Posters

Abstract

We assess the impact of spatial sampling on the accuracy of wind speed and turbulence measurements obtained from ground-based wind lidar profiling systems. Such systems execute a conical (velocity-azimuth display, or VAD) scan, sampling the wind around a scan disk at chosen heights. The lidar assesses a number of line-of-sight (LOS) wind speeds around the disk and uses this to calculate the 3D wind vector under an assumption of uniform flow via a wind field reconstruction algorithm (WFR). For some types of lidar system, data from throughout the disk contributes to the calculation, whereas other systems sample a select number of azimuth angles.      In uniform flow, excellent wind speed results can be obtained when the flow is sampled with fewer than 10 LOS values. However, when the flow complexity increases due to the influence of complex terrain or high turbulence levels, then the impact of the uniform flow assumption in the conduct of the WFR algorithm leads to greater uncertainty and possible bias.     The sampling effects have been investigated through Turbulence Box modelling - a wind pattern is generated via NREL’s TurbSim  and the LOS spectra for continuous-wave profiling lidar at different scan frequencies are calculated by modelling the continuous lidar probe as a discrete distribution of weighted samples.  The wind vector is calculated via WFR and averages are compared via RMSE to values representative of the desired signal (turbulence box input parameters and samples representative of wind resource assessment). The comparison is averaged over many Turbulence Box seeds, making them representative of the errors involved in the measurement process, rather than the noise inherent in turbulence. Care is taken to control for other factors directly related to simulation constraints, particularly the discretisation of the wind signal and lidar probe geometry.    Preliminary results suggest that the errors between lidar measurements of wind speed and actual values directly above the lidar are similar across different spatial sampling schemes. When compared to a single point in the middle of the scan (i.e. a cup measurement), RMSE increases linearly with TI, ranging from 0 to ~12% for TI between 0 and 25%, for all sampling schemes. Likewise, taking an average of measurements across the width of the scan, gives similar results, but with a maximum error of ~5%. In contrast, the difference in errors between few spatial samples (<10) and many (50) is more significant when comparing to the average of measurements around the VAD scan. Here, maximum errors amount to ~4% for <10 spatial samples, in contrast to ~ 2.5% for ~50 spatial samples. The accuracy of the LOS values is important for CFD corrections in complex terrain as well as the estimation of turbulence intensity.     Several simplifying approximations have been made; work is underway to refine the analysis and fully account for real-world complexities. The analysis will also be extended in future to investigate non-uniform flow in highly complex terrain, where effective spatial sampling assumes higher importance.