How New Methods Developed for Rotorcraft are Making This Possible
While wind turbines were still using blade designs of the past, with generic, suboptimal shapes that neglected complex terrain and multiple turbine effects, aircraft and rotorcraft computational fluid dynamics (CFD) was already starting to compute and optimise wing design and rotorcraft multiple blade flows. Recently, there has been promising new CFD research that could have a large impact on wind turbine siting and blade design. This article describes some of this research.
By Dr John Steinhoff, Professor at University of Tennessee Space Institute and consultant to Flow Analysis Solutions, Inc., USA .
The closest configuration to the wind turbine in the aircraft world is, of course, the rotorcraft. Below we focus on new developments in rotorcraft CFD and their incorporation into wind turbine blade and optimal site design.
The Power of CFD Until Now
Until recently most advances in CFD aerodynamics involved fixed wing aircraft, and essentially neglected the details of the wake. This earlier CFD could not treat the strong swirling flow (vortices) in the wake, and was not applicable to rotary blades (such as helicopters and wind turbines), where each blade is strongly affected by the vortices of the preceding blade. New technology described below is now enabling accurate, rapid computations of these flows.
Even though it could not treat the wake details, earlier CFD could make accurate, reliable predictions for flows where vortical regions in the boundary layer and wake are not dominant. Thus, starting with the first sufficiently powerful computers in the 1960s, a foundation of CFD methods, supplemented by models for vortical regions, was built up. For aerodynamics, the vortical regions are turbulent and had to be modelled. The main point is that these models involved complex mathematical equations that had to be solved numerically in very thin regions. These regions originate in the ‘boundary layers’ near the surface and convect away in the wake as thin vortex filaments (for rotorcraft applications, see Figure 2).
In the earlier CFD, these model equations were solved with mathematical equations requiring very fine computational grids (points where the solution is sampled). The computed solutions, even though they only treated approximate (often ad hoc) model equations in the vortical regions, generally required about a hundred times more computing time than the non-vortical regions (the rest of the flow).
New Developments in CFD
The new developments can be divided into three types: (1) new combinations of existing older methods; (2) extensions and improvements of known methods using rearrangements; and (3) radical new improvements. In this section, we review these as options for wind turbine simulations. It should be noted that these methods, which were originally developed for rotorcraft, have high accuracy and reliability requirements since they were to be used for manned vehicles.
For a wind turbine site computation VC has a very important feature. This feature enables it to compute flow over very complex topography with no additional computational requirements and automatically to input the topographical data (e.g. from satellite measurements) in minutes. This can be contrasted with method classes (1) and (2). These require surface-fitted computational grids, which are very time consuming to generate and difficult to use. By contrast, VC uses a uniform grid, which requires no special software or user intervention or longer computations. In VC the surface is simply ‘immersed’ in this grid and VC provides an effective turbulent boundary layer along the ground. A vorticity iso-surface of the solution is shown in Figure 3. In this computation, the vortices are captured over only three grid cells. This allows the use of coarse grids, which accounts for low computing time.
Problem of Multiple Wind Speeds
For some particular problems, such as flow over a typical transport aircraft in steady cruise, a relatively small number of conditions must be satisfied. This is because the aircraft is developed for optimum performance at cruise conditions within about 1% of design and the development period extends over a number of years, during which the vortical flow model can be developed and tuned to match the experimental data. Unfortunately, the atmospheric conditions where wind turbines (and rotorcraft) operate can be highly variable.
One hope for some models was that the time average (over some appropriate time intervals) could be modelled for practical results. If the wind turbine power were a linear function of the wind speed, this could work, because the average power could be directly calculated from the average wind speed. However, since the wind turbine power varies at a rate close to the third power of the instantaneous wind speed, the average generated power will be very different from that computed based on the average wind speed. Because of this problem, the most reliable computation for power should probably come from a sampling approach involving a number of runs with different conditions. These sampling approaches require hundreds of computer simulations to obtain reasonably accurate and reliable results. This approach leaves out the conventional CFD models. However, the VC model, due to its inherent accuracy and speed, can make such a sampling approach feasible.
Biography of the Author
Dr John Steinhoff, Professor at the University of Tennessee Space Institute, is the inventor and ongoing developer of Vorticity Confinement (VC), an advanced CFD aerodynamics model. He and his team of highly qualified researchers have successfully implemented VC in the rotorcraft industry. They are providing technology, experience and scientific consulting to the wind turbine industry through the recently formed company, Flow Analysis Solutions, Inc.{/access}
While wind turbines were still using blade designs of the past, with generic, suboptimal shapes that neglected complex terrain and multiple turbine effects, aircraft and rotorcraft computational fluid dynamics (CFD) was already starting to compute and optimise wing design and rotorcraft multiple blade flows. Recently, there has been promising new CFD research that could have a large impact on wind turbine siting and blade design. This article describes some of this research.By Dr John Steinhoff, Professor at University of Tennessee Space Institute and consultant to Flow Analysis Solutions, Inc., USA .
{access view=!registered}Only logged in users can view the full text of the article.{/access}{access view=registered}
Recent developments in CFD, which are the result of large investments in research from NASA, military budgets and the aircraft industry, are being incorporated into wind turbine design. This is leading to improvements in blade design, blade dynamic control and wind turbine optimal siting (including wind turbine interference), as demonstrated in Figure 1. These modern improvements will prove to be vital in realising the promise to produce low cost energy from wind turbine farms.The closest configuration to the wind turbine in the aircraft world is, of course, the rotorcraft. Below we focus on new developments in rotorcraft CFD and their incorporation into wind turbine blade and optimal site design.
