Análise da dinâmica de longo prazo do recurso solar no território brasileiro utilizando decomposição de séries temporais e técnicas de estatística multivariada

Abstract

The use of electrical energy generation systems based on the direct use of solar radiation is showing a rapid growth both on Brazil and worldwide, mainly trough photovoltaic systems, due to the continuous reduction of the prices associated with the components fabrication. As this technology does not present storage systems intrinsically associated, and external storage technologies still present high costs, it is important to understand the behaviour of the variability of the source. This dissertation explores spatiotemporal aspects of solar resource behaviour throughout the Brazilian territory, using satellite based data, from the NASA's project POWER (Prediction Of Worldwide Energy Resource), and meteorological stations data, from the international project SWERA (Solar and Wind Energy Resource Assessment), both freely obtained from their's respective websites. The analyses, focused on long-term variability, is performed using a methodology proposed by the author, based on the decomposition of historical irradiance time series into trend, interannual, seasonal, and monthly remainder components. Each component is separately evaluated, using multivariate statistic techniques, such as principal component analysis and cluster analysis. Moreover, a new model of index for pairwise evaluation of the solar resource dynamics is proposed, which presents greater practical meaning than the correlation index, commonly used. It is shown that the proposed methodology is useful for the comprehension of the solar resource dynamics, specially considering its use regarding energy planning. However, the analysis of some components, mainly the interannual one, shows that a larger contingency of data might be needed for obtaining more consistent results. Among the results found, there is the suggestion that the development of the solar energy on the state of Mato Grosso do Sul can be interesting, aiming at lessening global solar energy monthly variability, specially when this source of energy presents a larger participation on the national energy mix. Furthermore, the two datasets used on the study are compared, and considerable dissimilarities are verified between them. This result bringsome questions regarding the quality of the satellite based data and its use for understanding of local variability.