Photovoltaic energy conversion models with clustering-based classification of days from irradiance data

In the literature, many contributions compare the effectiveness of different energy models for photovoltaic (PV) generators in specific installation sites. Most techniques are analytical or based on electrical equivalent circuits. However, only a few works investigate the performance of the models in different weather classes. The parameter mostly used to classify the days into weather categories is the clearness index. An alternative parameter is the clear sky index. However, information about their values and ranges in different weather conditions is missing in the literature. This paper is based on data gathered in Italy, Brazil and Portugal, at different latitudes, with fixed and sun-tracking configurations of the PV generators, and for different weather conditions, and has a twofold objective. First, the clearness index is shown to be inappropriate to represent general situations, while the clear sky index is suitable to identify consistent ranges that represent different sunny, partly cloudy and cloudy days, at different latitudes and PV configurations, by using a clustering procedure. Next, the effectiveness of the PV energy conversion models for different installation sites is evaluated, showing which model is more suitable for each site and type of day. In this context, a model based on single diode model is proposed, formulated after numerically determining the parameters of the equivalent circuit. The parameters under reference conditions and the coefficients of the equations that quantify the dependency on weather conditions have been numerically optimized starting from experimental datasets of PV modules.