The Power of CFD Until Now
Until recently most advances in CFD aerodynamics involved fixed wing aircraft, and essentially neglected the details of the wake. This earlier CFD could not treat the strong swirling flow (vortices) in the wake, and was not applicable to rotary blades (such as helicopters and wind turbines), where each blade is strongly affected by the vortices of the preceding blade. New technology described below is now enabling accurate, rapid computations of these flows.
Even though it could not treat the wake details, earlier CFD could make accurate, reliable predictions for flows where vortical regions in the boundary layer and wake are not dominant. Thus, starting with the first sufficiently powerful computers in the 1960s, a foundation of CFD methods, supplemented by models for vortical regions, was built up. For aerodynamics, the vortical regions are turbulent and had to be modelled. The main point is that these models involved complex mathematical equations that had to be solved numerically in very thin regions. These regions originate in the ‘boundary layers’ near the surface and convect away in the wake as thin vortex filaments (for rotorcraft applications, see Figure 2).
In the earlier CFD, these model equations were solved with mathematical equations requiring very fine computational grids (points where the solution is sampled). The computed solutions, even though they only treated approximate (often ad hoc) model equations in the vortical regions, generally required about a hundred times more computing time than the non-vortical regions (the rest of the flow).
New Developments in CFD
The new developments can be divided into three types: (1) new combinations of existing older methods; (2) extensions and improvements of known methods using rearrangements; and (3) radical new improvements. In this section, we review these as options for wind turbine simulations. It should be noted that these methods, which were originally developed for rotorcraft, have high accuracy and reliability requirements since they were to be used for manned vehicles.
- Starting in 1986, when Norbert Kroll presented the first attempts at solving the full ‘Euler’ equations, people have been trying to generate reasonable solutions for rotary blade aerodynamics. Even though they completely neglected the thin, vortical boundary layers, they still were not able to solve these equations for the strong tip vortices that are shed from each blade, which, because they pass close to the following one, have a large effect. Since 1986, there have been many attempts at computing flows using these methods with ever larger and faster computers. The latest, at NASA, uses a 256 processor system for 19 hours to get a single solution for 80 million grid points. This, of course, is totally irrelevant for blade design or siting simulations. Furthermore, the details of the vortex solution are more a result of a build up of numerical artefacts than solutions of the exact physical equations, and these still represent a model of the vortices.
- People have rearranged the same equations in attempts to compute the vortex filaments in a more direct manner. Using a complex grid system, they have managed to get closer to the correct vortex interactions. However, the methods are still many times too slow, even on supercomputers, for routine engineer design analysis of blades or site computations. One example of a method known as ‘VTM’ (Vorticity Transport Model) does, however, represent an improvement, even though it still neglects the thin, vortical boundary near the blade surface. Furthermore, as in the section above, the details of the vortex solution are more a result of a build up of numerical artefacts than solutions to the exact physical equations, and these still represent a model of the vortices.
- A more efficient technique of simulating the thin, vortical regions has recently been developed. As in (1)and (2), this represents a model of the thin turbulent regions. Unlike (2), however, it also models the thin boundary layers. This method, known as Vorticity Confinement (VC), requires less computing power than (2) and much less than (1). However, it achieves results that are comparable in validation studies. VC is based on basic physical principles used in the last few decades in studies of condensed matter. These are known as ‘diffuse interface methods’. They take into account the build up of numerical error, unlike (1) and (2). More information can be obtained by searching on the internet for ‘Vorticity Confinement’.
For a wind turbine site computation VC has a very important feature. This feature enables it to compute flow over very complex topography with no additional computational requirements and automatically to input the topographical data (e.g. from satellite measurements) in minutes. This can be contrasted with method classes (1) and (2). These require surface-fitted computational grids, which are very time consuming to generate and difficult to use. By contrast, VC uses a uniform grid, which requires no special software or user intervention or longer computations. In VC the surface is simply ‘immersed’ in this grid and VC provides an effective turbulent boundary layer along the ground. A vorticity iso-surface of the solution is shown in Figure 3. In this computation, the vortices are captured over only three grid cells. This allows the use of coarse grids, which accounts for low computing time.
Problem of Multiple Wind Speeds
For some particular problems, such as flow over a typical transport aircraft in steady cruise, a relatively small number of conditions must be satisfied. This is because the aircraft is developed for optimum performance at cruise conditions within about 1% of design and the development period extends over a number of years, during which the vortical flow model can be developed and tuned to match the experimental data. Unfortunately, the atmospheric conditions where wind turbines (and rotorcraft) operate can be highly variable.
One hope for some models was that the time average (over some appropriate time intervals) could be modelled for practical results. If the wind turbine power were a linear function of the wind speed, this could work, because the average power could be directly calculated from the average wind speed. However, since the wind turbine power varies at a rate close to the third power of the instantaneous wind speed, the average generated power will be very different from that computed based on the average wind speed. Because of this problem, the most reliable computation for power should probably come from a sampling approach involving a number of runs with different conditions. These sampling approaches require hundreds of computer simulations to obtain reasonably accurate and reliable results. This approach leaves out the conventional CFD models. However, the VC model, due to its inherent accuracy and speed, can make such a sampling approach feasible.
Biography of the Author
Dr John Steinhoff, Professor at the University of Tennessee Space Institute, is the inventor and ongoing developer of Vorticity Confinement (VC), an advanced CFD aerodynamics model. He and his team of highly qualified researchers have successfully implemented VC in the rotorcraft industry. They are providing technology, experience and scientific consulting to the wind turbine industry through the recently formed company, Flow Analysis Solutions, Inc.{/access}
