Research On Short-term Wind Power Prediction Of Wind Farms

With the dual pressure of primary energy depletion and fossil energy pollution,wind power generation has been recognized by all countries in the world.It is the most mature technology in renewable energy utilization and the most suitable for large-scale development.However,the random fluctuation and uncertainty of wind power generation have brought some difficulties to the dispatching department of power system,resulting in the high rejection rate of wind power generation in recent years.The short-term wind power prediction of wind farms is an effective way to solve the problem of wind power abandonment.A lot of research work has been carried out for the wind power prediction,but the accuracy of wind power prediction has not been well resolved.Based on this,this paper proposes to study the short-term wind power prediction of wind farms,the main research work is as follows:Firstly,the relationship between wind speed and wind direction and the output power of wind farm is quantified based on the data between the daily and annual changes of wind speed and wind direction and the output power of wind farm.A four component wind speed model is established,which is studied from 16 directions of basic wind speed,gust,gradual wind speed and noise wind speed,in order to improve the accuracy of prediction.Secondly,aiming at the good regression effect of BP neural network,the convergence of RBF algorithm into a good prediction model,the stationarity and simplicity of time series model,and the advantages of SVM algorithm sample data learning,BP algorithm,RBF algorithm,time series model and SVM algorithm are proposed to predict wind power respectively.The wind power prediction models of BP algorithm,RBF algorithm,time series model and SVM algorithm are established.The wind power prediction accuracy of the above four prediction methods is verified by five performance indexes of absolute error,mean square error,average relative error,root mean square error and correlation error.Finally,the simulation calculation is carried out based on the measured data.The prediction results show that the prediction accuracy of the above four methods can meet the requirements of the actual project,and the SVM algorithm model can be closer to the real value with higher accuracy.