To address this problem, researchers from Argonne National Laboratory, the University of Texas at Dallas and the University of Utah conducted a study of the variability of the velocity and turbulence of wakes generated by onshore wind turbines. The study was featured on the March/April 2022 cover of the Journal of Renewable and Sustainable Energy.
The researchers used light detection and ranging (LiDAR) data together with meteorological data collected from a meteorological tower and supervisory control and data acquisition (SCADA) data.
“In this case, insufficient data is not the problem,” said Romit Maulik, an assistant computational scientist in Argonne’s Mathematics and Computer Science Division. “What is challenging –- and computationally demanding –- is the analysis of the enormous amounts of real-world experimental data to process in order to identify trends, data variability and features.”
The approach taken by the researchers involved representing the data on a lower-dimensional subspace. More specifically, they projected the numerous data samples onto a suitable basis obtained with so-called proper orthogonal decomposition, or POD, with the aim of reducing the dimensionality of the problem. The challenge here was to ensure that the subspace had a suitable reduced representation to enable a simplified analysis while permitting detection of the dominant components.
“Selecting the appropriate POD modes was crucial to enabling an efficient reconstruction of the wake variability. In contrast with the usual approach that just retains POD modes based on captured variance, we used physical arguments to select our POD modes, since experimental data has several sources of nonphysical bias,” Maulik said.
Next, the coefficients associated with the POD nodes were used for a clustering algorithm, which generated subsets of the LiDAR data that the researchers grouped together based on the turbine operational conditions. Once the clustering was done, the researchers analyzed the mean velocity and turbulence intensity fields associated with the wakes of the various clusters.
The results showed that the majority of the clusters were characterized by a quasi-symmetric wake. But in two clusters, the researchers detected a significant transverse deflection of the wake center.
“This would not have been detected without the data-driven approach of our clustering algorithm, and it might be important for the power efficiency of the entire wind farm,” Maulik said.
The clustered data also was analyzed in terms of wake-added turbulence. In the far wake, the researchers detected a decay in the wake-added turbulence intensity. Since this could affect performance, the researchers recommend further modeling studies of wake interactions with relatively small streamwise spacing among wind turbines.
For the full article, see G. V. Iungo, R. Maulik, S.A. Renganathan, and S. Letizia, “Machine-learning identification of the variability of mean velocity and turbulence intensity for wakes generated by onshore wind turbines: Cluster analysis of wind LiDAR measurements,“Journal of Renewable and Sustainable Energy 14(2), 023307 (2022); https://doi.org/10.1063/5.0070094
