Recent advances in CFD (computational fluid dynamics) codes and the availability of large scale computer “clusters” are guiding the design work of the next generation of wind turbines. Developers should be able to model a turbine’s power output versus wind velocity, and then the power output for a wind farm with direction data. In addition, detailed information on wake turbulence and velocity deficits in large arrays of wind turbines will avoid the under-prediction of power outputs that plagued previous methods. Those are the goals. The devil, as usually, is in the details
Trouble with turbulence
As companies build larger wind turbines, they place an increased emphasis on decreased spacing distances. This gives rise to a twofold problem: Shorter distances can reduce the production from downstream turbines, and result in a significant decrease to service life due to the increased load levels created by turbulence. Field tests and simulations show that operating in a turbulent wake field can increase equipment stress levels by 5 to 15%.
Typical turbine spacing is 6 to 8 times the rotor diameter.
Yellow turbine velocity: The velocity simulation is on a plane through the hub and nacelle. About 9 m/s wind slows a bit at the nose and behind the nacelle and speeds up at the sides.
The wind-resource use factor is independent of rotor diameter and ranges from 2.5 to 9 MW/km2. But a recent study shows that turbulence can be twice normal levels at 10 rotor diameters downstream, along with a 25% velocity deficit. In addition, some early indications are that the largest turbines may be particularly vulnerable to “horizontal wind shear”, which is created by the meandering wakes of upstream turbines.
Such conditions are of increasing concern to offshore developers and OEMs. To vividly illustrate the problem, Ed Salter, CTO and Co-founder of Greenward Technologies, Austin, Texas (greenward-technologies.com), points to an aerial photo of the Horns Rev wind farm, the world’s largest offshore development. The photo shows clouds forming in low-pressure turbine wakes. The image reveals a lot. First of all, the rotating wake, referred to as “swirl”, does not dissipate quickly and extends back many rotors diameters. More importantly, the photo shows the swirl building as it passes through the rotor of each turbine.
“There are no terrain features offshore to break up the wakes, so they persist for long distances,” says Salter. “Also, they are reinforced in an additive manner at each row. Turbines operating downstream of other turbines in highly turbulent flow can experience greatly accelerated fatigue damage to all components in the primary load path.”
Could there be a way to eliminate or reduce the effects of rotor-induced wake swirl? Salter and Greenward CEO Larry Haworth think they have an answer in what they call a Quad Array. It consists of four counter-rotating turbines that feature 3-blade flexible, lightweight rotors that were first developed by Salter at Wind Power Systems Inc. in 1977. The four turbines are mounted on a streamlined “X” frame that rotates to service each turbine. The Quad Array frame also uses what Salter calls “flexible lightweight rotor technology” that almost eliminate “tower shadow” noise and vibration.
Quad Array: places four turbines on crossed arms letting their blades counter rotate to minimize turbulence.
The wake dynamics of the Quad Array are of particular interest, and a simplified wake analysis was done in February of 2008. This was followed by the design and construction of a functional wind tunnel scale model that Salter subjected to a series of preliminary controlled velocity tests.
The results led him to formulate what he calls the Wake Convergence and Swirl Cancellation hypothesis. The concept is simple enough – get the counter rotating wakes to converge and the opposing swirls will cancel each other. The implications are not so simple. “If we can prove this, it will change the industry,” says Salter. “We could be looking at something like a 10-fold improvement in the wind resource use factor, combined with a large reduction in turbulence levels.”
To back up his comments, Greenward has launched a collaborative program to analyze the wake of the Quad Array using the latest CFD codes, along with a comprehensive wind tunnel testing program. The company is looking for qualified collaborators, and plan on presenting their first paper at Windpower 2010.
Wakes from Discs: a detailed site analysis shows a measure of turbulence behind turbine rotors
Learning from larger machines
As OEMs design larger turbines, they are finding variations to the wind in just the area swept by the rotor. “One question is: What is the real loading on the blades?” asks Dennis Nagy, vice president of business development with CAE software developer CD-adapco, Melville, NY (cd-adapco.com). “It’s the fluctuation that matters. The amount of turbulence at the top of a blade versus at the bottom of the rotation leads to vibration and that leads to fatigue. Blades are longer and flex more on larger turbines so their designers also worry about them bending to the point where they could hit the tower. If that happens, the turbine would be shut down to prevent damage. In addition, blades that pass near the tower momentarily pass through slower moving air for a change in loading. That once-per-revolution condition adds to the vibration. So a good vibration analysis ultimately leads to a fatigue-of-composites study and that needs good CFD and wind data.
“If you do a good job on wind-farm analysis and know where the wakes will form and dissipate due to terrain variations, that would be better information to guide turbine placement than to, say, spacing them equally,” adds Nagy.
Here’s another problem with wind-farm predictions and it’s a big one: Almost all farms end up generating at least 10% less power than the values predicted during the planning and financing stage. Understandably, that disappoints owners.
It turns out that prospects often want OEMs to predict that if they sell this many turbines for a particular placement, the turbines will produce this much power. “The wind data is available from all year round for strength, turbulence, and directions. After evaluating all this stuff, OEMs come up with an estimate for how much the farm can yield, assuming downtime for maintenance. And then they guarantee some level of power output. Apparently, if the farms don’t meet the predicted output, the turbine builder pays penalties,” says Nagy. So OEMs guarantee an output. And should a competitor forecast a higher figure, the second OEM might get the job. So an accurate prediction is essential. To assist, Nagy says his company developed what it calls an actuated disc as a way to more economically model a turbine’s wake effects.
Using the company’s CFD software, wind-farm designers would place a stick to represent a tower and mount a disc on it with thickness and diameter to represent the turbine rotating blade set. “Wind hits the disc’s front side, extracts energy, and modified wind flows out the backside. Each model then predicts swirl and turbulence based on previous detailed studies adjusted or calibrated for wind speed. It’s one way to simplify the analysis,” says Nagy. “Although still compute intensive, we’ve done runs with over a dozen turbines on a landscape to examine wake formations.”
An additional challenge for turbine OEMs, says Nagy, has been to integrate CFD software into the work flow of their field-siting tasks. “For example, a prospect wants to build a wind farm on a particular parcel. One European OEM would then examine the wind and weather data for the topology and suggest placing turbines in precise locations. The field engineer, in the past, would take this data and feed it into proprietary in-house software that would make an assessment of each proposed turbine and decide whether it would be a normal or “complex” situation. If the wind is turbulent enough, or if the terrain is rugged enough to produce local wind effects, such as from a canyon or ridge, the in-house software would require a more detailed CFD study for the turbine. Then the field engineer would send wind and topology files to company headquarters and a group there will do the CFD and generate reports and return them to the field engineer. That would take about three weeks, not because it ran for three weeks, but more likely because it took some human intervention, for example, to transfer files among multiple modules (meshing, boundary conditions, solving), adjust the mesh, and monitor the run. After about three weeks, the field engineer had something to present to the client.
Nagy says his company was able to help significantly streamline the process for the OEM. “Now that complex, three-week task is fully automated and reduced to two hours. It’s literally push button. The user supplies the needed files and the STAR- CCM+ software does the rest.” Nagy acknowledges that significant credit for the short run time goes to the OEM running this process on a HPC cluster of some 1,200 cores.
Another task wind farm operators are looking at is forecasted winds. If a lull is forecasted, the owner might schedule some preventative maintenance. But if good wind is in the offing, he might plan on producing power.
::Windpower Engineering::
Discuss these considerations for designing turbine hydraulics at www.EngineeringExchange.com
Filed Under: Featured, Turbines
