Research On Short-term Wind Power Forecasting Based On Improved K-Nearest Neighbor Method

The randomness and volatility of wind energy make large-scale wind power integration has become the biggest bottleneck restricting the development of wind power in China.Accurate prediction of wind farms output power is conducive to the absorption of more wind power and promote the development of wind power,which is of great significance to improve the safety and stability of the system and economic operation.Based on the power model of wind turbine,this thesis studies the main factors affecting the output power of wind turbines and introduces the principle,type,and error evaluation index of wind power forecasting method,which provides a good theoretical basis for the following forecasting modeling.Secondly,the shortcomings of short-term wind power forecasting based on the principal component K-nearest neighbor algorithm are proposed for the wind power forecasting method of traditional K-nearest neighbor algorithm,which has the disadvantages of computational complexity and high spatial complexity.The method has the advantages that the calculation amount can be reduced in each power forecasting process,so that the amount of memory occupied by the method is reduced,the efficiency is higher,and the space occupied by the data set can be reduced by the method.Taking the measured data of a wind farm in Jiangsu as an example,this method is compared with the traditional K-nearest neighbor algorithm and the support vector machine to obtain the wind power forecasting method.The wind power forecasting method based on the principal component K-nearest neighbor algorithm is superior in accuracy and efficiency.The traditional K-nearest neighbor algorithm and the forecasting method of the support vector machine.Finally,the influence of intelligent algorithm and support vector machine classification on the principal component K-nearest neighbor algorithm is studied.A short-term wind power prediction method based on K-nearest neighbor algorithm is proposed.Aiming at the problem that the principal component K-nearest neighbor algorithm output function weight is single,the reference combination prediction method is used to train it to obtain a set of weights so that the error is smaller.For the problem that the prediction error of some data in the prediction result is large,the support vector machine is introduced to judge it.For this type of data,there will be a set of independent parameters corresponding to it.In the prediction,it is first classified,and the prediction method of different parameters is selected by the classification result to predict it.Finally,taking the measured data of a wind farm in Jiangsu as an example,the simulation comparison with the traditional K-nearest neighbor algorithm and the support vector machine wind power prediction method shows that the error of the wind power prediction method is smaller than the traditional K-nearest neighbor algorithm and support vector machine.Compared with the wind power prediction method based on the principal component K-nearest neighbor algorithm,the prediction method is found to be smaller than the latter,which proves the effectiveness of the proposed method.