunswick EMC saw outages affect nearly 78,000 of its 92,000 meters.

“After the storm, getting around from point to point was difficult,” Shaw says. “We had several pockets where we couldn’t get across the roads, and access to the area from outside was very limited for several days. Getting materials in was very tough, which increased the need to plan ahead.”

Working around the clock and drawing on experience, Brunswick EMC employees were able to restore power to most of the membership within four days. But anything that could have increased their ability to respond efficiently would have meant fewer man hours and less expense.

“At NRECA, we talk a great deal about increasing the resilience of electric distribution networks,” Thomas says. Machine-learning systems “promise to increase the resilience of the networks and may help increase member satisfaction as well as reduce the cost of repair.”

Shaw says those capabilities could also help cooperatives manage a generational transition from veteran staff, who rely on their years of experience, to younger personnel who grew up in a digital environment.

“I think it will be of even greater value for the next generation to validate some of the assumptions we make going into a storm,” he says.

Overall, Shaw believes the potential for data integration and processing made possible through MultiSpeak and analytics software is just starting to be tapped.

“It’s exciting to think about the possibilities with AI and the knowledge we can gain making use of the data available,” he says. “It’s something of interest to all of us utility folks, and I can see that the bright hope of the future is finally putting all this data to use.”