Argonne Researchers Use Deep Learning to Study Avian Interactions with Solar Panels
Solar project installations across the United States are increasing, and scientists are trying their best to find out the impact of solar infrastructure on wildlife. Currently, the data collection methods being used are a bit outdated and time-consuming. Taking these things into consideration, the U.S. Department of Energy’s (DOE) Argonne National Laboratory has come up with a solution to monitor avian interactions with solar infrastructure cost-effectively.
The lab has been awarded $1.3 million (~₹98.6 million) from DOE’s Solar Energies Technologies Office to develop a technology that can effectively deal with birds and how they interact with the solar panels.
The technology uses computer vision techniques with artificial intelligence (AI) to keep a tab on these intruders at solar sites and collect the data on what happens when these fly by, perch, or collide with solar panels.
Speaking on this latest development, Yuki Hamada, a remote sensing scientist at Argonne, said, “There is speculation about how solar energy infrastructure affects bird populations, but we need more data to scientifically understand what is happening.”
According to an earlier study published by the Argonne lab, the collisions with solar panels at utility-scale solar projects across the country kill between 37,800 and 138,000 birds per year.
“The fieldwork necessary to collect all this information is very time- and labor-intensive, requiring people to walk the facilities and search for bird carcasses. As a result, it’s quite costly,” said Leroy Walston, an Argonne ecologist who led the study.
The new method uses technology to reduce the necessity of human surveillance by using cameras and computer systems and collect more and accurate data at a lower cost. The process involves three tasks, which include identifying which of the objects are birds and classifying events like perching, flying through, or colliding.
The scientists at the Argonne lab will use deep learning and AI method to teach the computers how to spot birds and behaviors by training them with similar examples.
“In an earlier project, the researchers trained the computers to differentiate drones flying in the sky. This latest project will build on this capability and add new complexities,” added Adam Syzmanski, a software engineer who developed the drone-detection model.
According to the new method, the cameras at the solar facilities will be angled toward the solar panels, and there will be more complex backgrounds. The system will have to differentiate between birds and other objects in the field of view.
The process will start with the setting of cameras at one or two of the solar energy sites. Hours of footage will then be processed and classified by hand to train the computer model.
Speaking on the significant amount of computing power required for the projects, Syzmanski said, “Model training requires a significant amount of computing power. We’ll be able to use some of the larger computers here at Argonne’s Laboratory Computing Resource Center for that.”
Once the computer model is trained and ready to go, the researchers will run the program internally on a live feed, classifying interactions on the fly.
The development required for tackling real-world challenges may be developed by using the Sage Cyber infrastructure initiative, led by Northeastern University and Argonne’s Waggle sensor system, to provide a faster computing platform.
This latest monitoring technology is being developed in collaboration with Boulder AI, a company with expertise in producing AI-driven cameras and algorithms that run on them.
The team of researchers at the Argonne lab will get help from a technical advisory committee, which comprises of machine learning experts from the Northwestern University and the University of Chicago. They will also get help from the Cornell Lab of Ornithology, conservation groups, and government agencies.
The resulting data from the above projects will be used to detect the types of birds that are more prone to strikes and also the time of day and year when these collisions increase. The data will also be used to find out whether the geographic locations of solar panels play a role in these types of interactions and also whether solar energy facilities in providing a viable habitat for birds.
“Once patterns are identified, that knowledge can be used to design mitigation plans,” Hamada added.
Recently, researchers at the University of Southampton announced that they had developed a technology that can map the locations of renewable energy projects across the globe and provide valuable insights into the potential environmental impact. The study shows the infrastructure density of solar and wind projects and also the power output in different regions.
Previously it was reported that the Ministry of New and Renewable Energy (MNRE) had issued a circular for retrofitting of transmission lines and wind turbines to avoid bird collision in Great Indian Bustard (GIB) habitats of Rajasthan and Gujarat. The circular was addressed to several wind developers such as Greenko and Suzlon Energy, among others.