Research On Short-term Wind Power Prediction Method Based On Fluctuation Process

The shortage of traditional fossil energy resources and the adverse impact on the ecological environment make the use of new energy sources such as wind power become the energy solutions of common concern of all countries in the world.The inherent intermittence,randomness and fluctuation of wind power make it more difficult for wind power Grid-connected than other power sources.Accurate short-term wind power forecasting can describe the fluctuation in the future.By coordinating with other power sources,the reserve capacity of power system can be reduced,and the operating cost of the system can be reduced,so as to mitigate the adverse impact of wind power on the power grid.Firstly,the development of wind power industry and the research status of wind power forecasting at home and abroad are summarized.On this basis,the main research directions of this paper are determined,including the design of historical data preprocessing scheme,short-term wind power deterministic prediction and uncertainty prediction.Firstly,on the basis of defining the significance of data pretreatment to improve the accuracy of wind power prediction,the scheme of anomaly data checking and missing data repairing for wind power data and measured power data is discussed.For wind measurement data,aiming at the rough problem of the commonly used methods for testing abnormal wind measurement data,the fine processing of index setting is carried out,and the link of singularity detection based on the maximum of wavelet transform modulus is supplemented.The situation of missing wind data is classified,and the repair schemes applicable to different situations are analyzed.For the measured power data,considering that the fluctuation and randomness of the power data are stronger than those of the wind data,the abnormal power data checking scheme based on the combination of time series diagram and P-V scatter plot and the joint missing power data repair scheme based on power curve and cubic spline interpolation are designed.Examples show that the above schemes have improved the inspection and repair effects.Then,a short-term wind power deterministic prediction scheme based on wave process matching technology is designed.The spectrum analysis of wind power fluctuation is carried out by using Hilbert-Huang transform.The intrinsic modal components are reconstructed into high frequency components with strong randomness and low frequency components reflecting fluctuation trend,and the frequency division prediction scheme is determined.The fluctuation process of low frequency components is identified and the characteristic parameter database is established.The matching relationship between the fluctuation process of wind speed and the fluctuation process of wind power is used to predict and correct wind power.Principal component analysis is used to reduce the dimension of wind resource elements and select appropriate training factors.BP neural network optimized by genetic algorithm is used to obtain high frequency prediction results.The deterministic prediction results are obtained by superposition.Examples show that the root mean square error of this method reaches 12.9%,which is obviously improved compared with the traditional forecasting methods such as persistence method,time series method and BP neural network method.Then,in order to supplement the information of wind power fluctuation which can not be provided by deterministic prediction,the uncertain prediction based on fluctuation process,namely interval prediction,is carried out.The difference of prediction errors between different fluctuation processes and different power levels in the performance of traditional error indicators and the distribution of error probability density is analyzed,and a classification interval prediction scheme is designed.The root cause of the multipeak phenomenon of error distribution,especially the occurrence of maximum error,is analyzed,and the prediction scheme of the fluctuation process is designed.Aiming at the lack of direct forecasting method combined with the actual situation of wind resources,a direct interval forecasting scheme for single peak fluctuation process based on numerical weather prediction is designed.Examples show that compared with the traditional interval prediction schemes,the three methods of classification prediction,segment prediction and direct prediction can greatly improve the pass rate and ensure the reliability and economy of interval prediction on the basis of ensuring the average width of the relative interval.Finally,the research results are summarized and analyzed,and the links that need to be optimized in the current method and the future research prospects are put forward.