Identification and analysis of wind speed patterns extracted from multi-sensors measurements

The understanding of the vertical as well as the horizontal behaviours of wind speed is of great importance in many applications such as aviation, meteorology and wind energy conversion. In this work, we propose to apply the principal component analysis (PCA) in order to extract probable components of wind speed. The idea behind the use of PCA is to introduce mixed sources signals to PCA algorithm as input in order to obtain a separated patterns as output. Hence, values of wind speed measured at three levels above the ground will be used as three separate sensors in order to extract the horizontal and the vertical components of wind speed. Once the principal components of wind speed separated, a process of recognition and identification is undertaken via the inspection of the statistical as well as the cyclical behaviors of the obtained components. For the examination of the statistical properties of wind speed, we propose to carry a comparison of the probability density of the extracted components with the Weibull distribution (commonly used to fit wind speed distributions). However, the spectral behavior of the obtained patterns is examined using time–frequency analysis rather than the traditional Fourier analysis. The time–frequency analysis has been chosen as it serves the purpose of following the diurnal and seasonal time variation of the wind speed spectrum. As a result, it has been found that the horizontal wind speed component fits the Weibull distribution and it is characterized by synoptic and intra-seasonal oscillations. On the other hand, the wind speed vertical component is better fitted by the extreme value distribution. It has been also found that the diurnal oscillations are the most significant oscillations in the vertical components especially in the summertime period.