Photovoltaic Power Ramp Prediction Based on Sequential Difference

The prediction of photovoltaic power ramp events is one of the important works for the grid to realize intelligent scheduling. A method was proposed to automatically predict the ramp events based on a set of steps including the sequence difference method for predicting cloud motion vectors. First, image preprocessing including mask and image inpainting was performed on the ground-based image to extract the sky area. Second, a cloud detection algorithm that reduced sunlight interference was used to identify clouds in images. Third, a geometric transformation that took into account the relative position of the observation point and the clouds was derived to adjust the position of the image pixels. Radial interpolation consistent with the transformation principle was proposed to repair the image and complete the distortion correction. Fourth, the sequence difference method is proposed to predict cloud motions, and the individual cloud motion vectors were provided to adapt to the characteristics of cloud motion anisotropy. Finally, photovoltaic power ramp events are predicted based on individual cloud motion vectors. The experimental results showed that the distortion correction with precise geometric transformation derivation and radial interpolation improved the accuracy and quality of the results. The sequence difference method improved the accuracy of cloud matching and realized anisotropic cloud motion prediction. Further, the photovoltaic power ramp event was accurately predicted.