Advanced digital signal processing methods applied to acoustic wind speed profiling for use in wind farm assessment
This project will develop and implement state-of-the-art signal processing methods on a newly designed SODAR acoustic radar to achieve significant improvements in wind profile data quality and height resolution - two major limitations for the use in wind farm site evaluation.
Most commercial SODARs measure wind speed by propagating a single-frequency acoustic pulse in each of three or five directions. Temperature and velocity fluctuations in the atmosphere scatter these pulses, some of the sound is backscattered towards the SODAR and is recorded. Due to the wind, the fluctuations are moving relative to the SODAR device, therefore the received, backscattered sound demonstrates a Doppler-shift. The amount of Doppler-shift in each beam is proportional to the radial wind velocity. Wind speed and direction, is then estimated using the radial velocity estimates from at least three beams.
One of the limitations of conventional SODARs is the loss of data due to the low level of the returned, backscattered signal. This means that conventional SODAR systems must transmit at high SPL levels in order to maximise signal-to-noise-ratio. By using more sophisticated coding strategies, the SNR ratio may be increased without the need to increase the transmission level.
The Acoustic Research centre has recently taken delivery of a special-purpose acoustic wind profiler developed by the University of Auckland. The new SODAR will be used to facilitate improved measurement performance with the help of advanced signal processing. This SODAR comprises a phased array of loud speakers and a separate phased array of microphones.
All channels can be individually controlled to achieve highly directional beam-forming. The intention is to operate the transmitting array and receiving array in close proximity so as to emulate a typical commercial system but still allow separate access to transmitter and receiver. The system is controlled using matlab and the playrec, Portaudio implementation which is an open-source audio I/O library.
SODAR wind speed profile
This is a dynamic wind speed profile, recorded yesterday at the UoS external site, and updated here daily. The colours depict the wind speed, while the x axis shows the time and the y axis the height. The data has been interpolated to make viewing easier. The blank regions represent regions where error codes were produced, this is often due to periods of rain or very stable conditions. Two methodologies are compared, the first uses average spectra to compute the doppler spectra, the second is know as the cluster method and averages the location of the peak in the returned spectrum.
SODAR signal simulation
Prior to the implementation of signal processing algorithms on the experimental test bed, a simulation of the backscattered signal from a simulated moving atmosphere will be carried out. This simulation couples a finite difference time domain (FDTD) acoustic wave simulation of propagation in inhomogeneous media, with a moving, stochastic simulation of turbulent temperature fluctuations. This simulation will be used to investigate the best signal coding strategies and identify any pit falls in the choice of signal coding.
The simulation is demonstrated in the video on the right, where the x-axis represents altitude. The video shows the scattered pressure resulting from a single frequency pulse propagated from the top left of the domain. To reduce the computational load the simulation is carried out in two dimensions rather than three. Perfectly matched layers (PMLs), which absorb sound without producing reflections, surround the main computational region in order to simulate propagation into an infinite domain.
Signal processingThe applicability of the coded signal - matched filter methodology, which is often used in RADAR applications, will be investigated. A matched filter is a filter whose response is designed to produce the maximal signal to noise ratio, in the presence of additive random noise, for a particular transmitted signal. The strength of the methodology comes from the choice of signal coding to apply to the transmitted signal. In radar, for example, the wider the bandwidth used by the coded signal, the higher the resolution of the position estimation for an object. Examples of coding strategies include frequency modulation such as chirps/sweeps or phase modulation such as the spread spectrum technique.
Publications - Conference papers
Paul Kendrick, Sabine von Hünerbein,(2010) 'Atmospheric sound scattering
model to test signal coding methods for acoustic wind profiling' at the
159th meeting of the Acoustical Society of America in Baltimore, Maryland, April
Annotated Powerpoint presentation
Kendrick, P & Hunerbein, S V 2010, Atmospheric sound scattering model to test signal coding methods for SODARs, in: 'ISARS', ISARS, Versailles, France. Conference details: ISARS, Versailles, France, 28-30 June 2010. Extended Abstract
Kendrick, P & Hunerbein, S V & Bradley, S G 2010, A flexible new SODAR design for chirped signals and conformable beam forming, in: 'ISARS', ISARS, Versailles, France. Conference details: ISARS, Versailles, France, 28-30 June 2010. Extended Abstract
Kendrick, P & Hunerbein, S V 2011, Finite Difference Time Domain Simulation
of Acoustic Wind Speed Profiling, in: 'Forum Acousticum', University of Aalborg,
Aalborg, Denmark. Conference details: Forum Acousticum, Alborg, June 2011.
Dr Paul Kendrick
Acoustics Research Centre
Newton Building, University of Salford, Salford M5 4WT, UK.
Tel +44 161 295 4618