Ultra-short-term Prediction Of Wind Power Output Based On Data-driven

In recent years,the installed capacity of wind turbines and wind power grid-connected capacity are increasing year by year.Due to the volatility and uncertainty of wind speed,large-scale wind power integration will bring problems such as frequency and voltage regulation to the power grid,affecting the stable operation of the power grid.The accurate prediction of wind power can effectively reduce the impact of wind power grid connection on grid stability.This paper deeply explores the characteristics of historical wind power data and studies the ultra-short-term prediction of wind power output.The main work is as follows:Firstly,a method of multi-model correction of historical wind power error data is proposed.Due to various factors,historical wind power data has erroneous data that is not conducive to wind power prediction.If wind power prediction is based on such data,it will inevitably cause large deviations.If the outlier data is directly eliminated only based on statistical characteristics,the physical concept is not clear and the accuracy is difficult to guarantee.Therefore,this paper analyzes the different causes of erroneous data,and proposes different screening methods and reconstruction methods.The wind curtailment data is reconstructed by the power curve method;outlier data is identified by the quartile method and reconstructed by the interpolation method;the electromagnetic interference data is identified by the fuzzy clustering method and reconstructed by the network topological structure method.Finally,the wind speed-wind power scatter diagram before and after the screening is given,and the accuracy of the reconstructed data is quantitatively analyzed based on the average relative error and accuracy.Secondly,a post-partial polynomial RBF neural network wind power prediction model based on particle swarm optimization algorithm is established.First,the input variables of the prediction model according to the correlation analysis of wind power is selected;then the sample data similar to the prediction point is selected as the training sample according to the wind speed characteristics and the subtraction clustering algorithm;and finally add the post-partial polynomial to the traditional RBF neural network model in order to solve the problem that the traditional model is easy to fall into the local optimal solution,and realize ultra-short-term wind power prediction.Finally,a wind power prediction error correction model based on reverse prediction is proposed.By analyzing the basic characteristics of the wind power time series,the consistency of the forward sequence and the reverse sequence is found,and the probability distribution of the wind power prediction error is depicted using the Gaussian distribution and its sample estimate is calculated.Taking single-step prediction as an example,a modified model of wind power prediction error is constructed,and the accuracy of the model is verified by an example.