Complexity Analysis of the Multivariate Wind Measurements: Renewable Energy Applications

Abstract

Complexity analysis of real world multivariate wind data is addressed using the recently proposed multivariate multiscale entropy (MMSE) analysis. Both the original (univariate) MSE and the multivariate MSE methods are shown to perform better than traditional complexity analysis techniques, since they operate on multiple temporal scales of the signals and are, thus, able to extract information regarding inherent long range correlations in the data, signatures of structural complexity. The MMSE method, in addition, can also quantify inter-channel correlations in multivariate data and is perfectly suited for the analysis of multichannel data, where the channels exhibit different dynamical properties, such as three-dimensional wind speed. To cater for the non-stationarity of wind recordings, a novel scheme is presented for obtaining data-driven scales from input data using multivariate extension of empirical mode decomposition (MEMD), in order to obtain robust estimates. Our method can thus characterise different wind dynamics regimes and cloud-cover conditions in complexity domain. Finally, we illuminate how the different dynamic complexities associated with different wind regimes, and their connection with atmospheric parameters, such as temperature, or cloud cover, can be used as baseline knowledge in several important settings in renewable energy.