Real Time Energy Forecasting Scheme for Large Scale PV Solar Power Plants - Cloud Image Segmentation and Contour Detection

Abstract

Solar power is primarily generated through large scale solar power plants. However, several factors can disrupt the output of these plants, with cloud cover being one of the most significant. To predict the effect caused by cloud cover, initially the cloud images should be segmented, and the contour lines of those images should be identified. This part of the study examines multiple algorithms including machine learning and deep learning models to segment the cloud images and detect the contour lines of them. The aim of this part of the study is to test, evaluate, and validate the segmentation and contour identification models to determine the velocity vectors of the clouds. The accuracy of velocity vector patterns is highly relying on the contour lines identified from this approach, making this a crucial step in forecasting cloud movement and its prediction of the outcome of large-scale PV solar power plants under grid 4.0 within the smart grid context.