| At present, the global wind power industry is in a stage of rapid development. Wind energy is unstable energy, wind speed and direction changes all the time, when a large number of wind farms connect to the grid, it will bring a huge impact to the stability of the power grid. In order to ensure the stability of the power grid, electric power dispatching department must know in advance of the wind power output in the next few hours of wind farms, to reasonable arrangements of power dispatch plan. It s requires us to forecast the short-term wind power of the wind farm. And the wind speed forecasting is first step of wind power prediction.In this background, we use the wind speed data for empirical analysis and study the short-term prediction of the wind speed. First,we use of statistical methods to analysis the time series characteristics of wind speed and find the wind series has a strong volatility. Then we established the time-varying model for prediction of wind speed fluctuations:After use ARMA (3,2) model for filtering wind speed data, we use of ARCH Tests to verify the time-varying characteristics of the wind speed fluctuations;Then the residual sequence were established by the GARCH and EGARCH model of the residual followed the normal distribution, t distribution and generalized error distribution; Finally, we use of the information criteria such as AIC verified to using EGARCH model residuals obey t distribution than the traditional GARCH model can more accurately depict the heteroscedasticity of wind speed fluctuations.Taking into account the fitting error of the parameters and the uncertainty of distribution function, We propose the method for prediction of wind speed fluctuations of non-parametric Bootstrap method, Finally, we use the loss function with other prediction method for evaluation and analysis, found the forecasting accuracy of non parametric Bootstrap is relatively good. In the end, the prediction method is extended to other GARCH models to predict. |
PDF Full Text Request