Paper presented to the 3rd Southern African Solar Energy Conference, South Africa, 11-13 May, 2015.
The characterization of the solar radiation variability is a fundamental step before prediction and is crucial to transform an intermittent source of energy into a stable one. As solar irradiance result from a nonlinear and non-stationary process, we use a multifractal approach based on the Hilbert-Huang Transform (HHT) consisting of an empirical mode decomposition (EMD) followed by a spectral analysis. In this paper, we will briefly introduce the HHT data analysis method. Such a recent adaptive data analysis method has been applied to Reunion Island global solar radiation time series of measurements with a sampling rate of 1/60 Hz over six years. Through the EMD, daily global solar radiation data were decomposed into several intrinsic mode functions (IMF). For each IMF, we estimate the amplitude, instantaneous frequency and Hilbert spectrum for the original data. From the comparison of Hilbert Spectrum and Fourier Spectrum, we find the calculated solar radiation power spectrum follows a power law behavior close to the Kolmogorov law. The method described in this paper provides an amplitude frequency representation of the global solar radiation sequences resulting in a probability density function and a scaling coefficient. The multifractal approach allows to extract parameters connected to the multifractal properties of the global solar radiation.