You are here:Research

Research


My main research interest is in:

  • Numerical simulations of wind turbine wakes
  • Stability of the tip spiral originating from a wind turbine
  • Simulations of wind turbine wake interaction
  • Simulations of production variation inside large wind farms
  • Optimization of large wind farms.
My research is mainly performed in collaboration with:
 
Prof. Dan Henningson, KTH, Prof. Jens Sørensen and Ass.Prof. Robert Mikkelsen at DTU.
  
 

Overall, performed work has resulted in increased knowledge about the flow structures behind wind turbines and the interaction of a number of turbines. The work has also established numerical methods utilizing possibilities to simulate production variation inside large wind farms, a necessary requirement in order to optimize the large wind turbine farms of the future. A third result of our work demonstrated a relationship between turbulence intensity and wake length, making it possible to optimize the spacing between turbines in wind farms.

Collectively, our research has established new simulation possibilities for the next generation of wind farm development. However, since these simulation methods are dependent on large computers and still demand extensive simulations, future work should concentrate on implementing these methods using an engineering approach to industrial codes. This does, however, require further development and especially computer resources to simulate more cases than have been possible within this project, i.e., to run simulations of a number of velocities and turbulence intensities corresponding to more complex terrain, and verifying these simulations with measurements. Engineering methods could then be based on a database with data from a wider regime of cases. Additional understanding of what occurs during interaction between a number of turbines can be achieved by running wake interaction studies using the actuator line method, instead of actuator disc methods, utilizing the computational possibilities of tomorrow. Such simulations, in conjunction with detailed wind tunnel measurements, could provide knowledge on how to control the breakdown of the wake behind the turbines and give further possibilities to optimize the placement of wind turbines in a park.

 

 

  • Numerical simulations of wind turbine wakes

Knowledge of wind power technology has increased over the years. Lanchester and Betz were the first to predict the maximum power output of an ideal wind turbine. The major break-through was achieved by Glauert who formulated the Blade Element Momentum (BEM) method in 1935.

The design codes of today are still based on the Blade Element Momentum method. It has however been extended to allow for dynamic events, with patch work and ad hoc engineering methods, sometimes of doubtful quality.

Therefore, the aerodynamic research is today shifting toward a more fundamental approach since the basic aerodynamic mechanisms are not fully understood and the importance of accurate design models increases as the turbines are becoming larger.

Recently, complete Navier-Stokes calculations have been performed and today supercomputers offer new possibilities.

The objective of our research is to evaluate existing aerodynamic simulation methods in order to run simulations that provide solutions satisfactory for evaluating the flow field behind one or a number of turbines, i.e., the wake and the wake interaction. From these simulations the basic physical behaviour is studied.

We perform all simulations with the code ElliSys3D developed at Riso and Denmarks Technical University(DTU). All simulations and methods are developed in cooperation with DTU.Figure 1, x=0-plane, pressure distribution; y=0-plane, streamwise velocity; iso-surface, constant vorticity.

One method used is the so called actuator line model (ACL) where the blades are represented by body forces. It that way we save a lot of computational power since we do not neet to resolve the blades and it boundary layer.

Using this method, the flow in the wake can be analysed. Figure 1 shows the structure of the wake. The figure depicts pressure distribution, velocity distribution and the spiral structure here identified by iso-surfaces of the vorticity. Figure 1 therefore contains significant information. Data are available for the entire 3D-domain. The resolution is of course dependent on the resolution of the computational mesh. This type of field would be impossible to reach with measurements. There are however powerful measurement tools available such as PIV (Particle Image Velocimetry). Still, it would be impossible to map an entire 3D field thus indicating that CFD (Computational Fluid Dynamics) opens up new possibilities. CFD simulations of this size can today be made on super computers during a time period in the order of weeks. It is then possible to study the influence of different parameters by running the simulations repeatedly. The same type of wind tunnel measurements is very expensive compared to the use of supercomputers today. Modern supercomputers are commonly clusters of many standard PC machines combined in a network that allows large tasks to run on a number of processors simultaneously. However the computational code has to be written in a parallel form to allow a task to run on many processors simultaneously.   

From the field plotted in figure 1 it is now possible to extract data for different evaluations.

 

 

  • Stability of the tip spiral originating from a wind turbine

In our research we try to understand the basic mechanisms resulting in the breakdown of the flow structure in the wake. This becomes especially important to understand when looking at interaction between turbine wakes. This knowledge becomes increasingly critical for correct placement of each turbine and spacing between each turbine in the large wind power plants being planned and built today. When building offshore parks, there is often an area with shallow water where investors want to concentrate as much production as possible. Onshore, there are frequently some limitations on area caused by other factors. To be able to optimize the number of turbines and their positions, knowledge about the length of the wake behind each turbine is necessary. Working knowledge about the basic mechanisms behind breakdown of the distinct tip vortex is therefore important.

