New research conducted by Stanford University using an AI called DeepSolar found that there are roughly 1.47 million solar rooftops across the US, higher than previous leading estimates. The findings can help utilities, regulators, solar companies and others understand who goes solar and why, as well as how to encourage local grid reliability.
The team of researchers created DeepSolar using 370,000 satellite images, covering about 100 feet by 100 feet. The program was told whether each image included or did not include a solar panel. Using that information the program learned to identify solar panel features including color, texture and size. After learning those features researchers had it evaluate roughly a billion satellite images to to find solar installations across the country.
“We can use recent advances in machine learning to know where all these assets are, which has been a huge question, and generate insights about where the grid is going and how we can help get it to a more beneficial place,” said Ram Rajagopa Stanford associate professor of civil and environmental engineering.
The research, led by Rajagopal and Arun Majumdar, Stanford associate professor of civil and environmental engineering, was published in the journal Joule on Dec. 19. In addition, they published results of DeepSolar, including interactive mapping tools at: http://web.stanford.edu/group/deepsolar/home.
DeepSolar was built by doctoral candidates Jiafan Yu and Zhecheng Wang, respectively seeking doctorates in electrical engineering and civil and environmental engineering. “We don’t actually tell the machine which visual feature is important. All of these need to be learned by the machine,” Yu said.
“We found some insights, but it’s just the tip of the iceberg of what we think other researchers, utilities, solar developers and policymakers can further uncover,” Majumdar said. “We are making this public so that others find solar deployment patterns, and build economic and behavioral models.”
To get a more complete picture of who is going solar and why, the researchers also integrated data from the US Census and other sources. They found, among other things, that household income remains an important factor in who goes solar.
They found that even in areas where low- and medium-income households could profit by going solar because of great solar resources and high electric prices, they often do not. The report authors suggest the upfront cost of solar could be a factor. However, when households have an income of $150,000 or more, income played less of a role in which homeowners went solar.
The authors said the research and findings can help regulators, solar installers and marketers, as well as utilities and others understand the needs of local markets better. It can help a utility balance supply and demand or a policymaker understand whether or not they should establish policy to encourage more solar power locally.Tweet