Classification of Types of Daily Solar Radiation Patterns Using Machine Learning Techniques

In this work, a new model is used for classifying solar radiation patterns, with the aim of studying the production and enhancement of solar energy efficiency. The model incorporates various clustering and pattern recognition methodologies, considering different criteria. To achieve a comprehensive and generalized recognition of these patterns, a methodology previously applied in similar approaches, which focuses on the analysis of time series data, is employed. Specifically, an exploratory analysis is initially conducted, followed by the conversion of the data into a daily polar representation. Subsequently, the process involves extracting relevant features and performing classification using solar irradiation data collected in the city of Cuenca, Ecuador, between 2014 and 2017. The analysis yielded four distinct clusters, accompanied by supplementary information and the corresponding average frequency of occurrence. The use of neural networks demonstrates satisfactory results when classifying solar irradiation patterns by not requiring prior knowledge of climatic and geographic parameters.