CLOUD TYPE CLASSIFICATION IN GROUND-BASED SKY IMAGES WITH DEEP LEARNING

Abstract

Clouds cover more than half of the Earth's surface and are the subject of intense research in climate modeling, weather forecasting, meteorology, solar energy research, and satellite communications. Determination of cloud types and characteristics is of great importance in developing and applying solar radiation forecasting models. Therefore, classifying clouds into different categories according to their optical properties is essential for developing solar radiation forecasting algorithms. In this study, we have tried to develop a more efficient, reliable, and cost-effective solution for cloud classification. In this context, a deep-learning CNN model that can classify six different cloud types is developed, and its performance and applicability are examined. The SWIMCAT-EXT dataset, available for research activities, is used for training and testing the model. The experimental results show that the proposed CNN model can successfully classify cloud types and can be integrated into the solar radiation forecasting process.