Tornado Tangential Velocity Prediction Using Multi-Source Data Fusion Method
Tornado has the characteristics of small scope of action,short duration and high intensity of action,and is the most frequent and destructive disaster in natural disasters.Due to the danger and the randomness of occurrence of the tornado,the field data is scarce and it is difficult to obtain complete wind field in the field measurement.In view of this,a data fusion method based on neural network model is proposed to realize the fusion of wind field data from different sources.The prediction effectiveness and the generalization ability of the model are verified.On this basis,the tangential velocity field of tornado is predicted.The results show that the average error of the driven model of measured data is more than 35％in the forecast of tornado with low swirl ratio,while the average error of the driven model with data fusion is less than 14％,which indicates that the fusion model has better prediction accuracy.In the prediction of tornadoes with high swirl ratio,the average error of the measured data-driven model is about 28％,while the average error of the data fusion driven model is less than 10％,indicating that the data fusion model still maintains high precision and has good generalization when predicting high swirl ratio.In the reconstructed fusion model,the vortex core of the low vortex ratio wind field is fractured,and the wind speed in the core area of the high vortex ratio wind field is significantly increased,and the coverage of the near-surface wind speed is also increased.The model can obtain wind field data near the ground and near the vortex core center,and improve the spatial resolution of wind speed in tornado field,providing important support for the improvement of structural wind resistance in tornado environment.
tornadomulti-source dataneural networkdata fusionwind field prediction
万方数据
维普
