A Hybrid Wind Speed Prediction Model Based On Clustering

With the depletion of traditional energy resources and the pollution caused by the use of traditional energy resources,the development and utilization of renewable resources has attracted increasing attention from many countries.As a kind of clean and safe renewable energy,wind energy is mainly used to generate electricity.Wind power generation will not cause pollution to the environment and cause damage to the ecology,it can also solve the problem of power shortage in areas where traditional energy is scarce and transportation is inconvenient,which is of great significance for promoting social development and sustainable human development.However,wind speed has great randomness and volatility,and is affected by weather factors,which causes certain difficulties to the stability of wind power generation.Therefore,in order to ensure the stable operation of wind power grid,accurate prediction of wind speed is particularly important.Wind speed prediction is conducive to the development of wind farm stroke power units on and off the plan is conducive to the timely adjustment of power dispatch departments to maintain the safety and stability of the power grid and improve economic and social benefits.In order to improve the accuracy of wind speed forecasting,on the basis of the existing machine learning algorithm,this paper proposed a hybrid wind speed forecasting algorithm based on clustering for the multi-step forward wind speed prediction,the model is based on the historical wind speed time series and meteorological data,combined with the k-means clustering,singular spectrum analysis,extreme learning machine,the radial basis network,ridgelet network,echo state network,as well as the social cognitive optimization algorithm and reciprocal method of variance.This method firstly using singular spectrum analysis to denoise original history wind speed time series through the decomposition and reconstruction,and the gray correlation analysis is used to select meteorological features that are highly correlated with wind speed time series,the selected meteorological characteristics are as the characteristic of K-means clustering to cluster the wind speed time series,Second,the best input lag of single neural network are determined by training the wind speed data after noise,the optimal cluster combined model of each cluster are determined too,and use the SCO and reciprocal method of variance to optimization the weights of cluster combined model,at last,the trained cluster combined model is used to forecast the wind speed time series,and the final forecasting results are obtained.In order to evaluate the forecasting performance of the model proposed in this paper,the data of the M2 tower of the national wind power technology center in the United States was selected for the training and forecasting.The experimental results show that the mixed forecasting model based on clustering proposed in this paper can improve the wind speed forecasting accuracy.