From a distance, a wind farm can seem almost placid, turbines turning slowly, steadily, churning out electricity. But there’s more to it than meets the eye.
The wind, though it can seem consistent, often has varying degrees of turbulence that impact wind turbine performance. Heating and cooling change the wind over the course of the day. A wind farm’s turbines interact in ways that reduce performance and add to structural loads on the turbines, increasing maintenance costs and the overall cost of wind energy.
Researchers at the U.S. Department of Energy’s (DOE) National Renewable Energy Laboratory (NREL) are learning how to better understand these issues and are working toward effective solutions for the wind industry. Their goal is to maximize turbine performance and minimize structural loads, which will ultimately result in lower-cost wind energy. Toward that goal, NREL researchers are leveraging the lab’s supercomputing resources and have developed high-tech modeling and simulation capabilities.
An Industry-Wide Concern
The market for wind energy continues to grow — and so do the wind turbines and farms themselves. Unfortunately, the power production of these energy plants has, in many cases, been lower than initially predicted. Wind plant underperformance has become a concern throughout the wind industry and could potentially cost developers millions of dollars over the life of a wind plant because of reduced power generation and increased maintenance costs.
“The wind industry is increasingly concerned with these underperformance issues,” said Pat Moriarty, senior engineer at NREL’s National Wind Technology Center. “The average underperformance is about 10%, with some seeing underperformance as high as 30% to 40%. This adds up to a lot of lost energy and high cost for the industry over the life of a wind plant and presents us with a big opportunity to improve wind plant efficiencies.”
Models Enhance Understanding of Performance Issues
Wake turbulence is a type of instability in the wind flow and is the result of wind flowing through the rotor of a wind turbine. Its effects and how they impact wind turbine and plant performance have not been well understood. To better understand these issues and move toward effective solutions, NREL researchers have developed sophisticated simulation tools to perform large-eddy simulation models that are designed to predict the performance of large wind plants with greater accuracy than any previous models.
Wind plant developers have used design tools going back to the 1980s, which are generally effective for basic optimization of the layout of a wind farm. However, none have been able to simulate with consistent accuracy how wakes propagate and how wind turbines interact with one another.
“Previous models were very simple and don’t capture a lot of the physics — how the atmosphere behaves and how wind turbines respond to changing conditions in the field,” Moriarty said. “It was clear that we needed better models, specifically understanding issues around wake-turbine and atmosphere-turbine interactions.”
Good models need good data. Data from operating wind farms provides validation for the models. So the project has been collecting data from offshore and onshore wind farms in both Europe and the United States to compare to the simulations. “We’re comparing the data from actual wind turbine performance in the field to the predictions from our models,” Moriarty said. “The models have been very accurate and very close to what is actually happening in the field.”
Big Computers Facilitating Big Ideas
The backbone of this new modeling capability is the high-performance computing resources that run the simulations.
Researchers are currently using RedMesa, NREL’s most powerful high-performance computing system, located at DOE’s Sandia National Laboratories and managed in collaboration with Sandia. Peak computational capability of RedMesa is about 180 teraflops, which means it can process 180 trillion floating point operations (flops) per second. For comparison, a basic calculator requires only 10 flops.
NREL will be adding additional high-performance computing capability in the next year with a new supercomputer on the lab’s Golden, Colorado, campus. The supercomputer in NREL’s new Energy Systems Integration Facility (ESIF) will be nearly a petaflop in scale (a petaflop is 1,000 teraflops) and will be the fastest computer system in the world dedicated to renewable energy and energy efficiency technologies.