Clustering and Prediction of Solar Radiation Daily Patterns

This paper addresses the problem of clustering daily patterns of global horizontal solar radiation by using a feature-based approach. A pair of features, referred to as Sr and Hr, representing a measure of the normalized daily solar energy and of the energy fluctuations, respectively, is introduced. Clustering allows to perform some useful statistics at daily scale such as estimating the class weight and persistence. Furthermore, the problem of one-day ahead prediction of the class is addressed by using both hid- den Markov models (HMM) and Non-linear Autoregressive (NAR) models. Performances are then assessed in terms of True Positive Rate (TPR) and True Negative Rate (TNR).