Combined ultra-short-term prediction method of PV power considering ground-based cloud images and chaotic characteristics

Abstract

To further improve the ultra-short-term prediction accuracy of photovoltaic (PV) power in mutant irradiation scenarios caused by moving clouds, a combined ultra-short-term prediction method of PV power considering ground-based cloud images and chaotic characteristics is proposed. Firstly, to further obtain more accurate prediction results of cloud distribution, a two-stage cloud speed calculation model is constructed by synthesizing different cloud image feature extraction and matching algorithms. Secondly, to quantify the effects of cloud distribution on PV power accurately, a dynamic modeling method of the “cloud images − PV power” mapping relationship is proposed using chaos theory. Furthermore, due to the strong dependence of irradiance on weather conditions, a combined prediction method of PV power is constructed based on the cloud classification model. Finally, the simulation analysis is carried out using the dataset containing mutant irradiation scenarios. The results show that the proposed cloud speed calculation model has higher calculation accuracy than the conventional method, while the proposed power prediction method maintains excellent prediction accuracy and applicability in mutant irradiation scenarios.