东南大学学报（自然科学版）2025，Vol.55Issue(5)：1399-1405,7.DOI:10.3969/j.issn.1001-0505.2025.05.021
基于时序差分与双重修正的雷暴风速短时预测研究
Short-term prediction of thunderstorm wind speeds based on serial difference and double-step modification
摘要
Abstract
A short-term prediction method of wind speeds jointly driven by serial difference and double-step modification was investigated to predict thunderstorm wind speeds effectively.First,a differential operation was applied to the original measured wind speed,and the serial differential wind speed was taken as the predic-tion parameter.Second,a prediction model for differential sequence of wind speeds was established using a long short-term memory(LSTM)network.Then,real-time forecasted results were enhanced using the double-step modification considering the correlation between prediction error and predicted wind speeds.Fi-nally,the predicted results of thunderstorm wind speeds were obtained.Taking the thunderstorm wind speeds measured at Sutong Bridge as background,the effectiveness of the developed approach was verified in both single-step and multi-step predictions.The results demonstrate that the differential operation of the original wind speed can consider the non-stationary characteristics of wind speeds,resulting in a more effective predic-tion than directly using wind speed as input.The double-step modification method can compensate for the er-ror of the initially predicted results and improve the prediction accuracy of local extreme values and global wind speeds.The predicted results are satisfied in both single-step and multi-step predictions.关键词
雷暴风/实测风速/短时预测/机器学习/双重修正Key words
thunderstorm winds/measured wind speed/short-term prediction/machine learning/double-step modification分类
交通工程引用本文复制引用
陶天友,邓鹏,王浩,徐帆..基于时序差分与双重修正的雷暴风速短时预测研究[J].东南大学学报（自然科学版）,2025,55(5):1399-1405,7.基金项目
国家自然科学基金资助项目(52278486,52338011) （52278486,52338011）
江苏省自然科学基金资助项目(BK20240177) （BK20240177）
中央高校基本科研业务费资助项目(2242024RCB0007). （2242024RCB0007）
