Revision Study Of Wind Speed Forecast Based On Optimal Weighted Combination Model For Complex Underlying Surface In Gansu Province

Accurate prediction of short-term wind speed is particularly important for renewable energy utilization and planning.Numerical weather forecasting models have become an irreplaceable tool for weather forecasting today,but direct use of model forecasts may not guarantee a high degree of accuracy due to unavoidable errors and other constraints.Based on this,a combined revision model based on machine learning approach is proposed in this study.Considering the influence of complex underlying surface and seasonal factors on wind speed,this paper selects one meteorological station in each of the six underlying surfaces in Gansu Province as the study object and divides the data set by season.Firstly,LASSO regression is used for feature selection,then several numerical weather prediction models based on extreme learning machine(ELM),ELMAN neural network,BP neural network and least squares support vector machine(LSSVM)are built for wind speed prediction.To further improve the revision effect,two combinations of equal-weighted averaging(EWA)and optimal weighting(OW)algorithms are introduced on top of the four optimization models to build the combined models.The results show that the optimization and combination of the models can significantly improve the model revisions under the complex subsurface sub-seasonal experiments,and the OW-based combination model has the best effect on the wind speed revisions.Compared with the ECMWF,the wind speed errors are reduced after the model revisions,and the minimum RMSE,MAE and r＿MAE are reduced from 0.8839m/s,0.6890m/s,39.1084% to 0.3430m/s,0.2580m/s,23.4830%,and the maximum from5.9744m/s,5.2320 m /s,98.5495% to 1.6679m/s,2.1321m/s,24.7340%.This study introduces the OW combined model for revisions,which can fully exploit the advantages of model revisions under complex underlying surface and subseasonal tests,effectively reduce the model forecast errors and provide more accurate wind speed forecasts,which is of reference value for the study of wind speed revisions in complex underlying surface of Gansu Province.