Wind power forecasting based on improved variational mode decomposition and permutation entropy

Abstract Due to the significant intermittent, stochastic and non-stationary nature of wind power generation, it is difficult achieve desired prediction accuracy. Therefore, a method based on improved variational modal decomposition with permutation entropy proposed. First, meteorological data farms, Spearman correlation coefficient used filter that are strongly correlated establish model set; then original decomposed using technique eliminate noise in data, reconstructed into new subsequence by entropy; as input variables, stacking deeply integrated developed; finally results obtained optimizing hyperparameters algorithm through genetic algorithm. The validity verified real set from farm north-west China. show mean absolute error, root square error percentage at least 33.1%, 56.1% 54.2% compared autoregressive moving average model, support vector machine, long short-term memory, extreme gradient enhancement convolutional neural networks memory models, indicating has higher