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

The energy problem is related to the development of the global economy,and the current global non-renewable energy is facing a shortage.Wind energy is highly valued by people because of its abundant reserves and clean and pollution-free mining process.Wind speed is one of the main influencing factors of wind power generation,and it is of great significance to predict the wind speed sequence.Wind speed is highly non-linear and intermittent.Direct modeling and prediction of wind speed sequence,its nonlinearity and intermittency have a greater impact on the prediction accuracy.the decomposition algorithm is used to preprocess the wind speed sequence.The accuracy of wind speed prediction can be improved after reducing the complexity of the sequence.In view of the difficulty of setting the internal parameters of algorithms such as Complete Ensemble Empirical Mode Decomposition with Adaptive Noise and Long Short-term Memory Networks,the Sample Entropy of the decomposed subsequence and the predicted mean square error of the model are used as performance indicators,PSO algorithm is used to noise amplitude,noise to the number of times and network weight Values and other parameters are optimized to reduce the difficulty of modeling.This paper proposes a combined prediction model method based on PSO-sample entropy-CEEMDAN decomposition and PSO-LSTM.First,the PSO-SE-CEEMDAN algorithm is used to adaptively decompose the wind speed sequence into several sub-sequences to reduce the complexity of the original sequence,and the PSO-LSTM algorithm is used to model and predict the sub-sequences one by one.The short-term wind speed data of Quebec,Canada is used for experimental analysis,and the ARMA model,BP neural network model,LSTM model and the combined model proposed in this paper are used to predict short-term wind speed.Mean Absolute Percentage Error,Mean Absolute Error and Root Mean Square Error.Three performance indicators of square error verify the wind speed prediction accuracy of the above four prediction methods.The prediction results show that,compared with the classic ARMA,BP and LSTM methods,the prediction accuracy of this method is the highest,and the prediction results are closer to the true value.