Short-term photovoltaic power prediction based on similar day clustering and VMD-SSA-LSTM

To mitigate the high variability and unpredictability in photovoltaic power generation, this paper introduces a prediction method for photovoltaic models using similar day clustering combined with the variational mode decomposition-sparrow search algorithm-long short-term memory neural network (VMD-SSA-LSTM). First, the Kmeans algorithm is used for clustering the photovoltaic data, and the main meteorological factors are selected using the Pearson correlation coefficient. Then, on the basis of selecting the historical similar days of the day to be predicted using a weighted comprehensive evaluation index based on grey correlation analysis and cosine similarity, the variational mode decomposition (VMD) is used to decompose the photovoltaic signal. Finally, the sparrow search algorithm-long short-term memory neural network (SSA-LSTM) model is used to predict each subsequence, and the results are aggregated to obtain the final prediction. Simulation results demonstrate that the proposed VMD-SSA-LSTM model significantly enhances the accuracy of photovoltaic output prediction under similar day clustering, outperforming single models and other combined models.