Research On New Method Of Wind Power Forecast Based On Wind Speed Data Preprocessing

As a new energy generation technology with zero emissions and low operation costs,the wind power has achieved large-scale development and utilization worldwide.However,the inherent random volatility of the wind energy still poses a huge challenge to the safe integration of the wind power.The high-precision wind speed prediction work is a basic guarantee for the normal operation of the power grid and a reliable basis for the timely adjustment of the dispatch plan.The traditional wind speed prediction model can’t meet the wind farm prediction accuracy requirements,and most of the prediction methods focus on the improvement of numerical algorithms,lacking the mining of the real wind speed data characteristics.In this paper,the original wind speed data is taken as the research focus.Using the time-frequency analysis method to explore the essential characteristics of the wind speed sequence,and the targeted selection for wind speed prediction methods,thus,effectively improving the wind speed prediction effect.The main work includes the following aspects:(1)Using the empirical modal decomposition method to adaptively decomposethe original wind speed sequence,and according to the fluctuationcharacteristics of the empirical modal function,the least squares supportvector machine and the RBF neural network prediction model are selected topredict the high frequency component and the low frequency componentrespectively.This establishes a combined wind speed prediction model;(2)According to the similarity of the original wind speed sequence,the data isfiltered to reduce the complexity of the historical data,and considering themodal aliasing problem in the empirical mode decomposition method,we startto study the decomposition effect of the vibrational mode decomposition onthe original wind speed sequence,and relying on the global optimizationability of genetic algorithm to improve the BP neural prediction network,finally establish an effective wind speed prediction model.(3)The wind speed change is affected by many factors.The principal componentanalysis model is built to compress the historical data,and extracts thecomprehensive index.Using the independent component analysis model toanalyze the comprehensive index and the wind speed sequence and to reducethe mutual interference between the data.Establishing the RBF neuralprediction model for every component respectively.Compared with thetraditional neural network prediction model(BP,RBF,ELMAN),theintegrated multi-factor wind speed prediction model effectively improves thewind speed prediction level and the prediction effect of the RBF neuralnetwork.In this paper,according to the essential properties of the wind speed sequence,the wind speed prediction model can be established,which greatly reduces the prediction error,effectively improves the prediction effect of the traditional prediction methods and satisfies the accuracy requirements of real-time scheduling in wind farms.The research work in this paper helps the power system dispatching department to accurately assess the grid operation risk,scientifically formulate the power generation plan,reduce the power operation cost,and promote the large-scale development of green energy.