Research On Combined Forecasting Model Of Wind Power Based On Support Vector Machine

The fossil energy has become the mainstay of global energy supply since the industrial revolution,which strongly promotes the progress of the society and the development of the world economy.With the continuous progress of science and technology,the exploitation and utilization of fossil energy are increasing on a daily broadening scale,and the energy crisis,environmental pollution and other problems are becoming increasingly prominent.Thus,how to find renewable clean energy becomes an urgent need for human's development.Around the world,wind energy resources have become the target to be developed because of its abundant reserves,wide distribution,small pollution and other advantages.The natural wind is volatile and intermittent,so there is strong uncertainty for the output power of wind power generation in the process of converting wind energy to electric energy.When large-scale wind power is incorporated into the grid,it will exert influence on generation,transformation,transmission and distribution of power grid and bring challenges to the safety of grid operation.This problem can be solved effectively by the wind power prediction technology,which can guide the grid to absorb more wind power and improve the safety of wind power grid connection.Thus,to study the wind power prediction technology can play a good role of promotion in the development of wind power industry.The background and significance of the research on wind power prediction technology are analyzed in this thesis at the beginning,as well as the current research status at home and abroad.A detailed description to the category of wind power prediction methods is proposed in this thesis.Among the many prediction methods commonly used at present,the prediction method based on support vector computer is widely used because of its advantages in solving small sample,nonlinear and other problems.The support vector machine is maked use of in this thesis to build wind power forecasting model with multiple influencing factors.However,the standard support vector machine is used to solve quadratic programming problems,leading to large amount of calculation,and its prediction results are easy to be affected by the parameter selection.As for the optimization of parameter penalty factor C,insensitive parameter ? and kernel parameter ? of support vector machine model,heuristic algorithm is used for optimization.Considering that the single optimization method has a high limitation,the bat algorithm(BA)and improved particle swarm optimization(IPSO)algorithm are combined to solve the problem that the support vector machine is easy to be affected by parameters.As for the wind speed and wind power series with nonlinearity and nonstationarity,the empirical mode decomposition(EMD)method is selected for treatment.The original sequence is decomposed into several subsequences with different frequency characteristics,which are respectively predicted,and then the predicted results are superposed to get the final prediction results of wind power.The existence of mode mixing could cause serious aliasing in the time-frequency distribution of the decomposed subsequence.Although the ensemble empirical mode decomposition(EEMD)alleviates this phenomenon,the complexity and computation of the algorithm are very large.To solve the above problems,the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)is advanced in this thesis.Finally,the combined forecasting model of wind power based on CEEMDAN and BAIPSO optimized support vector machine is proposed in this thesis.Through the case simulation analysis,the effectiveness of the combined forecasting model of wind power proposed in this thesis is verified.