Code underlying the publication: Wind pattern clustering of high frequent field measurements for dynamic wind farm flow control

Code used to generate the wind direction time series used in the publication "Wind pattern clustering of high frequent field measurements for dynamic wind farm flow control" by M. Becker, D. Allaerts and J.W. van Wingerden (in preparation for the TORQUE conference 2024)

The TenneT_BSA_* files convert the raw data from the KNMI [1] into one file with all data at 119m height. This is equivalent to the hub-height of the DTU 10MW reference turbine. Note that there is a channels switch in the data. That's why there are two functions to read in the data.

The output dataset is given in the CombinedDataAt199m.csv file.

The two hpc06_trajectories_* files are then used to segment the data into time series of requested length. This code also contains the filtering and interpolation of the data. The output are two .csv files, one with wind direction trajectories and one with wind speed trajectories.

Two examples are given by WindDirTraj.csv and WindVelTraj.csv - they have been generated with a length of 30 data points and with an offset of 30 data points (no overlapping).

The code of hpc06_cluster_dir* can then be used to cluster the given data.

The remaining files are supplementary to plot data, to calculate distances in radial data etc. including the kmeans360.m function which is the modified function of the Matlab kmeans function which also works for radial data.

[1] https://dataplatform.knmi.nl/dataset/windlidar-nz-wp-platform-1s-1