Short Term Wind Power Prediction Based On DWT-CE-BPNN Model

It is an important part of the transformation of energy structure in China to use wind energy instead of fossil energy to generate large-scale power.However,there are many technical problems in the power generation by wind energy.For example,the fluctuation of wind energy will lead to the problem of "abandoning wind power limit",and "abandoning wind power limit" will reduce the utilization rate of wind energy,increase the operation and maintenance cost of wind farm.the fluctuation of wind energy even threatens the safety of wind power in largescale grid connection.In view of the influence of wind instability and volatility on wind power system,a DWT-CE-BPNN model is constructed to predict the short-term wind power.Firstly,in order to reduce the outliers disturbance in wind power data on the prediction accuracy of the model,the thesis uses the interquartile range(IQR)and Markov chain Monte Carlo(MCMC)to eliminate and supplement the abnormal data in the data set.In order to reduce the influence of interference information on wind power prediction accuracy and the expense of redundant variables on model prediction,Pearson correlation coefficient(PCC)is used to analyze the correlation between various environmental variables and wind power,and the wind speed and air pressure with the highest correlation with wind power are selected as input variables.Then,the short-term wind power is predicted by using RBF neural network,Elman neural network and BP neural network.Then,BP neural network is selected as the basic model according to the predicted wind power value and various error evaluation indexes.Then,the model is established around BP neural network.In order to improve the current basic model prediction accuracy and convergence speed.this thesis constructs a DWT-CE-BPNN model.Firstly,the parameters of BP neural network are optimized by cross entropy optimization algorithm(CE).The input data is decomposed into different frequency components by discrete wavelet transform(DWT),and then the component is predicted by CE-BPNN model.Finally,the wind power prediction value is obtained by superposition of the prediction results.The experimental results prove the validity of DWT-CEBPNN model.In the end,in order to improve the prediction performance of DWT-CE-BPNN model,the CE algorithm is improved in this thesis.Firstly,the global search of flower pollination algorithm(FPA)is added to CE algorithm to increase the diversity of population.Then,a selective mutation operation is proposed to improve the convergence rate of the algorithm to the global optimal value.The experiment shows that the improved CE algorithm can effectively improve the stability and accuracy of model prediction.