PSU, Eversource to use big data to better predict storms that cause power outages

  • Andrea LaRocca, a graduate student in applied meteorology at Plymouth State University, describes a map of accumulated snowfall from March 14 in the PSU Weather Center on May 2, 2017. David Brooks—Monitor staff

  • Eric Hoffman, professor of meteorology at Plymouth State University, describes a chart May 2, 2017, of precipitation readings taken by a vertical Doppler radar, showing at which height snow (in blue) changed to rain (in red) as it fell. David Brooks—Monitor staff

Monitor staff
Tuesday, May 02, 2017

Ice storms have a bad habit of taking down New Hampshire power lines and leaving us in darkness, but Plymouth State University and Eversource Energy have a weapon we can use to fight back: mathematics.

“This takes a lot of statistical analysis – a lot of it,” said Eric Hoffman, professor of meteorology and director of outreach for the Judd Gregg Meteorology Institute at Plymouth State University, during a presentation Tuesday about a new research partnership with Eversource Energy aimed at honing weather predictions in order to reduce power outages.

The partnership is an expansion of a program that Eversource and the University of Connecticut have been developing for five years in which researchers are building systems to better predict wind damage from incoming weather patterns. Eversource is funding the project, to the tune of some $15 million at UConn, because preparing for storms by bringing in line crews and equipment is a big deal to an electric utility.

“I’ve been involved in storms where we’ve spent $50 million in advance of a storm hitting our system, bringing in resources,” said Bill Quinlan, president of Eversource’s New Hampshire operations. “Sometimes we guess right, sometimes we miss.”

They want to miss less, which explains the memorandum of understanding signed Tuesday. Under it, Plymouth State University’s well-respected meteorology program will join the company’s collaboration with UConn for a five-year research agreement that will kick off with a roughly $200,000 expenditure over two academic years.

The first step of the program will be for faculty and students – two graduate students and up to seven undergraduates to start – to take the wind-damage prediction methods developed in Connecticut and bring them to New Hampshire. That begins by taking detailed data about weather and power outages from past years and finding how one connects to the other, which is where the mathematics comes in.

“We’re identifying indications that are correlated with past events,” said Jason Cordeira, assistant professor of meteorology, as he launched into a quick explanation of how the school uses 84 linked computers to crunch weather data from myriad sensors, a process so computation-heavy that a single run can take 12 hours. Those correlations will then be used to create mathematical models and applied to weather patterns to see how well they line up with what winds actually do to power supplies.

The goal is to tell utilities as many as five days in advance that a weather pattern moving into an area has this or that likelihood of creating specific levels of damage in various locations. That ability exists in Connecticut, Quinlan said, to the point that the number of power outages per town has been predicted several days in advance of a major storm.

The second step for PSU will be for its researchers and UConn researchers to figure out how to predict outages due to snow and ice, not just wind.

“That’s a little bit tougher, scientifically,” because so many more variables are involved, Hoffman said.

For example, he pointed to a Doppler radar at PSU, on loan from the National Oceanographic and Atmospheric Administration. It points directly upward, which allows it to accurately determine how fast drops of precipitation are falling, from thousands of feet up clear to the ground.

Snow falls more slowly than rain – barely 2 meters per second for snow, compared to as much as 9 meters per second for rain – which means the radar shows at what point precipitation is freezing. That in turn reflects whether communities at different altitudes are about to see rain; snow; or the most damaging form of precipitation, freezing rain.

Quinlan said that such predictions and patterns can also be used to plan the company’s tree-trimming programs, by identifying those locations most likely to be hit by storms that bring limbs down onto power lines.

It can be combined with other research at PSU’s forestry and environmental departments and other locations throughout New Hampshire, to the point of determining which species of trees are likely to be a problem in different storm conditions.

At the Hubbard Brook Experimental Forest near Woodstock, for example, researchers with the USDA and other agencies have sprayed water onto trees in midwinter to simulate an ice storm and quantify the effect that the accumulation of ice has – exactly what electric utilities want to know when deciding which crews to send where, and when, as an ice storm approaches.

For PSU, Tuesday’s announcement fits well into its new academic structure that emphasizes clusters built around real-world problems and applications, rather than around traditional academic disciplines. A major incentive for making this change was to provide a better context for industry funding and cooperation.

(David Brooks can be reached at dbrooks@cmonitor.com or 369-3313.)