Sign In

Communications of the ACM

News

Algorithms to Harvest the Wind


View as: Print Mobile App ACM Digital Library Full Text (PDF) In the Digital Edition Share: Send by email Share on reddit Share on StumbleUpon Share on Hacker News Share on Tweeter Share on Facebook
wind turbines

Credit: Juan Enrique del Barrio

Wind-generated electricity has expanded greatly over the past decade. In the U.S., for example, by 2018 wind was generating 6.6% of utility-scale electricity generation, according to the U.S. Energy Information Administration. The criteria for efficient design and reliable operation of the familiar horizontal-axis wind turbines have been well established through decades of experience, leading to ever-larger structures over time, both to intercept more wind and to reach faster winds higher up.

As these gargantuan turbines are assembled into large wind farms, often spread over uneven terrain, complex aerodynamic interactions between them have become increasingly important. To address this issue, researchers have proposed protocols that slightly reorient individual turbines to improve the output of others downwind, and they are working with wind farm operators to assess their real-life performance. Beyond extracting more power from current farms, widespread use of these "wake-steering" techniques could allow denser wind farm designs in the future.

Back to Top

Bigger Is Better

"The tendency is to build higher and higher turbines," said Mireille Bossy, a fluid dynamics expert at Inria, the French national institute for computer science and applied mathematics, located in the Sophia Antipolis technology park near Nice, France. "We are talking in a new project about 300m [about 984 feet] in height." The wake of slower disturbed air typically extends 10 or more times the diameter iof a turbine, robbing downwind turbines of wind. Completely avoiding power loss for downwind turbines would demand several kilometers between turbines, incurring substantial additional costs in real estate and wiring.

These costs and the specific constraints of available sites often lead to less-than-optimal arrangements, however. Predicting the interactions is difficult, especially for farms located in terrain that may create turbulent flows. Bossy described one existing farm where the addition of a single turbine, in what seemed like a good place to minimize wake interactions, would decrease the output of the entire facility. "It's complicated," she stressed. "We cannot just do some wind-tunnel simulation," she stressed. "We need to simulate the site."

Back to Top

Wake Steering

Things get complicated quickly, because each combination of wind direction and speed, as well as other atmospheric details, requires a new simulation. Fortunately, the wake interactions can be reduced by "yawing" the turbines: rotating them slightly around a vertical axis, which deflects their wakes. Although a misalignment slightly reduces the power output of the upwind turbine, for some wind directions this can be more than offset by increased power from downwind turbines (which is proportional to the cube of the windspeed).

Still, the number of combinations of possible yaw angles for each turbine, as well as windspeeds and directions, quickly becomes computationally challenging. For this reason, the U.S. National Renewable Energy Labs (NREL) has developed both a calculation-intensive "large-eddy" computational-fluid-dynamics model, called Simulator fOr Wind Farm Applications (SOWFA), as well as a simpler tool for steady-state calculations called FLOw Redirection and Induction in Steady State (FLORIS).

Paul Fleming, who developed these tools with colleagues at NREL and the Delft University of Technology in The Netherlands, noted that although there is still debate about how to deal with rapidly shifting wind directions, there "seems to be some convergence toward steady-state modeling." Wind farm operators currently prefer to set a yaw angle and hold it for a while, "striking some balance between trying to keep up with the changing wind direction and trying to yaw as little as possible," he said. "Wake steering has to be built on the same structure."

A wind farm's electricity generation over the course of a year, known as annual energy production or AEP, is likely to see only small fractional increases from wake steering, in part because many wind directions would not create substantial losses in any case. For existing facilities, Fleming said, "a reasonable guess for total AEP gain is somewhere between 1% or 2%." The gains could be especially compelling for offshore generation, where winds tend to be steadier, turbines larger, and wakes more persistent, but land-based sites can benefit as well.

This improvement may seem modest, but it could amount to millions of dollars of revenue for very little cost. "It's garnered a lot of interest" from the industry, Fleming said. Indeed, at a September 2019 meeting attended by major wind developers, he said, "There was pretty broad agreement that something like this will be adopted more widely."

Back to Top

Out in the Field

To get this kind of buy-in, "field tests are critical," Fleming said. For this reason, he and his colleagues have worked with manufacturers, including NextEra, which he said is "the largest owner of turbines in the U.S.," to conduct field trials that have validated the simulation predictions. For one unusually close pair of turbines, spaced approximately three times the diameter of the turbines apart, the power from the downwind turbine for the worst wind direction was increased by about 14% when the upwind turbine was yawed to deflect the wake. This deflection produced in an overall 4% increase for the pair.

