Short-Term Forecasting for Photovoltaic Power Based on Successive Variational Modal Decomposition and Cascaded Deep Learning Model

Along with the improvement of new energy technology, the proportion of photovoltaic power generation is increasing. The accuracy of photovoltaic power generation prediction plays a key role in the stable operation of the power system. Hence, a hybrid deep learning framework based on successive variational modal decomposition, convolutional neural network and gated recurrent unit is presented. Firstly, the photovoltaic power is decomposed by successive variational modal decomposition. Secondly, convolutional neural layer is employed to extract the local key features between photovoltaic power data subsequences and weather data, and gated recurrent further explores the coupling time characteristics in each subsequence. Thirdly, the eventual photovoltaic power forecasting results are obtained via reconstructing the predicted values of each subsequence. The experimental results show that the proposed prediction model has lower computational complexity in data decomposition and exhibits higher prediction accuracy against other models.