Research On Short-term Wind Speed Forecasting Based On Deep Learning And Wavelet Decomposition

With the continuous progress of wind power generation technology in China,the proportion of wind power demand is increasing day by day.Wind speed forecasting plays an increasingly important role in power system,wind farm and power market.Due to the randomness of wind speed and the influence of many factors,the traditional short-term wind speed forecasting model is difficult to achieve satisfactory forecasting accuracy.In order to improve the accuracy of short-term wind speed forecasting,this paper combines deep learning theory with wavelet transform method and proposes a hybrid wind speed forecasting model wt-lstm-arma based on wavelet transform(WT),long short-term memory network(LSTM)and auto-regressive moving average(ARMA)model.In this paper,ARIMA,BP,Elman and LSTM four single wind speed forecasting models are studied,their basic principles and network structure are introduced,and the sliding window algorithm is used to realize multi-step wind speed forecasting.Through simulation,the forecasting results of each model are obtained and analyzed.A hybrid model wt-lstm-arma is proposed based on the LSTM model to solve the problem of low forecasting accuracy of a single wind speed forecasting model.The principle of the hybrid model is that first ly,the original wind speed sequence is decomposed into low and high frequency sub-series by using wavelet transform,then the LSTM and ARMA models are used to forecast the low and high frequency sub-series respectively,and finally the forecasting results of each sub-series are combined to obtain the final forecasting results.In order to verify the forecasting accuracy of the hybrid model wt-lstm-arma,four hybrid models and the corresponding single model were compared,and then wt-lstm-arma and the other three hybrid models were compared.In order to ensure the reliability of the conclusion,three different sets of wind speed data were used in the forecasting experiment.All the experimental results show that the hybrid model wt-lstm-arma is more accurate than other models in short-term wind speed forecasting and has satisfactory results.Deep learning theory and wavelet transform method are applicable to short-term wind speed forecasting.