Research on Photovoltaic Power Prediction Based on Mode Decomposition and LSTM-IDBO-GRU

In order to improve the prediction accuracy of photovoltaic power generation, a combined prediction model based on the rime optimization algorithm (RIME), variational mode decomposition (VMD), long short-term memory network (LSTM), and improved dung beetle optimization algorithm (IDBO) for optimizing gated recurrent unit (GRU) was proposed. Firstly, the minimum envelope entropy was employed as the fitness function of the optimization algorithm, and RIME was applied to optimize VMD to find the optimal parameter combinations for the number of intrinsic mode functions (IMF) components and the penalty factors. Secondly, these components were classified into low-frequency and high-frequency based on the zero-crossing rate, and LSTM model was used to predict low frequency components. To solve the problem that the prediction accuracy of high frequency components cannot be guaranteed, various strategies were employed to improve the traditional dung beetle optimization algorithm, and an IDBO-GRU model was utilized to predict high-frequency components. Finally, the predicted results were reconstructed to obtain the final photovoltaic power result. Comparative experimental results indicate that the proposed combined model outperforms the other models, including VMD-LSTM-GRU, RIME-VMD-LSTM-GRU, and RIME-VMD-LSTM-dung beetle optimization algorithm (DBO)-GRU, with the smallest error evaluation indexes and higher prediction accuracy.