Better Computer Models Needed for Mega Wind Farms
With wind power getting cheaper, wind farm developers are drawing up plans for farms an order of magnitude bigger than anything around today, some with more than 1,000 turbines. But there’s one big problem: the economics of wind farms depends on accurate predictions of power output, and it is far more difficult to model how such large wind farms will behave.
At the scale of several hundred to over a thousand wind turbines, simulating the interactions between so many wind turbines, in a range of different weather conditions, can be too complex for current computer models. The issue could have profound implications for the cost of wind power and its ability to scale up to replace large amounts of fossil fuel.
About five years ago, poor data gathering and ineffective computer models meant wind developers were prone to overestimating the energy production of their farms by over 10 percent, enough to destroy profits and in some cases prevent them from making loan repayments. Models have improved, but the jump to larger wind farms, and the increasing prevalence of ordinary wind farms built close enough together to interfere with each other, is raising the issue again.
Some researchers and industry experts say that while computer models have become good at estimating the production of typical wind farms of between 50 and 100 wind turbines, they will have trouble at much larger scales. “If you were to go to 1,000 wind turbines, the industry’s tools would start to break down,” says Keith Longtin, general manager of product management at GE Power and Water.
With billions of dollars on the line, developers of the new, very large wind farms are taking extraordinary measures to try to predict power output.
Construction on one proposed 1,000-turbine mega wind farm, the Chokecherry and Sierra Madre project at the Overland Trail ranch in Wyoming, is expected to begin next year.
The project’s developers have measured wind speed, direction, temperature, and other weather conditions at the site for five years (two years is standard). They’ve gathered data from 32 sensor-studded meteorological towers that rise 60 to 80 meters above the ranch’s rolling, sage-brush covered hills, along with measurements from sound-based radar (called sodar) that can scan the wind at altitudes up to 200 meters, the height of some wind turbines.
The developers also have used supercomputers to run the data through the latest computer models, ones developed to factor in the way wind turbines interact with one another—as wind moves through them, the turbines create wakes that can interfere with the performance of downstream turbines. They concluded that the wind farm could generate at least 8.76 billion kilowatt hours a year, or enough power for 770,000 homes.
One problem with the current models is they don’t accurately represent the variability of wind, not just at ground level, but even hundreds of meters above the level of wind turbines. Recent research suggests that in some weather conditions, “models can dramatically underestimate wake losses,” says Michael Drunsic, senior consultant at DNV KEMA, which helps developers estimate wind power production. “The wakes carry a lot further than previously estimated,” he says. This isn’t a big problem in wind farms that have only one row—such as those that follow ridges. But with large wind farms with multiple rows, the longer wakes could affect many wind turbines, lowering their output.
These sorts of findings can impact how wind turbines are arranged, especially in larger wind farms with many rows of turbines. Charles Meneveau, a professor of mechanical engineering at Johns Hopkins University, has developed models of the way very large wind farms disturb the air up to a kilometer above them. Based on some of his simulations, he’s shown that turbines should actually be placed twice as far apart as they usually are to get the most out of the wind. But he says his current models still can’t accurately predict very large wind farm performance—they look only at performance in average wind speeds, and wind turbine performance can vary greatly depending on differences in wind conditions, he says.
Yet while new computer models are identifying challenges for large wind farms, they’re also bringing some good news, Meneveau says. He’s shown, for example, that turbulence, while it can be a problem, is also essential for large wind farms. Without any turbulence, the first rows of a wind farm would essentially block the wind, limiting the number of rows that could be installed. He showed that turbulence actually pulls wind down from above the wind farm. As a result, the last row in a wind farm, while it gets somewhat less energy than the first row, can still generate electricity, as long as wind turbines are properly spaced. “The bad news is that for large wind farms, people didn’t know quite how to handle turbulence,” he says. “The good news is that you can make the farm as big as you want, because you can pull down air from above.”
Meneveau says the models can also be used to optimize the design and operation of wind turbines. Those in the front row, for example, might be programmed to orient their blades to allow more wind to pass by them, thereby improving the performance of turbines in subsequent rows, and so increasing the output of the entire wind farm. Related research is being conducted at the National Renewable Energy Laboratory in Golden, Colorado (see “Novel Designs Are Taking Wind Power to the Next Level” and “Will Vertical Turbines Make More of the Wind?”).
Improved computer models will be crucial to lowering the cost of wind power, Drunsic says. “The more accurate the models and the more certain a developer can be, the lower the cost of financing,” he says.
Geoffrey Hinton tells us why he’s now scared of the tech he helped build
“I have suddenly switched my views on whether these things are going to be more intelligent than us.”
Meet the people who use Notion to plan their whole lives
The workplace tool’s appeal extends far beyond organizing work projects. Many users find it’s just as useful for managing their free time.
Learning to code isn’t enough
Historically, learn-to-code efforts have provided opportunities for the few, but new efforts are aiming to be inclusive.
Deep learning pioneer Geoffrey Hinton has quit Google
Hinton will be speaking at EmTech Digital on Wednesday.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.