Improving Solar PV Nowcasting: Raindrop Detection on All Sky Imagers

Accurate short-term forecasting of Global Horizontal Irradiance (GHI) is essential for optimizing photovoltaic system operations. All sky imagers play a crucial role in providing real-time data for cloud detection and solar forecasting. However, raindrops on the camera dome introduce distortions, leading to inaccuracies in GHI predictions. This study proposes a raindrop detection algorithm in sky images. The image processing methodology involves binary thresholding, morphological operations, and geometric analysis to identify raindrop artifacts. The detection process relies on shape descriptors, such as circularity, to distinguish raindrops from clouds and other features. The analysis was performed on 7,321 high-resolution fisheye images captured at the Cabauw Baseline Surface Radiation Network site, achieving an accuracy of 83% for raindrop detection, validated against 50 manually labeled images. This study highlights the importance of addressing raindrop-induced distortions in sky images and provides a robust, scalable solution for enhancing short-term solar forecasting systems.