Application of Machine Learning for CIE Standard Sky Classification

The spectrum of skies in the world was classified into a range of 15 standard skies. These standard skies are crucial in estimating solar irradiance and daylight illuminance needed for energy-efficient building designs. Generally, using the sky luminance distributions to identify the standard skies is the most effective method, but these are sparingly measured. Alternatively, climatic variables can identify the standard skies. Nevertheless, it is necessary to determine if the available climatic variables could correctly identify these skies. Also, there are several climatic variables, but there is no criterion for selecting a climatic variable over the other. This study addresses the lack of luminance distributions measurement by classifying the standard skies using measured climatic data. Furthermore, sensitivity analysis was used to determine the relative importance of one variable over the other. Importantly, the proposed approach for classifying the standard skies was implemented using support vector machines (SVM). Findings showed that the SVM could classify the skies with an accuracy of 72.4% on the training data and 71.4% on the test data.