Machine Learning And Solar Farms


Large-scale real-world solar data sets and cutting-edge machine learning techniques were coupled by American researchers to examine the effects of severe weather on solar farms and identify the variables that influence energy production. The scholarly journal Applied Energy published its findings earlier this month. The National Renewable Energy Laboratory and the Department of Energy’s (DOE) Solar Energy Technologies Office collaborated on this study, which was funded by the DOE.


A solar farm is at risk from hurricanes, blizzards, hailstorms, and wildfires, both directly through costly damage and indirectly through obstructed sunshine and decreased electricity production. To evaluate the effects of severe weather on the facilities, two American researchers examined repair tickets from more than 800 solar farms spread across 24 states. They then integrated this data with information on power generation and weather records. They want to make solar farms more resistant to severe weather by identifying the elements that contribute to low performance.


If we want the renewable energy sector to be robust in the face of a changing climate, we must work to understand how future climate conditions can affect our country’s energy infrastructure. We are now concentrating on extreme weather events, but soon we will also include chronic exposure events, such as persistently high temperatures – Senior Scholar


The research team first searched six years’ worth of solar maintenance records for important weather-related terms using natural-language processing, a form of machine learning utilized by smart assistants. They later shared their analysis techniques, which other photovoltaic researchers and operators can access for free.


Even though hailstorms are notoriously expensive, they did not show up in solar farm maintenance records, most likely because owners often record hail damage through insurance claims. Instead, she discovered that the other weather phrases, such as snow, storm, lightning, and wind, were referenced in approximately 15% of weather-related maintenance records.


The framework that supports the panels, known as the racking, can sustain damage from storms’ high winds, according to the paper’s principal author. Flooding that prevents access to the site and delays the process of turning the plant back on is the second significant issue they have identified through the maintenance logs and discussions with our industry partners.


To evaluate the effects of severe weather on solar farms, the researchers merged more than two years’ worth of actual energy production data from more than 100 solar farms spread across 16 states with historical meteorological data. They discovered through statistical analysis that snowstorms, followed by hurricanes and a broad category of other storms, had the greatest impact on the production of electricity. After that, they employed a machine learning technique to identify the unknown elements that contributed to the poor performance caused by these extreme weather events.


According to the lead author, statistics only provide a portion of the picture, but machine learning has been quite useful in identifying the key elements. Machine learning was utilized to focus on the most crucial variables from the researchers’ suite of variables. The research team discovered that older solar farms were generally more susceptible to extreme weather. Older solar farms may have experienced greater wear and tear as a result of being exposed to the elements for a longer period of time.


To help the grid adapt to the changing environment and developing technology, the researchers are currently expanding their research to examine how severe weather affects the entire electrical system, add more production data, and address even more problems.