Research On Wind Power Forecasting Algorithm Based On Support Vector Machine Theort

With the increasing demand of wind power penetration and clean energy economy,efficient environmental protection system,reducing wind power waste and improving wind power utilization rate have become an important issue in wind power research.Therefore,it is particularly important to improve the power stability and power quality of wind farms.According to the characteristics of a wind farm in Liaoning Province,the wind power prediction based on support vector machine(SVM)is studied in this paper.Firstly,the principle of machine learning theory and related concepts in statistical learning theory are studied,including VC peacekeeping and empirical risk minimization.In the case of large sample data,by learning the mapping relationship between input and output,the SVM is comprehensively expounded and studied from two aspects of sample classification and regression.Secondly,the search process of cuckoo algorithm is studied.The cuckoo algorithm is improved by adding adaptive adjusting dynamic step size and simulated annealing process to cuckoo algorithm.Optimization of penalty factor parameters and kernel function parameters using improved cuckoo algorithm.The grid method,cuckoo algorithm and improved cuckoo algorithm are used to predict wind power in Matlab respectively,and the prediction results are analyzed.Empirical mode decomposition is introduced to stabilize wind power signals.Several important concepts of EMD,including instantaneous frequency,intrinsic mode function and characteristic time scale,are described.Emphasis is laid on the basic principles,processes,main properties and existing problems of EMD.Then,according to the existing problems,an improved method is proposed.Combining EMD with SVM of improved cuckoo algorithm,an algorithm model of support vector machine based on EMD is established,and the prediction accuracy of the model is proved by the actual data of wind farm.