"Right now, the algorithms we're implementing aren't very complicated; they're essentially a lookup table" of yaw offsets for a particular windspeed and direction, Fleming said. Over time, as the technique proves its value, he expects these algorithms can be refined.

John Dabiri, now at the California Institute of Technology, recently explored one such refinement with colleagues, and followed it up with field experiments. "What we were aiming for was to do site-specific optimization: for a given layout, a given terrain, a given location where the wind conditions are what they are, and to be able to incorporate historical data in a way that informs a physics model."

Other researchers have used such historical data, capturing how much energy each turbine generated under various conditions with no wake steering, to train machine learning models. "The challenge is that we don't typically have enough data," Dabiri said, so models can overfit the existing data but fail to generalize to different locations. He and his team combine the data with a simplified physics model to match each site. The model is efficient enough to optimize the entire set of yaw angles "on a laptop computer in a few seconds."

Dabiri's team, then at Stanford University, worked with wind farm operator TransAlta to test their optimization algorithm on a line of six turbines in Alberta, Canada. "That middle ground, between the two-turbine studies and a full wind farm, is important for us to investigate," he said, to give operators confidence about real-world operation.

"Academic research has largely been focused on numerical simulations, some wind-tunnel studies, and then, even in the field, it's typically maybe a pairwise study," Dabiri said. "We're finding there's still a pretty big leap from standard methods of investigation and what happens in a real wind farm." One concern is "secondary steering," in which a deflected wake is further modified by interactions with the downwind turbine, which is not important for just one pair of turbines.

As the researchers hoped, their algorithm increased electric output by almost 50% for slow winds directed along the line. Wake steering also significantly reduced fluctuations in power generation due to turbulence, another important consideration. However, these wind conditions are rare at this test site, so the improvement is expected to be much smaller when averaged over a year.

In evaluating long-term adoption of wake steering, operators also will need to know how it affects reliability. "Over a 10-year period of operating the turbines in this mode, what could the long-term impacts be on the blade health, et cetera?" he asked. "Those are important questions to consider."

Back to Top

Design for Steering

Although the results from existing farms are promising, "the bigger impact is in how we design future wind farms," Dabiri said. To date, "most wind farms are designed conservatively, such that the turbines are spaced far apart from one another," which is one reason the increases are modest.

Fleming agreed that as operators become comfortable they can mitigate wake losses, it could open "opportunities for densification of wind farms," perhaps significantly. More speculatively, there may even be ways to harness the wake interactions. "When we first modeled wake steering, it was more or less as a horizontal displacement of the wake," and the goal was to "navigate these wakes into the gaps between other turbines" Fleming said. "But when you look at the three-dimensional flow out of CFD (computational fluid dynamics), there's an additive effect to wake steering because of the generation of counterrotating vortices that persist through the flow." These vortices could suck down faster, higher-altitude winds, which he described as "different from just avoiding wake losses."

Dabiri suspects these interactions could be even more important with vertical-axis turbines, although so far such designs are less mature and reliable. "Vertical-axis turbines individually tend to be less efficient," Dabiri acknowledged, but "they perform better when they are in close proximity. We see possibilities of 10X improvement, as opposed to 10% improvement."

Even without such dramatic enhancements, however, the combination of real-time yaw-control algorithms for wake steering and simulations to improve the collective output of entire farms look to help drive the continued growth of wind farms and their implementation at high densities in previously inhospitable terrain.

* Further Reading

Howland, M.F., Lele, S.K., and Dabiri, J.O.
Wind farm power optimization through wake steering, Proc. Nat. Acad. Sci. 116, 14495 (2019), http://bit.ly/36FvZx2

Fleming, P., et al
Initial results from a field campaign of wake steering applied at a commercial wind farm – Part 1, Wind Energ. Sci. 4, 273 (2019), http://bit.ly/32jZz7J

Renewable & Alternative Energy, U.S. Energy Information Administration, https://www.eia.gov/renewable/data.php#wind

Wind Energy Research, U.S. National Renewable Energy Laboratory, https://www.nrel.gov/wind/

Back to Top

Author

Don Monroe is a science and technology writer based in Boston, MA, USA.


©2020 ACM  0001-0782/20/3

Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and full citation on the first page. Copyright for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, to republish, to post on servers, or to redistribute to lists, requires prior specific permission and/or fee. Request permission to publish from permissions@acm.org or fax (212) 869-0481.

The Digital Library is published by the Association for Computing Machinery. Copyright © 2020 ACM, Inc.


 

No entries found