G. Koudouris, Stochastic investigation of temporal variability of solar radiation, Diploma thesis, 104 pages, Department of Water Resources and Environmental Engineering – National Technical University of Athens, Athens, November 2017.
With the progress of technology and the increasingly exploitation of the mineral wealth of the earth, there is an urgent need to turn humanity into renewable energy sources. Since the beginning of the 21st century, the scientific community has made huge leaps to exploit all renewable energy sources. One of the most widely known is solar radiation, which is one of the most important forms in which the human factor is going to rely in the future. However, the variability of solar radiation has a significant impact on solar energy conversion systems, mainly in photovoltaic systems, characterized by a fast and non-linear response to incident solar radiation. The performance prediction of these systems is based on hourly or daily data because these are usually the only ones available. However, given that the presence of clouds modifies global radiation and knowing that the processes concerning the clouds tend to be very dynamic, great variability of the daily and hence hourly values is expected. The aim of this work is to investigate the stochastic nature and time evolution of solar radiation process in a daily and hourly step on a monthly basis scale, with the ultimate goal of creating a stochastic model capable of generating hourly solar radiation. For this purpose, an analysis was initially made at stations in the Greek area and then the research was expanded on a global scale. We propose a distribution that can adequately describe daily radiation and a new distribution consist of the sum of two distributions that is capable of fitting hourly solar radiation. Also, we use the clear sky index coefficient (ΚT) due to the double periodicity of the process, so a complete description of the solar radiation in all scales was achieved. As an extension we use statistical case tests and selection criteria, in order to verify the good fitting of the distribution. Finally, we propose the first pseudo-cyclostationary with long term persistence behaviour model capable of reproducing the clear sky index coefficient (KT) and so the hourly solar radiation. The model can maintain and preserve the probability density function and so the first four central moments i.e. (mean, standard deviation, asymmetry and kurtosis) and also the parameter Hurst which indicates correlation and a long-term persistence behaviour.