Outliers Detection in a Quality Control Procedure for Measurements of Solar Radiation

Abstract

The correct design of Photovoltaic Systems and Solar Heating Systems should be based on solar irradiance measurements (global, sky-diffuse and direct-beam) acquired on land surface. This paper presents a quality control procedure for global (IG), sky-diffuse (ID) and direct-beam (IB) solar irradiance measurements on horizontal surface, considering the application of a impossible values identification method with predefined thresholds and a statistical techniques for outliers detection (identifying measures suspected to be correct). To evaluate and quantify the occurrence of physically impossible and suspected correct measures, it was also proposed to apply this procedure in six years of measurements (2005 to 2010) of the Solar Radiometry Laboratory of the School of Agricultural Sciences of the Sao Paulo State University (UNESP) of Botucatu (SP) - Brazil. The procedure was based on the International Commission on Illumination (CIE) fundamental quality control of solar measurements and statistical ranges according to the different sky cover classification (KT). The results indicated that the dataset used for the demonstration contained approximately 1.00%, 1.49% and 1.88% of global, sky-diffuse and direct-beam irradiance values on horizontal surfaces, respectively, flagged as invalid by the procedure. This amount represents measures physically impossible or suspected to be correct (outlier). The techniques for outliers detection, proposed in this paper, in combination with the CIE quality control criteria, should be applied in solar measurements to provide greater security for future investigations pointing the valid and invalid measures.