Power Prediction Based On Classification Of Large Wind Farm Units

The wind power output of large wind farms has fluctuations.The volatility of wind power will impact the power system when it is connected to the grid,affecting the safety,stability and economic operation of the power system.It will bring challenges to the safe and stable operation of the power system.Wind power prediction can be Tracking the wind power output of the wind farm and conducting power prediction,so that the power system can make a reasonable scheduling plan is one of the effective methods to solve this problem.The general wind power prediction method uses the wind condition of a specific fan position in the wind farm to predict the wind condition of the entire wind farm.However,for large wind farms,the prediction method is difficult to guarantee the prediction accuracy of the wind farm power;Modeling and prediction of each wind turbine in the field,the predicted calculation time is too long to meet the requirements of the power system for wind farm power prediction.Wind turbines of large wind farms are classified into wind turbines with similar characteristics in wind farms.The results of unit classification are used to predict wind farm power,which can improve prediction accuracy,reduce the complexity of prediction process and forecast time..Therefore,the power prediction based on the classification of large wind farm units is one of the methods to improve the prediction accuracy and improve the calculation efficiency of large wind farms.This paper studies the power prediction problem based on the classification of large wind farm units.The main work includes:(1)Using K-means clustering algorithm modeling,the classification clustering simulation of 33 wind turbines was carried out,and the clustering results of the number of unit classifications K=3~7 were obtained.Using the clustering algorithm's classification evaluation index Davidson Ding(DBI)index to determine the optimal classification number of 33 units of the wind farm.(2)The physical characteristics of the measured wind power are analyzed.The volatility characteristics are the most obvious.In order to facilitate the calculation of the clustering model and improve the clustering accuracy,the empirical modal decomposition(EMD)is used to decompose the measured wind power data.The complex measured wind power data is decomposed into an easily modeled eigenmode component(IMF)and a residual component matrix.(3)Based on the improved K-means clustering algorithm model of EMD,the improved clustering algorithm is used to classify and cluster the 33 wind turbines of the example,and the clustering results of the unit number K=3~7 are obtained.The classification index of the classification index is used to determine the optimal number of units in the improved clustering algorithm.The advantages and disadvantages of the clustering algorithm before and after the number of classifications K are compared.The results show that the improved K-means based on EMD Class algorithm clustering is more rational and better.(4)Based on the classification results of unit classification with improved clustering algorithm,a wind power prediction model with strong applicability is established.The power prediction results based on unit classification and unclassified mode are compared.The results prove that The power classification accuracy of the unit classification is higher and the prediction time is shorter.