A Distributed Photovoltaic Ultra-short-term Power Forecasting Method Based on Dynamic Graph Network with Shape-amplitude Loss Function

Accurate distributed photovoltaic power forecasting has significant function for the safety and stability of active distribution network. However, there are some drawbacks with the majority of the existing research for distributed photovoltaic power forecasting: 1) lacking consideration of the dynamic correlation among power sites; 2) lacking of a training loss function that can match the amplitude and shape of predicted value and true value simultaneously. Therefore, a dynamic graph network with a shape-amplitude loss function based distributed photovoltaic ultra-short-term power forecasting method is introduced. Firstly, a data-driven method is used to mine the dynamic correlation and the dynamic graph data can be generated to ensure the effective characterization of the correlation among distributed photovoltaics. Then, the dynamic graph network is constructed as the power forecasting model to realize the effective utilization of spatial-temporal correlation features. Finally, the shape-amplitude loss function which combines the Dynamic Time Warping and Mean Square Error is used as the criterion of model training to ensure the consistency of the forecasting value and the real in situ shape and amplitude. The forecasting performance of the introduced approach is demonstrated via a distributed photovoltaic power dataset in China.