Researchers spy on marine energy machines using data analysis and a raft packed with ocean monitors

UW Mechanical Engineering graduate student Emma Cotter wants to know if marine energy machines are harming sea life. She is using machine learning to create tools for processing massive amounts of underwater monitoring data and to fine-tune data collection. (University of Washington Photo / Matt Hagen)

In the basement of a nearly century-old, Gothic building on the University of Washington campus, engineers are conducting research essential for developing clean energy sources of the future.

In the basement of the UW’s Harris Hydraulics Lab, Cotter stands in front of the frame of a new Adaptable Monitoring Package (AMP). Once outfitted, it will carry sonar, cameras and other devices to spy on marine life. (GeekWire Photo / Lisa Stiffler)

A team from the UW’s Department of Mechanical Engineering is building monitoring devices and fine-tuning algorithms to understand the impact of tidal and wave energy machines on seabirds, marine mammals and fish.

Because just as hillside wind turbines can harm and even kill birds, there are also concerns about the effect of spinning underwater turbines or bobbing wave-energy generators on marine life.

“It’s the question that comes up every time people talk about deploying technology, particularly when you talk about utility-scale deployment,” said Brian Polagye, a UW associate professor leading the work on marine energy and environmental modeling.

The biggest worries focus on animals colliding with the power-producing machines, and behavior changes caused by the noises the devices make.

Often researchers begin examining the harm to birds, fish and other creatures after erecting power generators such turbines and dams. The UW scientists are hoping to avoid that fate. Engineers are still experimenting with different marine turbine designs and the U.S. does not yet have any large-scale underwater energy installations.

“If we can understand the environmental effects before we converge on a design,” Polagye said, “we can design more benign systems.”

Back in the basement of the UW’s Harris Hydraulics Laboratory, Emma Cotter, a graduate student in Polagye’s lab, walks around a raft-like device made of white plastic. She points to spots where different monitoring tools will be attached.

The device, called an Adaptable Monitoring Package, or AMP, carries long- and short-range sonar, two optical cameras and lights, four hydrophones and a sensor for measuring water currents. When deployed, the AMP is tethered to the seabed, submerged 10 meters deep or more. An energy-producing turbine can be attached to the AMP or installed near it to observe its effect on marine life.

The UW lab is unique for developing a system that combines a suite of monitoring devices. The sensors corroborate each other, building a clearer image of underwater events. But that approach comes at a price.

“It makes a pretty big challenge for data processing and storage,” Cotter said, given the massive volume of images and data that the monitors generate.

Images of a seal as detected by sonar and an optical camera. Images courtesy of Emma Cotter, University of Washington.

In tests conducted offshore of Sequim, along the northern coastline of Washington state, an AMP collected data for 45 seconds every 15 minutes. If the AMP collected data continuously, Polagye estimates that a year’s worth of data from an AMP would fill a metric ton of hard drives.

So in addition to designing and building a robust monitoring system, the researchers also needed to figure out how to handle the data.

That’s where Cotter comes in. With help from other UW students, they analyzed sonar readings and camera images from an AMP to match sonar blips with seals, diving seabirds or schools of fish. Cotter then used machine learning algorithms to teach the computer how to do the time-consuming analysis on its own. Her work marks an important step forward, Poyagye said, by using data from multiple types of sensors, integrating that information and making sense of the giant volumes of data.

“That capability, Emma is the first one to do that,” he said.

The next step is fine-tuning the system to analyze the sonar images in real time, triggering more frequent data collection — snapping photos and gathering sound — when the monitors sense that an animal is nearby.

“No one has detected a collision [with a marine-power machine],” said Cotter. “So we’re looking for something that if it happens, will be really rare.”

The UW scientists are not alone in the search. Thirteen countries including the U.S. are part of an International Energy Agency Ocean Energy Systems project to share information on the effects of marine energy on sea life. The U.S. Department of Energy (DOE), the Bureau of Ocean Energy Management, the National Oceanic and Atmospheric Administration and the Pacific Northwest National Laboratory (PNNL) all support the effort. The UW project is funded by DOE’s Water Power Technologies Office.

Among the participants, the United Kingdom is the first to install a commercial tidal energy facility. It generates enough power for roughly 6,000 homes, with plans to add more production. In the U.S., energy prices and a lack of government subsidies currently makes wide-scale marine energy unfeasible, Polagye said.

Retrieving an AMP from Pacific Northwest National Laboratory’s underwater testing site at Sequim, Washington. AMPs can collect data underwater for up to six months. (University of Washington Photo / Matt Hagen)

In America, the first installations will likely serve areas off the grid, Polagye predicts. There’s an isolated village in Alaska that runs on diesel fuel that’s flown in, he said, and they’re interested in testing marine power. There could be applications for remote aquaculture farms, military projects or even to power underwater data centers like those being tested by Microsoft.

Later this month, the UW researchers are taking their an AMP to Oregon to partner with scientists at Oregon State University. For six weeks they’ll gather data in deeper conditions in the Pacific Ocean.

The researchers are also working on monitoring devices called Drifting Acoustic Instrumentation Systems, or DAISYs, that float on the surface, gathering acoustic data using submerged equipment.

The team hopes that the AMPs will be ready for use by global marine-energy testing facilities in the next year or two. A former graduate student from Polagye’s lab who is now a UW employee is launching a startup called MarineSitu to market the monitoring devices.

When it comes to marine energy, “we still need to understand what the environmental effects are,” Polagye said. “So the tools that we deploy, the AMPs and the DAISYs, will play a critical role in developing comfort with the technology.”