The Research On Indirect Wind Power Forecast Of Wind Farm Based On Wind Speed Decomposition

Due to energy shortage and China’s strong support for new energy generation technologies,China’s new energy generation technologies are developing rapidly.The installed capacity of wind power generation ranks top three among new energy generation,and it is an important way of generating electricity.The intermittent and instability of wind not only have a serious impact on large-scale wind power grid connection,but also are not conducive to the reasonable dispatching of power grid.Reasonable prediction of wind power in wind farms is of positive significance for improving the problems existing in wind power grid connection,strengthening the consumption capacity of power grid,and reducing the phenomenon of "abandoning wind" in power grid.This paper proposes a method of indirect wind power prediction based on wind speed decomposition.Firstly,the wind speed is decomposed and predicted,and then the wind power is predicted based on the predicted wind speed results.The main research contents and results are as follows:(1)The decomposition of non-stationary nonlinear wind speed sequence is completed by improving the standard FEEMD.FEEMD algorithm can realize the fast decomposition of time-frequency signal,but the result of decomposition is not good due to the disadvantage of end effect in the decomposition.This paper proposes an improved FEEMD algorithm to improve the end effect.And through testing and comparing the decomposition effect of the FEEMD algorithm and the improved FEEMD algorithm,the results show that the improved FEEMD algorithm can improve the end effect problem.(2)An improved wind speed prediction model is proposed.The kernel parameters and regularization parameters determine the prediction accuracy of the LSSVM model,and the relevant parameters determined by manual methods are not accurate,resulting in unsatisfactory prediction results.This paper improves the FOA algorithm and verifies the improved FOA algorithm by using test functions.The parameters of the LSSVM model can be optimized using the improved FOA algorithm.Combined with the decomposition results of the improved FEEMD algorithm,an improved FEEMD-FOA-LSSVM model is established for each intrinsic mode function.The test proves that the improved FOA algorithm is used to optimize the parameters of the LSSVM model before prediction,and the prediction result is better than using LSSVM directly.(3)By fitting the wind power curve of the wind farm and combining the predicted wind speed results,the short-term wind power prediction of the wind farm is realized.Firstly,direct method,maximum method and Bins method are used to fit the wind power data,and secondly,a wind speed power curve model is established based on the improved GWO algorithm.In addition,In order to coordinate the global optimization and local optimization of the GWO algorithm,the chaotic population initialization and nonlinear convergence factors are used to optimize the GWO algorithm.Finally,the wind power prediction is completed by combining the wind speed prediction results.Through simulation verification,the power curve model based on the gray wolf optimization algorithm proposed in this paper is better than the maximum method,the direct method,and the Bins method when fitting the wind power curve.The result is more precise.