Ultra-Short-Term Prediction Of Photovoltaic Power Generation Considering Cloud Shielding

Under the strategic background of "carbon peak" and "carbon neutral",solar power generation technology has been fully developed.Affected by meteorological factors,photovoltaic power generation has strong volatility and randomness.When photovoltaic is connected to the grid in a large scale,the fluctuation and flicker of system voltage will be caused,and even the power balance of the system will be broken,threatening the safety of the power grid.The ultra-short-term power prediction can effectively reduce the influence of photovoltaic output fluctuation on the power grid system.The fast and irregular movement of clouds is the main factor causing fluctuations of surface irradiance and pv output.In order to further improve the prediction accuracy,a super-short-term pv power prediction method considering cloud shielding is proposed in this paper.The method is studied from three aspects:cloud trajectory,surface irradiation intensity and power prediction model.In order to achieve high precision prediction of surface irradiance and power,the cloud movement trajectory should be studied first.In this paper,the study area is firstly determined according to the location information of photovoltaic stations,and the target clouds with potential influence on photovoltaic output are identified in the area.Then,the proposed 2d convolution algorithm is used to determine the cloud displacement vector,and the proposed wind speed prediction method is used to determine the cloud movement vector in the future period.Finally,cloud trajectory tracking prediction is realized.Surface irradiance can determine the output of photovoltaic power station to a certain extent.The high-precision prediction value of surface irradiance is beneficial to improve the prediction accuracy of photovoltaic power generation.In this paper,an irradiance prediction model considering cloud shielding is proposed based on the moving trajectories of future cloud clusters.Firstly,the clear sky model is established by using astronomical formula to calculate the surface irradiance under ideal conditions.Then,a cloud shielding model was established based on the location relationship between solar,cloud and photovoltaic power station,and the shielding situation of photovoltaic power station by cloud cluster was quantitatively analyzed,and the attenuation coefficient of surface irradiance was obtained.Finally,the ideal irradiance is combined with the corresponding attenuation coefficient to predict the surface irradiation intensity.The result of numerical example shows that the prediction effect of this method is better than that of traditional numerical weather forecast,especially in the weather condition with big fluctuation,it still has high prediction accuracy.Based on the predicted value of surface radiation intensity and the improved LSTM neural network,the ultra-short-term prediction of photovoltaic power generation is realized.First of all,the raw data is preprocessed to ensure the data quality of pv power generation prediction samples.Secondly,the maximum mutual information coefficient was used to screen the characteristic variables with strong correlation.Then the nearest neighbor clustering propagation algorithm is used to divide the data sample set to improve the similarity between the data samples.Finally,the power prediction model is constructed based on the improved LSTM neural network.The results of a practical example show that the proposed model can effectively improve the accuracy of ultra-short-term prediction of pv power under cloudy weather,and has a certain promotion value.