Series temporais de irradiação solar: recuperação de dados mediante componentes principais.

Abstract

Time series o f global irradiation of two stations (Barra de Santa Rosa and Pombal) of the solanmetnc network of Paraiba were simulated using the technique of multivariate analysis. The data from 16 stations of solanmetnc network of Paraíba and 3 stations of the National Institute of Meteorology (INEMET- MA) located in the neighbourhood of Paraiba in the states of Pernambuco. Ceara and Rio Grande do Norte were used in this study The data collection at the above stations during the period of six years (1975 to 1980) for the months of March (in the rainy season) and September (in the dry season) were utilized in this study. Groups of stations with homogeneous characteristics were previously determined. The method of grouping adapted, among the methods tested, was that of Ward winch presented satisfactory characteristics of separation into groups and it looks to be the recommendable method for this work. The homogeneous groups of stations were obtained by grouping various types of variables, for example: crude data and standardized data of irradiation, 'factor loading" obtained from the eigenvalues which explained 90% of variance and the first the.three, eigenvalues. The number of appropnate groups for the 19 stations was five, but due to the limitations of data dming the period of study, less number of groups were used in some cases. A simulation of the time series was applied using the data of the group to which Barra de Santa Rosa f.BSR) and Pombal (POM) peitam. Applying the method developed by CEBALLOS and BRAGA (1991) which is based on the matrix of the aurovectors associated to the correlation between the stations of a homogeneous group. For the pertment calculations, only the principal auto vectors o f the matrix (which explamed 90% of the variance and the first two autovectors) were utilized. The simulations present good results of deviation of the order of 0,45 of the shnulated series with relation to the original series. The magnitude of the derivation is of the order of the error encountered ui the irradiation data of the stations studied.