Research On Ultra Short Term Wind Power Prediction Of Multisource Wind Speed Information

In recent years,with the depletion of fossil fuels in the world and the vigorous development of renewable energy in China,wind power generation technology plays an important role.However,wind power generation with wind volatility,intermittence and uncertainty brings serious challenges to the stable operation of power grid.Therefore,the accurate prediction of wind power has a huge impact on the development of the whole wind power industry.In view of this,improving the accuracy of wind power prediction has become an important topic.First of all,the wind speed measured by the wind tower of the wind farm and the wind energy data of numerical weather prediction(NWP)are analyzed.Its data characteristics are the precondition to improve the prediction accuracy.Through quantitative analysis,the NWP wind energy information and the correlation degree of the measured wind speed information of the wind tower are obtained,and then the NWP wind speed is modified by long short term memory(LSTM).Secondly,an ultra short term multi-step prediction model of wind power based on atomic sparse decomposition(ASD)and chaos theory is proposed.Firstly,the wind power time series is decomposed into several atomic trend components and one residual random component by using the good trend tracking characteristics of ASD;then the ultra short term prediction of the two components is carried out by using the adaptive prediction method and chaos theory respectively;finally,the prediction results of the two components are superposed to get the final wind power prediction results.Thirdly,by considering the future wind speed information to improve the prediction accuracy of power mutation,first of all,considering that the ideal wind power curve can not replace the real power output,a fitting method of wind power curve is proposed.Secondly,because of the difference of wind power curve,an improved FCM clustering is proposed to get the optimal division results,and a "switching mechanism" based At last,the validity of the model is verified by the examples of different wind farms.Finally,the hierarchical compensation method for the probability distribution of generalized error fraction is proposed.Firstly,the generalized error distribution is established for the historical predicted value and the historical actual power value,then the confidence intervals of the distribution under different confidence degrees are given,and different error layers are established according to the predicted intervals.Different error level compensation values are different,and the original prediction value is compensated.The results show that the compensation method can effectively improve the ultra short term prediction accuracy of wind power in large-scale wind farms.