Day-ahead generation power forecasting for wind-photovoltaic clusters considering complementary generation characteristics

With the large-scale development of wind and photovoltaic (PV) power stations, accurate power forecasting for regional wind-PV clusters is crucial for the stability and reliability of regional power systems. Therefore, this paper proposes a power forecasting method for wind-PV clusters considering output complementarity. Firstly, a multi-dimensional indicator system is established to quantify wind-PV complementarity, so as to systematically assess the output complementarity between any two wind and PV power stations. Secondly, based on the quantified output complementarity between plants of different generation types and the output correlation between plants of the same generation type, a graph convolutional network (GCN) based day-ahead power forecasting model is constructed for wind-PV clusters. Finally, a multi-task learning mechanism is introduced to jointly optimize forecasting tasks for different seasonal scenarios. Simulation results demonstrate that the integration of wind-PV complementarity and multi-task learning mechanism effectively improves the accuracy of power forecasting for wind-PV clusters.