基于DOA-BP神经网络的电离层TEC短期预测

Short-term prediction of ionospheric TEC based on DOA-BP neural network

影响全球导航卫星系统(GNSS)所需导航性能(RNP)的最大误差源之一是电离层延迟,该延迟与电离层总电子含量(TEC)成正比,因此TEC的准确预测直接影响到GNSS的RNP。探索性地使用2021年提出的澳洲野犬优化算法(DOA)优化反向传播(BP)神经网络,构建DOA-BP神经网络TEC短期预测模型,以欧洲定轨中心(CODE)提供的电离层TEC值作为数据集,训练、测试DOA-BP TEC短期预测模型,实现全球范围和中国区域不同电离层格网点处TEC值的高精度短期预测,并将预测结果分别与传统BP神经网络模型、麻雀搜索算法(SSA)优化的BP神经网络模型(SSA-BP)预测的TEC值进行对比分析,结果表明,相较传统BP模型,DOA-BP模型的TEC预测精度明显提高,且相比其他优化模型(如SSA-BP模型),预测的TEC精度也占一定优势,能准确反映全球不同时空下电离层TEC的变化特征,可作为电离层TEC短期预测的一种新方法。

One of the largest error sources affecting the Required Navigation Performance(RNP)of the Global Navi⁃gation Satellite System(GNSS)is ionospheric delay.This delay is proportional to the Total Electron Content(TEC)of the ionosphere,so the accurate forecasting of TEC directly affects the RNP of GNSS.The Dingo Optimization Algo⁃rithm(DOA)proposed in 2021 was used to optimize the Back Propagation(BP)neural network,and a TEC shortterm forecasting model was constructed based on the DOA-BP neural network.Using the ionospheric grid dot TEC values provided by the Center for Orbit Determination in Europe(CODE)as the data sets,the forecasting model was trained and tested to achieve high precision short-term forecasting of TEC values at different ionospheric grid points globally and in China.The predicted TEC values were compared with those predicted by the traditional BP neural net⁃work model and Sparrow Search Algorithm(SSA)optimized BP neural network model.It was found that compared with the traditional BP model,the DOA-BP model has significantly improved the forecasting accuracy of TEC.Com⁃pared with other optimized models(such as SSA-BP model),the DOA-BP model can also obtain better forecasting accuracy of TEC.The results show that the DOA-BP model can accurately reflect the change characteristics of iono⁃spheric TEC in different time and space around the world,and can be used as a new method for short-term forecast⁃ing of ionospheric TEC.