Ultra-short-term Probability Prediction Method of Photovoltaic Power Considering Satellite Image

As cloud shielding is one of the main factors leading to the fluctuation of photovoltaic power, an ultra-short-term probability prediction method of photovoltaic power considering satellite cloud images is proposed. Firstly, a cloud displacement vector algorithm for the outer edge of the cloud cluster is designed, based on the block-area-center-positioned strategy, which improves the calculation efficiency and precision; Then, based on the Gaussian mixture model of multi-station output, a conditional probability model considering cloud occlusion is constructed. The model can give the probability prediction of power for target photovoltaic station, using the historical power of reference stations and predicted power of target station. The model corrects the correlation of adjacent power stations by introducing time delay, which is calculated using the result of cloud displacement vector and Pearson correlation coefficient of power. The cases show that the proposed method is superior to the existing probability prediction methods both in reliability and sharpness.