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

Abstract

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