Comparing various solar irradiance categorization methods - A critique on robustness

Traditional ways of planning and operation of electricity networks have been challenged lately by the spread of variable renewable energy sources, especially solar photovoltaics, and the need for better forecasting has increased interest in various solutions. Categorization of solar irradiance data, as one of the earliest applied techniques, is a frequently discussed topic in the literature, but the efficiency of different methods may be significantly variable. The aim of this paper is to compare various categorization methods using a one-year-long solar irradiance dataset and reflect on their inefficiencies and the need for more timely solutions. Six methods have been selected and implemented, including deterministic and non-deterministic ones. The number of groups created by the methods varies between three and five, and they also use data with different temporal resolution. The aim of the comparison was to reveal the strengths and weaknesses of the implemented methods and to highlight possible contradictions among them, based on two tests: uniformity in identifying days with clear or cloudy conditions and contradictory identifications. The results have shown that the usability of certain methods is limited as they are very sensitive to input data, and categorization is often inconsistent, which limits the usability and dissuades users of such methods. (C) 2020 The Author. Published by Elsevier Ltd.