Teaching ‘selfish’ wind turbines to share can boost productivity
The exact amount of energy gained depends on factors such as farm layout and site wind conditions. However, when tested on a commercial farm in India, the algorithm increased energy production between 1 and 3%, depending on wind speed, which would be equivalent to powering 3 million homes if the software were deployed on existing farms around the world, the study authors believe.
And reaching that point isn’t as far-fetched as it sounds. One of the advantages of the approach is its potential for real-world scalability. “Usually, to increase the unit of output, you either have to install a larger rotor or a more powerful generator, or change some of the hardware,” says Xavi Vives, control engineer at wind turbine manufacturer Siemens Gamesa. (Vives was not involved in the study, although Siemens Gamesa staff participated in the research.) “But it’s pure software, so it’s very promising at a very low cost.”
According to Varun Sivaram, one of the study’s co-authors, who at the time served as chief technology officer at renew the power, India’s leading renewable energy company, testing the technology in India was also important. “I wanted to find a way to translate lab-scale technology into real-world experience. And I also wanted to do it in an emerging economy because that’s where the real need for clean energy solutions is, in these emerging economies where energy demand is increasing,” he says.
In addition to increasing turbine power output, the algorithm could also help wind farms by extending turbine life and reducing wear and tear that can decrease their output over time. “I think the most important finding from their study is that if you can equalize the loads, if you can actually let more wind through to the subsequent turbines, you’re going to reduce the wear on the first turbine,” says Mark Z Jacobson, professor of civil and environmental engineering at Stanford University. Vives agrees: “The higher the turbulence, the greater the wear…if you can reduce or move the wake away, you also give the turbines more slack so they can run longer.”
Although the study showed promise, Jacobson believes more experimentation is needed before the software can be deployed, as initial testing focused on a setup involving three turbines under specific conditions. In reality, there are endless potential turbine configurations, wind speeds and topographies, he explains. “I think they need to test more complex setups and try to come up with general rules that apply regardless of the setup,” he says. “You don’t want to try to optimize every turbine and every farm.”
As wind power grows, Sivaram believes algorithms like this will be needed to generate as much electricity as possible. Ideal land sites for wind farms require specific circumstances – places with very fast wind speeds and plenty of land to place the turbines far apart from each other. In the future, wind turbines will likely be placed close together as land becomes less available.