Using LSTM Network Based on Logistic Regression Model for Classifying Solar Radiation Time Series

An accurate knowledge of the types of solar radiation is important. Some machine learning and deep learning models have been used to classify solar radiation time series. However, applying some methods on a large spatial and temporal scale has only been hardly investigated. Therefore, in this study, two algorithms were used: long-term memory (LSTM) networks and the traditional statistical method represented by the logistic regression (LR) model to classify solar radiation patterns based on some meteorological time series data over a large time range. The results show that LSTM achieves acceptable results with clear superiority of the LSTM method for classification tasks. As a result, it is feasible to deduce that the capability of LSTM networks to learn long-term and determine the relation between time series of solar radiation and meteorological data has been more significant for more sophisticated applications, and hence the study underscores the importance of deep learning models, including LSTM networks, for large-scale applications comparing to LR and other traditional models.