When Winds Get Rough
PSC Experts and Resources Enable Simulation of Non-steady Forces on Wind Turbines
Rough winds are an issue for generating electricity from wind power. They can cause early failure of turbine components that limit the method’s monetary bottom line. Researchers formerly at Penn State University, working with real wind-turbine data from GE Global Research and with the help of PSC experts, used high-performance computers at PSC and its partners in XSEDE to simulate the wind loads driving a 1.5-MW GE wind turbine. They discovered three different types of non-steady changes. In particular, a set of strong, sub-second changes in forces they found may contribute to early failure of bearings. Their results offer the possibility of better wind-turbine component design that extends component lifetimes and improves economic feasibility for wind power.
Why It’s Important:
Wind power is clean and renewable. In the right geographic areas, it’s also dependable. The wind drives a giant propeller, or “rotor,” like the windmills of old but much more efficiently. The blades turn a driveshaft, which in turn transfers that mechanical energy to a generator that creates electricity. Simple. But the winds that create wind power can also buffet the wind turbine with violent changes of speed and direction.
Circling pockets of air, called “turbulent eddies,” put wind turbines through their paces. These winds can cause large, rapid variations in the forces that drive the wind turbine to create the electricity we use. Eddies are particularly strong on days when the sun is baking the ground, causing the ground to heat powerful updrafts of air. Over time, that buffeting wears down the turbines’ internal parts. The bearings that hold the shafts and gearbox see particular punishment. Early failure of components is a major financial drain on wind-energy plants, limiting the expansion of wind power.
Image Left: Development of the flow over and in the wake of an individual wind-turbine blade after half blade revolution (looking from the front of the wind turbine).
“[Wind] forces interacting with individual turbine blades combine to create torques and bending moments on the hub, rotor, shaft, gearbox and generator. These fluctuations in torque cause fluctuations in what we call ‘out-of-plane bending moments’ … What this work addresses is how large those ‘non-steady responses’ are, how rapidly they change, what are their characteristics and how they are tied to the turbulent eddies in the winds that pass into front row [of wind turbine arrays.]”—James G. Brasseur, University of Colorado, Boulder
A team led by James G. Brasseur at the Pennsylvania State University (now at the University of Colorado, Boulder) and Tarak Nandi (now at the National Institute of Standards and Technology), in collaboration with Andreas Herrig at GE Global Research in Munich, worked with experts at PSC and their colleagues at the Texas Advanced Computing Center (TACC) on this problem. The team used high-performance computers (HPCs) at PSC and TACC to optimize their “large-eddy simulation” studies of the effects of atmospheric turbulence on wind turbines. Their end goal is to increase the lifetime of wind-turbine components in response to these non-steady wind stresses. If components last longer, wind energy will cost less and be feasible in more places.
How PSC and XSEDE Helped:
The researchers began with a gold-standard dataset in their pockets. GE Wind Energy had performed a unique field experiment in Germany. They had collected measurements of the changing wind direction and speed experienced by rotor blades on a full-scale 1.5-megawatt GE wind turbine in the field. The Penn State group used this dataset to ensure their simulations produced realistic results and to extend the analysis of their simulation data.
With the help of PSC’s Anirban Jana, the team carried out a series of simulations on HPCs there. These included the former Blacklight and the current Bridges systems. The researchers simulated the transfer of forces across the very thin “blade boundary layer” just above the surface of the turbine’s blades. The boundary layer is incredibly thin—a few micrometers, or a thousandth of an inch. But this layer proved to be important to the wind forces on the rotor blades. The researchers’ simulations showed a stunning range of scales, varying from boundary-layer thickness to hundreds of yards. Both the rotors and the turbulent eddies that hit them are about 100 yards across.
“The relationship with Anirban was really important. We had so many problems [developing the simulations]; XSEDE came to the rescue. If it weren’t for XSEDE, we wouldn’t have been able to do it.”—Tarak Nandi, National Institute of Standards and Technology
With help from TACC’s Si Liu, they used the Stampede1 system there to simulate the chaotic interactions of large-scale eddies with the wind turbine. The Stampede1 simulations covered a longer time scale in a simplified way. The Bridges simulations were of shorter duration and used a much more accurate and detailed model. The researchers carried out both studies using the GE field measurements as a benchmark. This ensured the simulations provided results that mirrored the real-world results.
“Bridges proved to be almost twice as fast as [other systems available]. This helped a lot because [we] had almost 16 to 20 million cells to compute on the order of 10-microsecond timescales.”—Tarak Nandi, National Institute of Standards and Technology
The team found three major time scales in the non-steady forces affecting the turbines. Turbulent eddies passed by the blades on a 25- to 50-second scale. The blades rotated on a 3-second scale, adding a sideways force to the incoming wind. Finally, the researchers discovered large, sub-second changes in force caused by internal winds within the eddies. Of the three findings, this last was possibly the most important. That sub-second force ramps up quickly and violently and probably puts enormous stress on the turbine’s bearings. It may be a particular target for improving turbine-component durability. The researchers reported their results in 2017 in the journal Philosophical Transactions A of the Royal Society (DOI 10.1098/rsta.2016.0103). Their current work focuses on simulating the effects of the forces they’ve discovered on the components of the turbine.