Research And Application Of Neural Network And Its Hybrid Model To Short Term Wind Speed Prediction

Wind energy is a kind of green renewable energy,it has been widely used in the world recently.However,due to the randomness and intermittence of wind speed data,wind power generation has great instability,Therefore,the prediction of wind speed is of great significance to the stability of wind power generation.Wind speed data has nonlinearity,the statistical model can only predict the linear characteristics of data.With the rapid development of machine learning,neural network can fit the nonlinear characteristics of data,so it has been widely used in the field of wind speed prediction.In order to solve the problem of low prediction accuracy of existing single model,This paper is based on three-layer feed forward neural network,combined with the advantages of statistical model.A new hybrid model is proposed to predict wind speed based on neural network as follows:Firstly,the neural network(NN)trained by two different training rules and used to predict the x-axis sequence of lorenz-63 system and wind speed data,and study the prediction results of first order transition rules(FOTR)and fuzzy first order transition rules(FFOTR).The results show that the predicted value of the neural network trained by the fuzzy first-order transformation rule is closer to the actual value.Secondly,the hybrid system of the feedback neural network is used and Auto Regressive Integrated Moving Average model to predict wind speed in the different stations.The weights of two single models are obtained by dominance matrix.The results of several numerical experiments show that the hybrid model is better than single model ARIMA?NN-FOTR and NN-FFOTR models.Thirdly,the hybrid system of the ARIMA-NN-FFOTR is used to predict the wind speed of different stations.The experimental results show that compared with ARIMA-NN-FOTR model,the hybrid model has better prediction effect.NN-FFOTR model makes fuzzy calculation on the values in the partition,so as to improve the utilization of the data in the partition and improve the prediction results.Fourthly,the hybrid system of EMD-ARIMA-NN-FFOTR is used to predict wind speed of different sites.The results show that the hybrid model has high prediction effect on wind speed data of different stations and seasons,Compared with ARIMA-NN-FFOTR,EMD decomposes the original wind speed data to get several components,these components can be predicted by selecting appropriate models,and obtained the higher prediction accuracy.