Ultra-short-term photovoltaic power prediction based on ground-based cloud images: A review

Solar power is a key option for sustainable development and clean energy transition due to its advantages of renewability, almost zero carbon emissions, and flexibility in scale. However, the intermittent variation of solar radiation makes it a huge impact on power quality and power dispatch after large-scale access to the power grid. Photovoltaic (PV) power Prediction is the best way to solve this problem. In practical application, the ultra-short-term PV power prediction can not only improve the stability of PV power generation system, but also has a great reference value for the spot trading in the power market. Since cloud cover significantly and intermittently affects the amount of incident solar radiation, the high-resolution cloud images are helpful for detecting fine-grained cloud variations. Therefore, studying ultra-short-term PV power prediction based on ground-based cloud images is significant. This paper firstly introduces the ultra-short-term PV power prediction systems based on ground-based cloud images. Subsequently, it reviews four parts: ground-based cloud images acquisition, preprocessing, cloud identification and trajectory prediction, and PV power prediction. It also summarizes the ground-based cloud images acquisition device and existing available datasets, and conducts a comprehensive comparison and analysis of the methods involved in ground-based cloud image preprocessing, cloud cluster identification and trajectory prediction, and photovoltaic power prediction. Finally, it puts forward the outlook of the future technological research with respect to the current research status.