Ultra-short-term Photovoltaic Power Prediction Based on STCN Model

It is crucial to ensure the stability and security of the power system by accurately predicting photovoltaic power. Furthermore, the spatial and temporal dependence of regional photovoltaic power stations enhances the accuracy of ultra-short-term power predictions. In this paper, we propose applying a hybrid depth learning model to the irregularly distributed PV sites in the region, namely, the Spatio-temporal causal convolution network (STCN). It utilizes the advantages of a classical convolutional neural network to extract spatial features and a causal convolutional neural network to extract temporal features to predict ultra-short-term solar power generation. Based on the photovoltaic power data provided by the National Renewable Energy Laboratory (NREL), we compare the performance of the proposed model to that of other models. Experimental results indicate that the STCN model can improve the accuracy of prediction.