Optimize VMD-LSTM photovoltaic force combination prediction model based on SSA parameter optimization

In order to solve the problem of low prediction accuracy caused by intermittency and uncertainty of photovoltaic output, In this paper, a parameter optimization variation mode decomposition (variation mode decomposition) based on sparrow search algorithm (SSA) is proposed. Combined prediction model of VMD and long short-term memory neural network (LSTM). Firstly, Pearson correlation coefficient (PCC) was used to analyze the factors affecting PV output. Secondly, the core parameters of VMD (k value and penalty factor coefficient α) are automatically optimized by SSA. After decomposing the original PV output time series by SSA-VMD, the learning parameters in LSTM are optimized by SSA, and the SSALSTM prediction model is established for each sub-sequence obtained by decomposition. Finally, the predicted values of each subsequence are superimposed and the final predicted values are obtained. The measured data of a PV power station in Xizang Autonomous Region are used to verify the results. The results show that the prediction accuracy of the proposed combined model SSA-VMD-LSTM is significantly improved compared with the original model LSTM and the unoptimized model VMD-LSTM. Therefore, the SSA parameter optimization method can effectively improve the prediction accuracy of VMD-LSTM combined model, and is more adaptable in PV output prediction.