[M-AG32] Renewable Energy

[MAG32-02] Advances in retrieving bias-free turbulence measures with Doppler wind lidars

*Alfredo Pena¹, Jakob Mann¹, Charlotte Bay Hasager¹ (1.Technical University of Denmark)

Keywords:Turbulence, Wind lidars, Remote sensing, Wind energy

Doppler wind lidars, which are laser-based devices, are currently the most advanced measuring technique for remote sensing of winds and turbulence. In wind energy, lidar technology has replaced traditional anemometry for several applications. Floating lidars, which consist of a lidar wind profiler on a buoy, are the standard for offshore wind resource assessment, and scanning lidars are required to determine flow characteristics at multiple locations within an area or at inaccessible locations. However, wind lidars, like any other remote sensor, suffer two main challenges when attempting to measure turbulence. The first relates to what is commonly referred to as turbulence contamination: wind lidars measure the radial or line-of-sight or beam velocity along the laser beam and so a scanning pattern is needed to reconstruct the velocity components. Since the measured radial velocity is almost never aligned with the wind as this rapidly changes in time and space, the radial velocity second-order statistics will include elements (variances and covariances) involving at least two of the velocity components. Depending on the scanning configuration, such elements can increase or decrease the beam variances. The second relates to what we refer to as turbulence filtering, which appears because with a lidar (as with any sensor) we do not measure at a point but over a probe volume. One can estimate the amount of turbulence contamination and filtering, but this requires knowledge about the turbulence characteristics; since we want to determine these characteristics with the lidar measurements, the problem becomes paradoxical. Peña et al. (2025) suggested a possible countermeasure of this paradox, which makes use of numerical datasets generated with a physics-based model for the turbulence measured by a conical scanning wind lidar profiler covering a wide range of turbulence conditions. The physics-based datasets are then used to train neural networks (NNs) and these NNs are then used to predict “true” turbulence values using lidar measurements.

Here, we extend this approach to measurements from a six-beam wind lidar profiler. The six-beam lidar scans the atmosphere with a beam pointing vertically and five beams at equally separated azimuthal positions along a cone. Computing turbulence using the radial velocity variances of each of the six individual beams eliminates the challenge of reconstructing winds for turbulence computation, thus alleviating issues with regards to turbulence contamination. However, the instrument measures over a probe volume so filtering is a challenge and, as described, the degree of filtering depends on the turbulence characteristics. We therefore generate physics-based datasets of the radial velocity variances of all beams covering a wide range of conditions and train NNs with these. We plan to use measurements from a six-beam lidar that was placed beside a 250-m meteorological mast to predict the true turbulence values and compare those with the turbulence measured by the anemometers from the mast.

Peña, A., Yankova, G. G., and Mallini, V.: On the lidar-turbulence paradox and possible countermeasures, Wind Energ. Sci., 10, 83–102, https://doi.org/10.5194/wes-10-83-2025, 2025