Short-term Wind Speed Prediction Based On VMD And Improved LSTM Model

Due to the rapid economic development,the consumption of nonrenewable energy has increased,and the development of renewable energy has become a hot topic of social concern.In clean energy,sufficient wind energy resources can effectively alleviate environmental pollution and other issues,which is of great significance for resource conservation and environmental protection.Wind speed is characterized by instability,randomness,and uncontrollability.Predicting wind speed can reduce the operating costs of wind power systems and effectively improve wind energy conversion rates.Therefore,this paper proposes a combined model for short-term wind speed prediction to improve prediction accuracy.The main contents are as follows:Firstly,for the non-stationary nature of wind speed,the variational mode decomposition algorithm(VMD)is used to preprocess the data and decompose the original wind speed series into several subsequences.At the same time,the sparrow search algorithm(SSA)is selected to automatically optimize the decomposition levels and penalty factors of the decomposition algorithm,using envelope entropy as a fitness function.The simulation results show that the components decomposed by this algorithm are clearer,the frequency decomposition of the original wind speed series is more accurate,and its non-stationary nature is reduced.Then,in order to reduce the complexity of the combined model prediction,sample entropy(SE)is used to reconstruct the components to obtain a new subsequence.Secondly,aiming at the problem of artificially determining the parameters of the long short-term memory network(LSTM),the whale optimization algorithm(WOA)is selected to optimize its initial parameters.At the same time,the number of input neurons of the model is determined by calculating the number of lag periods of each component through the partial autocorrelation function(PACF).The simulation results show that the optimized LSTM model reduces the impact of human factors and improves the accuracy of prediction results.Finally,in order to solve the problem of poor performance of a single prediction model,this paper proposes a combined model based on the idea of decomposition,reconstruction,and prediction,which combines an improved VMD and LSTM.Empirical analysis using real-time wind speed data from the Wind Speed Observatory in Boulder,Colorado,USA,shows that the model has higher prediction accuracy and robustness in various evaluation indicators compared to other benchmark models,demonstrating its practical value in short-term wind speed prediction.