Conventional PV power stations can only rely on local surface meteorological observation information for irradiance forecasting,and it is difficult to tap the spatio-temporal correlation characteristics of wide area photovoltaic resources around the power station for these kinds of stations,which limits the forecasting accuracy of irradiance and PV power.To solve the above problems,this paper proposes a mapping method for the spatial distribution of wide area irradiance around PV power station based on satellite remote sensing,and establishes an ultra-short-term spatio-temporal correlation forecasting model for surface irradiance based on graph convolutional network(GCN).The method makes full use of multi-channel satellite data and considers the spatio-temporal correlation characteristics to improve the ultra-short-term prediction accuracy of surface irradiance.The feasibility of the inversion model of surface irradiance is verified through the simulation analysis of a photovoltaic station,and the progressiveness of the corresponding spatial-temporal correlation prediction model is also proved.
关键词
卫星/特征选择/辐照度/反演/图卷积神经网络/地表辐照度超短期预测
Key words
satellite/feature selection/solar irradiance/inversion/GCN/ultra-short-term forecasting of surface irradiance
维普
万方数据