Google has put itself on the map by funding more than $2 billion in renewable energy projects.
Now the search giant wants to map the amount of sunlight your rooftop receives — and help you decide if it makes financial sense to go solar.
Dubbed Project Sunroof, the recently released online tool enables you to type in your address and find out how much space you have for solar panels on your roof, how many hours of rooftop sunlight you’ll get a year, and how much money you’d save.
After typing in your address, a Google Earth image of your home and the surrounding neighborhood appears, with the roofs appearing in colors ranging from purple to yellow to indicate how much sunlight is striking the surface. The photos, however, appear in some cases to be at least one or two years old.
“As Google, we knew we had the expertise to do this well using Google maps and aerial imaging,”
“As Google, we knew we had the expertise to do this well using Google maps and aerial imaging,” said Barry Fischer, a Google spokesperson. “Project Sunroof draws upon the same high-resolution imagery that powers Google Earth.”
The project — which started as the “20% time” effort of a Cambridge, Massachusetts–based Google software engineer — is a latecomer to the solar mapping scene.
City and state governments, the United States Department of Energy, and solar installers such as Sungevity have developed similar tools enabling curious homeowners to type in their address and determine how much solar power they can produce from a photovoltaic system.
But Google’s tool — which is still a pilot and so far only available for residents of the San Francisco Bay Area, Boston, and Fresno, California — appears to be unique as it uses machine learning to distinguish a rooftop from, say, adjacent trees or lawns.
The team found that the accuracy of estimates of a roof’s solar capacity was increased by reducing the computer’s chance of mistakenly including surrounding trees or lawns when calculating the rooftop’s exposure to sunlight.
“Based on our testing, we found that using machine-learning techniques alone helped reduce classification errors by 75% compared to using traditional methods of determining the boundaries of buildings and rooftops,” Fischer said.
The method is also being applied to solar in other ways. Students at Duke University, for instance, are using machine learning to estimate U.S. solar capacity.
Project Sunroof will also recommend the size of solar system that you should install based on your average electricity bill. And it calculates how much you’ll save depending on whether you opt to finance the solar array with a loan, by leasing, or by purchase. Then it presents the option to request a consultation with Google’s solar provider partners — SunWork, Vivint, SunEdison, SunPower, or NRG Solar.
Fischer said in the future Google will collect a referral fee from those companies when a potential customer contacts it through Project Sunroof.
Google plans to expand the tool to other areas of the United States.
“During the coming months we’ll be exploring how to make the tool better and more widely available,” Fischer said.