An Effective Model Presentation for Solar Irradiance Prediction using Deep Learning Neural Network

World energy demand is increasing and among renewable energy sources, solar as a clean energy source plays an important role in energy supply. The variability of the solar energy source makes it difficult to operate and manage the power system. Therefore, a very short-term forecasting of solar irradiance is required to operate the power grid efficiency and reliably against these fluctuations. In this study, a model is proposed for a very short-term solar irradiance prediction based on sky images and deep learning. A series of whole sky images are performed to detect and track the movement of clouds for ۱۰ minutes ahead using dense optical flow. Then, the solar irradiance is forecasted using predicted images of cloud motion via ResNet۵۰ deep learning algorithm. The proposed model is evaluated by the Root Mean Square Error (RMSE), R-Squared Correlation ( ) between the actual and forecast values of solar irradiance.

Keywords:

Deep learning , Sky image , Convolutional Neural Networks , End-to-End mapping model , Irradiance