A Ground-Based Cloud Motion Prediction Method

Cloud conditions play a significant role in affecting the output power of solar photovoltaic (PV) panels, and ground-based cloud images can provide valuable data for cloud monitoring and ultra-short-term PV power forecasting. From the perspective of image sequence prediction, a cloud motion prediction method based on Cascade Causal LSTM is proposed by using the ground-based cloud image data taken by Total Sky Imager, realizing the prediction of cloud contour change. On this basis, digital image processing technology is used for features extraction from the ground-based cloud images, including clear sky irradiance, clusters, color statistics, and texture that affect PV power generation. Finally, the ultra-short-term PV power forecasting model with multi-information fusion is established. Experimental results demonstrate that the proposed PV power prediction method can effectively improve the accuracy of ultra-short-term PV power forecasting. It is further shown that the effectiveness of the proposed cloud motion prediction method.