SolarNet: A Deep Convolutional Neural Network for Solar Forecasting via Sky Images

Abstract

The exponential growth of solar energy poses challenges to power systems, mostly due to its uncertain and variable characteristics. Hence, solar forecasting, especially short-term solar forecasting (STSF), has been adopted to assist power system operations. The STSF takes inputs from various sources, among which sky image-based STSF is not yet well-studied. In this paper, a deep convolutional neural network (CNN) model, called the SolarNet, is developed to forecast the 10-min ahead GHI by only using sky images without numerical measurements and extra feature engineering. The SolarNet, composed of 21 convolutional, max-pooling, and fully-connected layers, learns latent patterns between sky images and global horizontal irradiance (GHI) in an end-to-end manner. Numerical results based on ten years data show that the developed SolarNet outperforms the benchmarking persistence of cloudiness model and machine learning models by up to 11.88% under various weather conditions.