Abstract
In 2003, the International Commission on Illumination (CIE) adopted 15 standard skies that cover a broad spectrum of the usual skies found in the world. Each sky represents a unique sky luminance distribution, which is the most effective way to classify the 15 CIE Standard Skies. However, luminance distributions for the whole sky vault are far from being widely available. Alternatively, the standard skies can be categorized by various climatic parameters but the criteria to distinguish individual skies are not always clearβcut and may lead to ambiguous results. The artificial neural networks (ANNs) represent a powerful tool for pattern recognition. This paper presents the work on the classification of the standard skies using a new form of neural network architecture, namely the probabilistic neural network (PNN), which is particularly apposite in classification problems. Five meteorological variables, viz. zenith luminance, global, directβbeam and skyβdiffuse illuminance, and solar altitude are employed as input data. Totally, 9000 samples covering the time span between 1999 and 2005 are shuffled. The findings suggest that the PNN is an appropriate tool for sky classification. Copyright Β© 2009 Royal Meteorological Society