Modern wind turbines feature extremely large physical dimensions and fast tip speeds, which result in high reflectivity and complicated spectra contamination, respectively. These make WTC difficult to remove by means of classic clutter filters. In order to find the best way to remove WTC, its radar signature needs to be fully characterized w.r.t. the yaw angle, phase and rotation rate of wind turbines. However, due to the lack of controllability of wind turbines, this has never been done by real measurement. For such a purpose, a scaled experimental system emulating the radar-wind turbine scenario has been built at the Atmospheric Radar Research Center (ARRC) of the University of Oklahoma.
The system includes a 1:100 scaled wind turbine model manufactured specifically for laboratory-based measurement. The movement of the model is fully controlled by a horizontally axial motor and a vertically axial motor. Also, a scatterometer has been built to emulate the weather radar. It is a low-power pulse-Doppler radar featuring fine range resolution better than 1.5 meters. The pulse repetition time has a wide range of adjustability, and it is fully polarized to best approximate the future upgrade of NEXRAD. The system is configured within an anechoic chamber to isolate the direct reflection from the model from other interference. Dynamic measurements have been made with rotor blades spinning at different rate and the model rotated to different yaw angle. The scaled measurement shows clear match with wind turbine radar data as far as the spectrogram is concerned.
A knowledge-based template retrieving method is being studied as a first attempt to mitigate WTC from weather radar observation. The method needs to pre-establish the knowledge base from observations of WTC in clear weather, and then retrieve the Optimal Prediction (OP) in terms of Minimum Mean Square Error (MMSE) aided by the knowledge base. A linear relation is assumed between OP and WTC contaminated radar data, so a Wiener filter is further applied to finally remove WTC on a scan-to-scan basis. The method is validated by the scaled measurement. Significant signal to clutter ratio improvement is obtained by filtering the simulated weather signal super-positioned on laboratory data. Furthermore, the proposed method performs well with limited number of pulses, so is well prepared for current weather radar running in operational Volume Coverage Pattern (VCP) mode.
The method assumes well established knowledge base for all yaw angles of wind turbines, which can be less feasible for some actual radar operations. The technique of adaptively learning and updating the knowledge base will thus be studied in the future.