Researchers at Stanford University now estimate that there are 1.47 million solar panels in use across the contiguous 48 states, a number that's higher than any other previous estimate. But what's really cool about this stat is the way they got it.
The Stanford scientists built an A.I. called DeepSolar to count up solar panels from space. The machine-learning system "analyzes satellite imagery to identify the GPS locations and sizes of solar photovoltaic (PV) panels," according to .
“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,” says Ram Rajagopal, an associate professor of civil and environmental engineering who helped lead the project, in a .
DeepSolar's data reveals that despite , income still plays a large role in determining who is and who isn't likely to invest in solar panels. According to , "low- and medium-income households do not often install solar systems even when they live in areas where doing so would be profitable in the long term." Researchers theorized that the upfront costs of solar installation are still hindrances, even if solar could save money down the road.
DeepSolar was fed around 370,000 images, all of them covering about 100 feet by 100 feet of real territory. These pictures were identified as either "having solar panels" or "not having solar panels." With that as a starting point, DeepSolar was able to learn the differences between solar panels—color, texture and size. About 93 percent of the time, DeepSolar was able to correctly identify a solar panel. Scientists estimate it missed around 10 percent of images that did have solar installations.
The resulting database comes from over a billion satellite images, looking at residential, business, and utility solar panels. Admittedly, the data set isn't perfect. Most rural areas were skipped in the study because of an assumption that they would not have solar panels, or they would not be connected to a greater energy grid. The scientists estimate that they missed a further 5 percent of all solar panels because of the choice and plan to improve DeepSolar to include these regions in the future.
As for now, the team hopes that their project can help answer the big questions about solar energy: Who's using it, who isn't, and what's stopping people from using it more? Data shows that in 2016 solar energy became the second most common new power source in America. Anyone interested in learning how it could become number one could find something to sift through in DeepSolar's maps. , professor of mechanical engineering, says:
“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. We are making this public so that others find solar deployment patterns, and build economic and behavioral models.”