The large-scale integration of intermittent renewable energy sources into electricity distribution is one of the significant challenges in meeting energy demand in the near future. The solar energy resource has a fluctuating spatial and temporal character intrinsic to the radiative processes with the atmosphere and the Earth’s surface. Efficient use of solar energy requires technologies that consider such variability and its impacts on the electrical system. In this context, a reliable computational model to estimate the surface solar irradiance is beneficial for operational issues. This work presents a methodology based on Markov chains for the generation of a global irradiance time series with one-minute temporal resolution using observational database acquired by a basic weather station. The statistical model was developed using irradiance data acquired in Sao Martinho da Serra/RS and Petrolina/PE to verify their performance in two different climates. The transition probability matrices were determined for the following cloud conditions: cloudy, partly cloudy or clear sky using ten years of observational data. Next, Markov’s walking technique was used to generate a data series with one-minute temporal resolution taking into account the cloudiness classification. Model validation was performed using solar irradiance data not considered to calculate the transition probability matrix (MPT). The mean squared deviation of the relative distribution frequencies indicated a deviation of only 1.5%, and the Kolmogorov-Smirnov test indicated that the synthetic and observational series presented similar cumulative frequency distributions. Thus, the proposed methodology showed high reliability in reproducing the temporal variability due to the stochastic nature of the incoming solar energy in both sites.
Read full abstract