Solar panels have a love-hate relationship with nature. They should be placed in exposed places, which receive a lot of sunlight, but cloudy weather markedly reduces their production. Less obviously, more extreme weather – from blizzards to hurricanes – can completely damage or even break solar hardware. new research The study, carried out by Sandia National Laboratories and published in Applied Energy, shows how weather events can reduce the amount of energy produced by solar farms in the United States.
To study this relationship, the researchers deployed machine-learning algorithms on large sets of data from private solar farms. “It was a huge, collaborative effort,” Tushara Gunda, one of the paper’s authors and a researcher at Sandia, told Ars. Going forward, Gunda wants to expand this research to look at other extreme weather events and renewable energy such as wind, geothermal and marine energy. He said his team is in the early stages of this work.
The team hopes that this research can be used to make decisions about future solar operations. This is especially true as climate change increases the frequency of extreme weather events, potentially leading to more issues affecting solar output. “We recognize that with the shift to renewable energy, there has been an increased reliance on local environmental conditions,” Gunda said.
To start, Gund and his co-author and fellow Sandia researcher Nicole Jackson collected more than 800 maintenance stamps from solar farms in 24 states. They then went back and forth with solar companies trying to understand the data set—for example, different companies sometimes used different words for the same thing.
The team had to do a lot of analysis to determine what each company meant when they used the word “hurricane,” as some firms included snow events or even hurricanes on their maintenance tickets. was classified as “Certainly on the industry and every day practice side of things, a hurricane can happen on any sunny day,” Jackson told Ars.
“Just because someone wants to share data doesn’t mean you can automatically analyze it. There are nuances with regard to how the data was collected,” Gunda said.
The researchers also obtained more than two years of electricity generation data from more than 100 solar farms in 16 states, along with historical weather data for those regions. From there, the authors ran a machine-learning algorithm on the data set to explore the relationship between rates of energy production and severe weather events. The algorithm allowed the team to identify points at which a weather-related power drop coincided with maintenance tickets and other variables.
highlight the situation
The team found that snow events caused the greatest reduction in performance (54.5 percent), followed by hurricanes (12.6 percent) and thunderstorms (1.1 percent). Somewhat surprisingly, hurricanes were mentioned in about 15 percent of maintenance records. Other factors that can lead to low performance include plant size, age, and location. “We noticed that older farms were more likely to be affected by performance issues,” Gunda said.
However, there are some nuances to this work. For one, older sites being more affected does not mean they are unproductive – rather, they are exposed to more weather than their younger counterparts (even older farms are relatively young, between three and five years). In addition, the sites the researchers collected data from were biased toward North Carolina and California. These states have severe weather events that may not occur in other parts of the US.
The team was also surprised to find that neither hail nor forest fires appeared in the data. That’s not to say that these events aren’t happening – the West Coast gets a lot of fires. Instead, maintenance tickets were notably absent at these events because companies only make them if there is something there for them. Most likely, given that hailstorms are covered by insurance, these events will show up in the insurance database.
“But we know from our interactions with industry and attending conferences that these special events are certainly of interest,” Gunda said.
Insurance data really tells a story when it comes to how much damage solar farms can do during a hailstorm. a report of National Renewable Energy Laboratory, published last year, uses data gathered from Verisk – an insurance services company – to dig into the amount of damage that weather events can cause solar operations. (The insurance data also includes vandalism and theft numbers).
Data collected between 2014 and 2019 shows that hailstorms caused the largest number of insurance claims with solar hardware, weighing in at 7,979 cases with an average cost of $2,555. “Jay is a big deal for solar panels,” Andy Walker, a senior research fellow at NREL, told Ars.
Fires were less common (1,282 cases), but they had larger-than-average claims at $17,309. There were 79 cases of a cold, including snow and ice, an average of $5,288. However, these averages include the cost of both commercial and residential solar operation. For example, in the case of residential freezing, the average claim value was $4,195, but it was $32,964 for commercial operations.
The unpublished NREL research also suggests ways that solar panels can better withstand extreme weather, Walker said. Methods include water-tight enclosures, modules mounted on three rails (instead of two), thick glass, wind-cool fences, marine-grade steel, and through-bolting (instead of clamps). “It turns out that clamps are the smoking gun in a lot of Module Liberation, as it’s called when a [photovoltaic] The module blows up a rack,” he said.
Walker said the upgrades cost a few cents for each watt of output. Some of these methods can help with a wide variety of weather events that solar panels will see and can increase the magnitude of the danger that the panels can survive – from being crushed by excess snow to being blown off their racks. To hail bombardment. “Solar panels are one of the most exposed things you’ll find in the built environment,” he said.
Applied Energy, 2021. DOI: 10.1016/j.apenergy.2021.117508 (About DOI)