Conclusiones y Recomendaciones

For the wind turbine nominal operating frequency, fwt, of 0.67 Hz, large amplitude

spikes are expected at ± 0.67 Hz, ± 1.34 Hz (2x fwt), ± 2.01 Hz (3x fwt) and so

on. The length of time taken for the 50 kHz bandwidth signal to be transmitted, the sweep frequency time, is a standard 0.26 s. This is then the sample time for the received signal at each range and each sweep is analysed with an FFT to give one sample per range. The Doppler bandwidth is 1/sweep frequency time and, for the standard 0.26 s, results in a sample frequency of 3.846 Hz and Doppler spectra frequency range of −1.923 Hz to 1.923 Hz. As the Doppler spectrum cuts off at ± 1.923 Hz any frequencies larger than this will appear in the next measurement cell. The third WTI harmonic expected at 2.01 Hz would exceed the immediate cell by 0.087 Hz. It is suspected that this will result in the third WTI harmonic being located 0.087 Hz into the adjacent Doppler spectrum. Counting down 0.087 Hz from 1.923 Hz you arrive at 1.84 Hz, exactly what is observed in the initial inspec- tion of the averaged Doppler spectra in figure 9.6. This evidence suggests that any

9. Wind farm clutter mitigation

wind turbine frequency modulation outside the Doppler spectral frequency range −1.923 Hz to 1.923 Hz would appear in the adjacent Doppler spectra. This idea is consistent with the sampling of the continuous data collection by the processing software.

Considering then unambiguous wind turbine frequency modulations in the stan- dard operating mode of 0.26 s, two wind turbine modulations are expected and ob- served in the positive and negative Doppler spectral regions. If the sweep frequency time were then decreased the Doppler bandwidth at each range would increase. The HF radar allows for up to six 10 minute measurement periods every hour, incor- porating data collection and processing into this time. As discussed in chapter 4, NOCL makes three measurements per hour utilising four of the possible six mea- surement periods (a longer averaging time for one of the three measurements takes up two of the possible six measurement periods). During the month of April 2010 both the Llanddulas and Formby radar were configured to make additional measure- ments at the 10 minute and 50 minute measurement slots. This allowed for the use of two shorter sweep frequency times, 0.195 s and 0.21667 s, extending the Doppler spectral frequency range to ± 2.564 Hz and ± 2.308 Hz, respectively. Full oper- ating parameters for the month of April 2010 are given in table 9.4 for Llanddulas and table 9.5 for Formby. With the chirp frequency large enough it is expected that a peak would be observable at ± 2.01 Hz and that there would no longer be a peak at ± 1.84 Hz. Time past the hour Frequency (MHz) Chirp Length (s) Sample Frequency (Hz) Doppler spectra range (Hz) Doppler spectra resolution (Hz) 00 13.430 0.26 3.836 ± 1.923 0.0075 10 13.430 0.195 5.128 ± 2.564 0.0100 20 13.465 0.26 3.846 ± 1.923 0.0075 40 13.395 0.26 3.846 ± 1.923 0.0075 50 13.430 0.21667 4.615 ± 2.308 0.0090

Table 9.4: The Llanddulas radar operating parameters for April 2010.

Figure 9.8 shows the average Doppler spectra for the extended Doppler fre- quency range at ten minutes past the hour in April 2010. As was the case for the twenty past the hour average shown in figure 9.6, both Llanddulas and Formby Doppler spectra are seen in figure 9.8 to consistently have high energy spikes at identical frequencies. With the extension of the frequency range the spike, which did occur at ± 1.84 Hz for the standard frequency range shown in figure 9.6, is now

9. Wind farm clutter mitigation

absent. Based on the first two spikes being exact multiples of each other it was expected that the ± 1.84 Hz peak would be identifiable in the extended frequency Doppler spectra at ± 2.01 Hz. This is what is seen in the Doppler spectra average for the extended frequency range at ten past the hour (figure 9.8). The spikes in the Doppler spectra have behaved exactly as predicted confirming the existence of subsequent harmonics in adjacent measurement cells to those directly observing the RFWF.

Figure 9.8: Identification of the wind turbine spectral energy from the direct HF radar cell in the Doppler spectra average. Ten minutes past the hour measurement, April 2010 from cells co-located with the Rhyl Flats wind farm.

Time past the hour Frequency (MHz) Chirp Length (s) Sample Frequency (Hz) Doppler spectra range (Hz) Doppler spectra resolution (Hz) 00 12.450 0.26 3.836 ± 1.923 0.0075 10 12.450 0.21667 4.615 ± 2.308 0.0090 20 12.465 0.26 3.846 ± 1.923 0.0075 40 12.435 0.26 3.846 ± 1.923 0.0075 50 12.450 0.195 5.128 ± 2.564 0.0100

9. Wind farm clutter mitigation

Figure 9.9: Identification of the wind turbine spectral energy from an adjacent HF radar cell in the Doppler spectra average. Ten minutes past the hour measurement, April 2010 from Llanddulas HF radar cells co-located with the Rhyl Flats wind farm. Multiples of the wind turbine modulation frequency are indicated by a number followed by fwt.

With the first and second harmonics of the suspected wind turbine modulation identified, as well as the predicted shift of the third harmonic confirmed, the ex- pected frequencies of the fourth, fifth, sixth and seventh harmonics can more confi- dently be calculated.

Figure 9.9 shows the Llanddulas averaged Doppler spectra for ten past the hour, as in figure 9.8. The expected locations of the forth, fifth and sixth harmonics of the wind turbine modulation can be seen to have located additional high energy peaks in the averaged Doppler spectra. For the standard chirp length used by NOCL, also the chirp length for figure 9.6, fourth and fifth harmonics would be expected at ± 1.17 Hz and ± 0.5 Hz. Figure 9.6 shows this to be the case but the energy at these frequencies, represented by their spectral amplitudes, are starting to reduce. Any further harmonics would then be expected to appear in a third cell with continued diminishing amplitude.

9. Wind farm clutter mitigation

dard chirp length have fallen outside the frequency range where wave measurements would be expected. Based on their predicted, and confirmed, location a seventh (shorter chirp length) and fifth (standard chirp length) harmonic can however be expected to fall within or near to the frequencies utilised for ocean parameter mea- surement.

The amplitude of the wind turbine harmonics does appear to gradually reduce. For the Formby radar measuring currents with large SNR, as reported in chapter 8, the HF radar still has very good agreement with ADCP data. For the Llanddu- las radar measuring currents with much small SNR, it is possible that the seventh (shorter chirp length) and fifth (standard chirp length) wind turbine harmonic are being picked out as the Doppler shifted Bragg peak. This could be the cause of the very large current measurements that are occasionally being made by the radar when the ADCP is reporting currents close to zero, as shown in figure 8.4 d).

The additional energy from the wind turbine modulation at the frequencies used for wave measurement could be responsible for large Hs measurements by the radar. This is because the WTI will only act to add energy to the measured sea state there- fore only increasing the Hs measurement. Removing this energy would then reduce the Hs measurement.

With the nature of the WTI confirmed for the nominal operating frequency it is important to remove energy added to the Doppler spectra in these circumstances. As suggested by Robinson et al. [2013], this will hopefully improve HF radar per- formance when measuring waves but may also have some benefit to current mea- surement as well.