Research On Ultra Short Term Prediction Of Wind Power

With the rapid development of wind power technology,Large scale wind power integration will have a great impact on the power grid.Accurate prediction of wind power can reduce the negative impact of wind power integration on the grid.But the randomness of wind power makes the historical wind power data nonlinear.Therefore,on the basis of summarizing the existing prediction methods,this paper applies two modal decomposition methods in the field of signal processing to ultra-short-term wind power prediction,which can reduce the randomness of historical wind power data and improve the prediction accuracy.The main research work is as follows:1.Single wind power prediction models are introduced,and BP,RBF,GRNN neural networks and SVM based on polynomial kernel,RBF kernel and Sigmoid kernel are modeled.The SVM based on RBF kernel is used as the basic wind power prediction model by example simulation.2.The combination of EMD and SVM is applied to wind power prediction.Aiming at the shortcomings of EMD such as endpoint effect,the envelope fitting and endpoint continuation of EMD are improved,and the EMD-SVM wind power prediction model is established.Simulation results show that the image extension EMD-SVM model has better prediction accuracy.3.Because of the difficult problem of mode aliasing in EMD,a combined prediction model of VMD and SVM is proposed.Aiming at the problem that the number of modes in VMD is difficult to determine,mutual information is applied to VMD to determine the number of modes,and the VMD-SVM wind power prediction model is established.Simulation results show that the VMD-SVM model has higher prediction accuracy than the EMD-SVM model.4.Because the parameters of SVM are difficult to choose,GWO is proposed to optimize the parameters of SVM.Aiming at the disadvantage that GWO is easy to fall into local optimal solution,IGWO with mixed strategies is proposed,and the VMD-IGWO-SVM wind power prediction model is established.Simulation results show that IGWO has better optimization ability than traditional swarm intelligence algorithm and the VMD-IGWO-SVM model has better prediction performance.