基于神经网络模型误差补偿技术的对流层延迟模型研究

摘要

Aiming at the low accuracy of the traditional tropospheric delay model,a high precision fusion model for the northern hemisphere is established based on the Hopfield model,using the neural network model error compensation technique.Taking the zenith tropospheric delay (ZTD) as the approximate "true value" of the meteorological sounding data of more than 120 observing stations in 2010 provided by the University of Wyoming,this paper analyzes and compares the Hopfield model,the traditional BP model,and the computational accuracy of the fusion model.The results show that the root mean square error (RMSE) of the Hopfield model is 35.31 mm,the RMSE of the traditional BP model is 30.34 mm,and the RMSE of the fusion model is 23.31 mm.%针对传统对流层延迟模型精度较低的缺点,基于神经网络模型误差补偿技术,在Hopfield模型基础上建立一个适用于北半球的高精度融合模型.以Wyoming大学提供的2010年全球120多个观测台站的气象探空数据精密解算的天顶对流层延迟(ZTD)作为近似“真值”,分析比较Hopfield模型、传统BP模型和融合模型的计算精度.结果表明,Hopfield模型的均方根误差(RMSE)为35.31mm,传统BP模型为30.34mm,融合模型为23.31 mm.