AI-Backed Sensors Help Reduce Wind Turbine Risks to
Protected Birds
On a Wyoming wind
farm, there’s an unusual addition amid the familiar, massive wind turbines. A newly built tower, which stands only about 33 feet tall, appears small in comparison to the neighboring 328-feet-tall wind turbines.
Atop the miniature tower sits an array of cameras — a layer of devices is arranged in a circle and one camera is stationed above swiveling in different directions.
Suddenly, to the west, a speck appears in the sky. The top camera swivels to face the speck, which is steadily growing larger on the horizon.
Within seconds, the camera has identified the approaching object as a golden eagle, one of only 40,000 left in the United States — and one that is flying directly toward the wind turbine’s deadly spinning blades.
Although wind farms utilize
a renewable source of energy and therefore do not contribute to greenhouse gas
emissions, wind turbines’ blades, revolving at 170 miles per hour, can kill
birds on contact.
The harm that turbines have already imposed on protected avian wildlife, such as the golden eagle, has dampened conservationists’ enthusiasm for them, but a camera detection system called IdentiFlight is helping the wind energy industry become safer for protected birds.
The harm that turbines have already imposed on protected avian wildlife, such as the golden eagle, has dampened conservationists’ enthusiasm for them, but a camera detection system called IdentiFlight is helping the wind energy industry become safer for protected birds.
IdentiFlight scans the sky in a 360-degree radius and can spot any object flying within 0.6 miles. Once it detects a flying object, the system uses artificial intelligence to analyze the image against 200 image attributes, including body length, wing span and wing posture, in order to classify what species the bird is, and determine if the bird is one of the protected species that IdentiFlight has been programmed to watch out for.
Initially designed in the United States to detect golden as well as bald eagles, IdentiFlight is expanding its species classification capabilities as the technology spreads to geographies with different avian identification needs. Once IdentiFlight makes its species classification, it produces a confidence value to measure how perfectly the object matches the system’s image attributes for a certain species.
That’s not all.
IdentiFlight can also track the bird to determine its flight path and assess
whether the bird is flying in a direction that takes it beyond the reach of the
spinning turbine or if the bird is flying directly into danger.
In addition, IdentiFlight uses captured images and distance information from its Stereo Unit to calculate the bird’s velocity — if the bird is flying at a speed that allows it to pass between the spinning blades unscathed or if its progress will not be fast enough to avoid harm.
Within 5 seconds of detection, IdentiFlight has gathered enough information to determine whether to shut a turbine down. Thirty seconds after a shutdown is triggered, the turbine’s blades have slowed to a safer speed for the bird to pass through unscathed.
In addition, IdentiFlight uses captured images and distance information from its Stereo Unit to calculate the bird’s velocity — if the bird is flying at a speed that allows it to pass between the spinning blades unscathed or if its progress will not be fast enough to avoid harm.
Within 5 seconds of detection, IdentiFlight has gathered enough information to determine whether to shut a turbine down. Thirty seconds after a shutdown is triggered, the turbine’s blades have slowed to a safer speed for the bird to pass through unscathed.
IdentiFlight’s ability
to protect birds from wind turbines exceeds what human monitors can achieve. A
2016 study conducted by The
Peregrine Fund, Western EcoSystems Technology and American Wind Wildlife
Institute examined the effects of four IdentiFlight units installed at Duke
Energy Renewable’s Top of the World Windpower Project, a wind farm located
outside of Glenrock, Wyoming.
The study showed that IdentiFlight was able to detect 562 percent more birds than humans could, although it missed 4 percent of the birds that humans spotted and misclassified 9 of 149 eagles as non-eagles and 287 of 1,013 non-eagles as eagles.
The study showed that IdentiFlight was able to detect 562 percent more birds than humans could, although it missed 4 percent of the birds that humans spotted and misclassified 9 of 149 eagles as non-eagles and 287 of 1,013 non-eagles as eagles.
Since this research,
IdentiFlight has transitioned from traditional machine vision techniques to an
artificially intelligent neural network technology.
This shift is part of an effort to reduce the misclassification of non-eagles as eagles and avoid costly turbine shutdowns for non-protected birds, as well as to reduce the misclassification of eagles as non-eagles and decrease risks to protected species.
IdentiFlight is also now capable of shutting down wind turbines on its own, when earlier models relied on a human operator to receive IdentiFlight’s alert and initiate the shutdown.
This shift is part of an effort to reduce the misclassification of non-eagles as eagles and avoid costly turbine shutdowns for non-protected birds, as well as to reduce the misclassification of eagles as non-eagles and decrease risks to protected species.
IdentiFlight is also now capable of shutting down wind turbines on its own, when earlier models relied on a human operator to receive IdentiFlight’s alert and initiate the shutdown.
IdentiFlight hopes to
keep improving the technology. The company aims to soon be able to judge how
aware a bird is of its surroundings. If a bird is looking ahead, it may be able
to see the spinning wind turbine and adjust its flight path to avoid contact.
On the other hand, if a bird is hunting for prey and focused on the ground, it
may not notice the turbine ahead.
“We ought to be able to look at their flight patterns, classify those in some way, and begin to make some judgments about whether this bird knows what he’s doing so we don’t worry about him, or whether this bird is not paying attention and we need to worry about him,” says Tom Hiester, president of IdentiFlight.
“We ought to be able to look at their flight patterns, classify those in some way, and begin to make some judgments about whether this bird knows what he’s doing so we don’t worry about him, or whether this bird is not paying attention and we need to worry about him,” says Tom Hiester, president of IdentiFlight.
IdentiFlight is now
operating at wind farms around the world, even as the company continues to test
the system. Duke Energy Renewables’ has already purchased 47 units for its
Wyoming wind farm. Operations in California, Oregon, Washington and Utah are
also in the process of setting Identiflight units, or have done so already.
In September 2018, IdentiFlight received its first international purchase for a trial in Germany. Since then, IdentiFlight has been bought for full scale commercial application in a wind farm in Tasmania, Australia, and is being considered for projects in Spain, France and Sweden.
In September 2018, IdentiFlight received its first international purchase for a trial in Germany. Since then, IdentiFlight has been bought for full scale commercial application in a wind farm in Tasmania, Australia, and is being considered for projects in Spain, France and Sweden.
A lot of the interest
in IdentiFlight comes from companies already in hot water for the deaths of
protected species that have been killed by turbines. While IdentiFlight can
help wind farms correct and improve following accidents, the company aims to
become more than just a fix for protected bird losses, and actually facilitate
the expansion of wind power.
“What I hope for
IdentiFlight is that we can allow wind companies to go where there may be
eagles but lots of wind such as in Wyoming and Montana, and they can do more
wind projects,” says Hiester. “I’d like IdentiFlight to become an enabling
technology rather than a fixing technology.”
Angela
Sivak is a freelance writer
and graduate student at Johns Hopkins University.