Research On Short-Term Wind Power Forecasting Based On Combination Method

The rapid growth of energy demand has made the global energy crisis increasingly severe,and traditional energy sources such as coal and oil cannot meet the needs of green economic development.Therefore,the development and utilization of new energy sources have attracted widespread attention from all over the world.As a kind of low-carbon and environmentally friendly renewable energy,wind energy occupies an increasing share in the new energy power generation industry market.However,the characteristics of wind power,such as volatility,intermittentness and non-stationarity,make large-scale wind power grid-connected,which will cause great harm to the large power grid and seriously affect the safe and stable operation of the power system.The accurate prediction of short-term wind power is an effective way to reduce the impact on the grid,and it has become a research hotspot in the wind power industry.This paper studies the short-term wind power forecast based on the measured data of a wind farm.The main research contents are as follows:(1)The wind power data measured by the wind farm has strong nonlinearity and non-stationarity.Based on the summary and comparison of multiple data decomposition methods,this thesis uses the complete total empirical mode decomposition(CEEMD)method to decompose the wind power time series to obtain a series of relatively stable intrinsic modal components IMF,which reduces the original data The volatility and non-flatness of the wind power are conducive to the improvement of wind power forecasting accuracy.(2)Aiming at the problems of slow convergence speed and poor local search ability of traditional gray wolf optimization algorithms,an improved gray wolf optimization algorithm(IGWO)is proposed from the aspects of convergence factor and gray wolf position update method.Through experiments on several sets of benchmark test functions,it is verified that IGWO effectively overcomes the shortcomings of traditional GWO,improves the global and local search capabilities of GWO in functions,enhances the convergence speed of the algorithm,and improves the stability of the algorithm.Combined with least squares support vector machine(LSSVM),several component signals obtained by data decomposition are respectively established IGWO-LSSVM prediction model for prediction.Experiments with measured data from a wind farm in Shanxi have proved that the IGWO-LSSVM prediction model based on the CEEMD decomposition algorithm has high prediction accuracy.(3)Analyze the different characteristics of high-frequency and mid-low frequency components,establish a wind power combined forecasting model,and study the influence of different frequency components on the forecast results.Among them,the medium and low frequency components have small frequency fluctuations and gentle changes.The IGWO-LSSVM model is used for prediction;the high frequency components have obvious nonlinearity,and the volatility is still strong.The neural network has powerful processing capabilities for nonlinear data.Four kinds of neural networks establish a non-linear combined neural network prediction model.At the same time,the non-equal weight superposition method is adopted for the prediction results to achieve the purpose of improving the prediction accuracy.The simulation results show that the proposed combined model is an excellent wind power prediction model,which can further improve the accuracy of short-term wind power prediction.(4)In order to perform short-term wind power prediction operations conveniently and efficiently,a set of short-term wind power prediction system for wind farms was developed based on the MATLAB visualization platform GUI.The system includes three prediction models: based on LSSVM single method model,based on ANNs The single method model and the LSSVM-ANNs combined model proposed in this paper have a safe and reliable system,simple and friendly interface,and have a certain use value.