Identifying power profiles in the photovoltaic power station data by self-organizing maps and dimension reduction by Sammon's projection

This paper presents results of the identification of clusters in the hourly recorded data of power from a small photovoltaic power station. Our main aim was to find a method of how to identify typical patterns of generated power. Although one can think that sunny days are the same, the power of the sun light is very volatile during a day. We were not interested in finding the absolute values of this power but just its patterns according to the day's maximal power. Our proposed method is based on several techniques. We used network algorithm as a method for removing noise from the data, Sammon's projection for visualization and dimensionality reduction and final clustering by the self-organizing maps.