To be able to observe instabilities in the wake structures, a harmonic perturbation is introduced close behind the blade tip. The perturbation is introduced by including a body force. By changing the amplitude and frequency of that perturbation the response from the spiral system could be evaluated. The result shows that the growth rate of the introduced perturbation was dispersive, i.e., different frequency regimes results in different type of modes. When changing the amplitude for a specific frequency it is possible to conclude that the growth rate is not related to the perturbation amplitude, meaning that a linear growth rate could be extracted. A larger amplitude leads however to an earlier breakdown of the spiral structure since it has a larger amplitude at the starting point. This also makes it possible to relate the introducedFigure 2, An iso-surface of the vorticity. perturbation to turbulence intensity. This analysis thus results in a relationship between the ambient turbulence and wake length. It is difficult to quantify the turbulence intensity level to a specific perturbation level, but the general physical behaviour can be explained. The result shows that the relationship between the turbulence intensity. Our studies of the influence of the perturbation frequency resulted in an identification of two types of modes. In the first mode the entire spiral system is oscillating in phase. That is, all spirals are extending from the free stream velocity in the same direction at some azimuthal position. This can occur in all three directions (axial, radial and azimuthal). The result did however show that the main extension was in axial and radial directions.

The second mode consists of out of phase motions of the spirals. This mode corresponds to the case when every second spiral extends from the free stream velocity in opposite directions at some azimuthal position. That is, every second spiral extends in positive axial or radial directions while the other spiral moves in the opposite direction at some point. When identifying different modes, the interaction of the two spiral arcs located closest together is of most interest. The complex spiral system consists of three spirals within each other but the interaction is done in pairs. Therefore, when analysing the interaction the local interaction between the spiral arcs and its closest spiral arc is identified regardless from which of the three spirals it originates.

The results show that the mode with every second spiral out of phase results in the largest growth rate. In both cases the growth results in pairing of the vortices due to non-linear effects. The mode with every other spiral out of phase, results in an earlier breakdown of the spiral system. That mode occurs at around 0.66 and 1.64 Hz as can be identified in figure 2.

Figure 2 identifies the spiral system by an iso-surface of the vorticity. The colour coding at that surface corresponds to the axial position. The mode with every other spiral out of phase described above can clearly be identified in the figure. It is possible to see how the instability grows from about z=17.5 which corresponds to about 4 radii behind the turbine. The unit of z is [R], i.e., the radius of the turbine. In this case the varicose mode has a wavelength that corresponds to 4.5 wavelengths for one revolution.  

The second step of the project enhances our understanding of the basic breakdown of the wake structure and provides important data about the effect of different spacing between turbines.

 

 

  • Simulations of production variation inside large wind farms

Our goal is to be able to simulate an entire park. Clear perception about a suitable method and its limitations and basic mechanisms behind the breakdown of the flow structure, etc., are however necessary before setting up an advanced simulation model of a park. When that is achieved and a simulation model for an entire park is created, studies can be made not only on how to optimize one or two turbines but also on clusters of many turbines. This opens possibilities to study how local energy extraction, turbine spacing, yaw angle and park design affects total park efficiency. The results from wake interaction studies will also be important from a fatigue load point of view. Figure 3 shows an example where we use a turbulent atmospheric boundary layer.

 

Figure 3, Vorticity at hub height. The figures illustrate different flow directions. The flow direction is identified by a number of stream-lines. (a) 265°, (b) 270°, (c) 285°.

Figures 4a and 4b show an example with 9 turbines. In the left figure, the turbulence intensity is 0% and in the right figure the turbulence intensity is 5%.     

 Figure 4, the figures illustrate 9 turbines with and without atmospheric turbulence. The figures are generated by an iso-surface of the vorticity. Figure 4, the figures illustrate 9 turbines with and without atmospheric turbulence. The figures are generated by an iso-surface of the vorticity.

 

 

 

 Figure 5, the figures illustrate 9 turbines with and without atmospheric turbulence. The figures show downsream velocity at hub height. Figure 5, the figures illustrate 9 turbines with and without atmospheric turbulence. The figures show downsream velocity at hub height.

 

  

 Figure 6, the figures illustrate downstream velocity at a vertical cut. Figure a shows results when the wind direction is aligned with the turbines. Figure b shows an wind direction 15 degrees from the alignmet of the turbines.s depicted.

Figure 6 shows a 3D field of all simulated turbines. The wakes are illustrated by an iso-surface of the vorticity. Note, that what appears to be the ground surface, is the same iso-surface as that which appears locally around each turbine and it is located at a height above the ground surface. The colour coding depicted at the iso-surface represents the pressure, the levels can be identified by the legend.     

 

 

 

 

  • Optimization of large wind farms.

When designing large wind farms there are many parameters to consider in optimizing cost efficiency. Offshore the foundation cost is strongly related to water depth. The position of the turbines therefore becomes important. It is not unusual that the planner of the park has a limited area with reasonable water depth. (Onshore, similar limitations appear because of houses, urban areas, restricted areas, etc.) Therefore there is an optimization of water depth versus losses in production caused by wake interaction. The overall objective is therefore to use information from single wake simulations and stability analysis to determine how the production varies inside large wind farms. That knowledge will provide important guidelines on the relationship between park losses and distances between turbines inside a wind farm. That relationship will however depend on parameters such as turbulence intensity and geometry of the farm related to the main wind direction.   

You can find more information under projects