Ultra-short-term Forecast of Photovoltaic Power Based on VMD Error Correction and CNN-GRU-AM

Abstract

Accurate forecast of photovoltaic(PV) power is important for the security, stability and economic operation of power system. Existing prediction models based on variational mode decomposition (VMD) and deep learning network have not considered the decomposition error of VMD, and deep learning network is can be further improved. Therefore, a photovoltaic generation forecast method based on VMD error correction and multiple Gated Recurrent Unit(GRU) model is proposed. First, the VMD is used to decompose the power sequence of photovoltaic generation, and the decomposition error is analyzed. Then, the modal components are divided into three categories according to their volatility characteristics, and multiple GRU models are built to predict the modal components. The advantages of different GRU structures are used to achieve fine prediction of the modal components. Finally, an error correction model based on XGBoost is built to correct the decomposition error of VMD and improve the accuracy of photovoltaic power prediction. The correctness of the proposed method is verified by actual data.