| With the growth of the proportion of wind power integrated into power grids, wind power forecasting has become an indispensable tool for keeping the power grid operating safely and economically. This paper summarizes the concepts and classification of wind power forecasting and the practical application of wind power forecasting system and theoretical study of the latest developments, and discusses the performance evaluation and supervision measures of wind power forecasting. Study shows that wind power forecasting model based on chaotic time series has a high accuracy. In order to reduce the short-term wind power forecasting error based on chaotic forecasting model, two methods are proposed to improve the forecasting accuracy.Firstly, chaos system is sensitive to initial conditions according to chaos theory. In order to reduce the error brought by input data, this paper draws the idea of selecting sample based on similar days form load forecasting, and uses trend similarity to select similar days as the input of the forecasting model. In order to validate this method, a large number of simulation analyses are done considering different forecast horizons and wind power seasonal variance. The results show that the proposed method improved the accuracy effectively.Secondly, studies show that combination forecasting model can improve the wind power forecasting accuracy compared with single forecasting model, and the analysis of original data with decomposition of time series method can help to dig its internal information. Based on chaotic forecasting model, this paper proposes a short-term wind power combination model based on Fourier transform. Firstly, the original data will be decomposed into low-frequency trend component and high-frequency random component by Fourier transform. Secondly, according to the characteristics of each component, chaotic prediction model and ARM A prediction model are established separately to forecast the short-term wind power. In order to validate this method, some simulation analyses are done considering wind power seasonal variance. The results show that the forecasting errors of the proposed combination model reduce a lot compared with single model, and this method improves the performance of the short-term prediction of wind power effectively. |
