ABSTRACT
Downbursts have enormous potential to jeopardize the serviceability performances of the long-span bridges, which are generally located in the mountainous and coastline areas. In order to investigate the behaviors of long-span bridges and improve the safety and comfort condition of driving road vehicles on the bridges under such wind events, it is essential to predict the short-term wind speed of downbursts. This paper introduces an improved multivariable prediction method wind forecasting model utilizing the vector auto-regressive with exogenous variables (VARX) model and compares this method with other forecasting models to verify the feasibility and effectiveness of the proposed strategy for wind forecasting application. First, correlations between environmental factors and wind attributes were analyzed to select the appropriate and relevant variables. Then, diverse forecasting models were generated and compared through error measures, verifying the good performance and accuracy of the chosen model. Further, the result of the improved wind speed forecasting model is compared with general model.
