Cloud Motion Estimation with ANN for Solar Radiation Forecasting

Abstract

The most critical issue in integrating solar energy into the electricity grid is the variability of solar energy. Cloud cover and motions are the most fundamental factors in the formation of this variability. In the related study, a 167-degree camera is placed in the main campus area of Afyon Kocatepe University, and sky images are recorded at regular intervals. Using the obtained images, cloud motion estimations are made for a 10 second time horizon at a 1 second time scale. Firstly, within the scope of this purpose, the points to be tracked by the Shi-Tomasi algorithm were determined. Then, using the Lucas-Kanade optical flow algorithm, the points found are followed on sequential images. Finally, cloud motion estimations are obtained using the Feed Forward Backpropagation Artificial Neural Network. The results obtained showed that the approach could be used successfully in cloud motion estimation.