考虑沙戈荒地区复杂气象因素的超短期光伏功率预测OA

Photovoltaic(PV)projects in the Gobi Desert and other arid areas serve as important supports for new power systems.However,such project sites are often characterized by complex climatic conditions,with frequent sandstorms,strong winds,extremely high temperatures,and other meteorological phenomena.These meteorological factors exhibit strong coupling and time-varying characteristics,posing significant challenges to PV power prediction.To address this issue,this paper proposes an ultra-short-term PV power prediction method that considers the impact of complex meteorological factors in Gobi Desert and other arid areas.First,a dynamic graph modeling method is introduced to represent the temporal correlations among multiple meteorological factors.The correlations are expressed as edge features to model coupling information.Second,an edge convolution improved graph attention network(EC-IGAT)prediction model is proposed.The proposed model captures both local geometric information and global context to improve prediction accuracy.Finally,the superiority of the proposed method is validated using actual data from a PV power station in the Gobi Desert and other arid areas of Northwest China.