With larger wind turbines and higher hub heights, SODAR (Sonic Detection and Ranging) is becoming an increasingly popular alternative to the conventional method of collecting wind data with cup anemometers mounted on meteorological (met) towers. SODAR can offer many advantages to anemometry, such as ease of installation and the ability to collect wind data over the entire rotor diameter. SODAR measures wind speeds by emitting high frequency acoustic pulses into the atmosphere which are then reflected by fluctuations in the refractive index of air. The backscattered energy is recorded by the SODAR, the frequency content of the signal is analyzed at heights of interest (range gates) and the wind speed at each range gate is calculated based on the observed shift in frequency. At each range gate, the reflected acoustic pulses are summed up over a volume and the frequency content is determined. The size of the volume will depend on the desired FFT (Fast Fourier Transform) size as well as the sampling rate and can be ~20 m in height. In this paper, a new method for processing SODAR data using CWTs (Continuous Wavelet Transforms) is presented. Wavelet analysis allows the user to continuously analyze both the time and frequency domain simultaneously. Using the complex Morlet wavelet, raw SODAR data was reprocessed and the frequency shifts were determined at every height. The resulting wind speed profile was compared to the calculated wind speeds obtained from the FFT analysis as well as to the wind speeds found from an anemometer on a 40 m met tower. It was found that the wavelet transform approach produced approximately the same wind speeds as found from the FFT approach however it also provided a more detailed snapshot of the returned signal's frequency content. The reason for this is because the CWT was calculated at all heights whereas the FFT was only calculated at the specified range gates. The one disadvantage to the wavelet approach is the computational time required to process the data. It was found that the processing time was ~10 times slower than with the FFT approach and this method will therefore not be feasible until faster processors or more efficient algorithms are available.
展开▼
