Analysis of Short-Term Photovoltaic Power Prediction Using EEMD-GRU Hybrid Neural Network

Due to the intermittent and highly volatile nature of PV power generation, the output power of PV devices fluctuates frequently and is difficult to control, making it challenging to precisely meet power generation plans and electricity demand. To address the issue of low prediction accuracy in PV power generation, this paper proposes a prediction model based on EEMD and GRU neural network. The EEMD algorithm is used to decompose the PV power signal into several stationary subsequences. Then, a GRU neural network model is employed to establish prediction models for each IMF component, ultimately constructing an EEMD-GRU-based PV power prediction model. Using a set of PV power data as an example, the simulation results demonstrate that the EEMD-GRU prediction model achieves higher prediction accuracy and smaller prediction errors compared to other models.