Minimizing errors in interpolated discrete stochastic wind fields

• A new method for interpolating turbulent wind data is suggested.
• It avoids loss of turbulence intensity experienced by linear interpolation.
• It does not require much more computational effort.
• It shows significantly better conservation of turbulence statistics.

For many unsteady processes (e.g. turbulent wind, electricity demand, traffic, financial markets, space physics, etc.) data is only available at discrete points, be it due to data storage or data gathering limitations. However, derived forms of that data are often used in further studies where the discretization may be different from the discretization of the original data. This paper addresses the question of how to obtain values between discrete data points, for example, when sampling turbulent wind. Linear interpolation is often the standard answer. Yet, it is shown that this is a poor choice for unsteady processes where the sample step size is significantly larger than the fluctuation scale. An alternative employing probability density functions of data increments is suggested. While this new method does not require much more effort than linear interpolation, it yields significantly more accurate results. Unsteady wind is used to exemplify this: turbulent wind speeds on a (rotating) wind turbine blade are synthesized from a coarse data grid via the introduced method of velocity increments. Thus the superiority of the presented approach over linear interpolation is demonstrated – with important implications for blade load and power output computations.

Figure optionsDownload as PowerPoint slide