Wind Power Prediction Based On VMD Decomposition And MRMR Feature Information Selection

Wind power data has strong volatility,intermittence and randomness.There are many defects in the current wind power prediction methods,which can not adapt to the characteristics of wind power data,and it is difficult to meet the security and stability of power system operation.For these reasons,this paper puts forward a new idea of "decomposition-feature selection-multi model prediction-integration".The main research ideas are as follows:In order to make full use of the fluctuation,intermittence and randomness of wind power data and improve the decomposition process of traditional decomposition methods such as Empirical Mode Decomposition(EMD)and Ensemble Empirical Mode Decomposition(EEMD)for the modal aliasing phenomenon,this paper,firstly,the variable mode decomposition(VMD),which is completely different from EMD and EEMD,is used to decompose the wind power data and obtain the modal components with different wave characteristics.In order to consider the influence of meteorological factors such as wind speed,air pressure and temperature on wind power data changes and make up for the lack of information in the decomposition process of VMD,this paper uses the feature selection method of max-relevance and min-redundancy(mRMR)based on mutual information(MI)to analyze and select the relevant feature set of each component.This method not only considers the correlation between variables,but also considers the redundancy between variables,reducing the dimension of each component feature matrix.In order to avoid the limitation of single prediction model,the back propagation neural network(BPNN)with strong self-learning ability and good nonlinear fitting ability and the least squares support vector machine(LS-SVM)with simple structure,fast learning speed,good generalization and good regression prediction performance will be used in this paper to predict each component.In the process of prediction,a new input matrix is established by combining the feature set selected by mRMR with each component.After the prediction of each component,BPNN is integrated to get the final wind power prediction value.In this paper,we use two groups of data of the wind farm in Northern Shaanxi and Yunnan to test the model,and compare the experimental results with other prediction methods.