由于电离层电子密度随时间变化,且空间分布不均匀,对不同频段的无线电波产生延缓和折射,因此电离层电子密度变化是影响短波通信、卫星通信、全球导航卫星系统和其他空间通信质量的一个主要因素,本文对全球电离层电子密度(Number of elec...由于电离层电子密度随时间变化,且空间分布不均匀,对不同频段的无线电波产生延缓和折射,因此电离层电子密度变化是影响短波通信、卫星通信、全球导航卫星系统和其他空间通信质量的一个主要因素,本文对全球电离层电子密度(Number of electron,Ne)的预测工作对短波通信设备三维射线实时追踪定位提供必要条件。本文采用国际电离层参考模型提供的2016年电离层Ne数据,根据数据的三维空间时间序列特征,搭建了自编码器和卷积长短期记忆(Convolutional Long Short-Term Memory Network,Conv LSTM)网络组成的网络结构,在不引入地球自转周期之外任何先验知识的条件下,对Ne数据进行深度学习并实现预测,首先通过实验对比了SGD、Adagrad、Adadelta、Adam、Adamax和Nadam六种优化算法的性能,又对比了三种预测策略的均方根误差(Root Mean Square Error, RMSE),1h-to-1h预测策略的全球平均RMSE为1.0 NEU(最大值的0.4%),1h-to-24h和24h-to-24h预测策略的全球平均RMSE为6.3 NEU(2.6%)。由实验结果得出以下结论,一是Nadam优化算法更适合电离层Ne的深度学习,二是1h预测策略的性能与之前类似的电离层TEC预测工作(RMSE高于1.5 TECU,最大值的1%)相比有竞争力,但预测时间太短且对数据的实时性要求较高,三是两种24h预测策略虽能实现长期预测但性能不理想,要实现三维空间时间序列的长期高精度预测需要进一步改善神经网络、模型结构和预测策略。展开更多
Ionospheric delay is one of the major error sources in GNSS navigation and positioning.Nowadays,the dual-frequency technique is the most widely used in ionospheric refraction correction.However,dual-frequency measurem...Ionospheric delay is one of the major error sources in GNSS navigation and positioning.Nowadays,the dual-frequency technique is the most widely used in ionospheric refraction correction.However,dual-frequency measurements can only eliminate the first-order term of ionospheric delay,while the effect of the second-order term on GNSS observations may be several centimeters.In this paper,two models,the International Reference Ionosphere (IRI) 2007 and International Geomagnetic Reference Field (IGRF) 11 are used to estimate the second-order term through the integral calculation method.Besides,the simplified single layer ionosphere model in a dipole moment approximation for the earth magnetic field is used.Since the traditional integral calculation method requires large calculation load and takes much time,it is not convenient for practical use.Additionally,although the simplified single layer ionosphere model is simple to implement,it results in larger errors.In this study,second-order term ionospheric correction formula proposed by Hoque (2007) is improved for estimating the second-order term at a global scale.Thus,it is more practicable to estimate the second-order term.More importantly,its results have a higher precision of the sub-millimeter level for a global scale in normal conditions.Compared with Hoque's original regional correction model,which calculates coefficients through polynomial fitting of elevation and latitudes,this study proposes a piece-wise look-up table and interpolation technique to modify Hoque model.Through utilizing a table file,the modified Hoque model can be conveniently implemented in an engineering software package,like as PANDA in this study.Through applying the proposed scheme for the second-order ionospheric correction into GNSS precise positioning in both PPP daily and epoch solutions,the results have shown south-shift characteristics in daily solution at a global scale and periodic change with VTEC daily variation in epoch positioning solution.展开更多
文摘由于电离层电子密度随时间变化,且空间分布不均匀,对不同频段的无线电波产生延缓和折射,因此电离层电子密度变化是影响短波通信、卫星通信、全球导航卫星系统和其他空间通信质量的一个主要因素,本文对全球电离层电子密度(Number of electron,Ne)的预测工作对短波通信设备三维射线实时追踪定位提供必要条件。本文采用国际电离层参考模型提供的2016年电离层Ne数据,根据数据的三维空间时间序列特征,搭建了自编码器和卷积长短期记忆(Convolutional Long Short-Term Memory Network,Conv LSTM)网络组成的网络结构,在不引入地球自转周期之外任何先验知识的条件下,对Ne数据进行深度学习并实现预测,首先通过实验对比了SGD、Adagrad、Adadelta、Adam、Adamax和Nadam六种优化算法的性能,又对比了三种预测策略的均方根误差(Root Mean Square Error, RMSE),1h-to-1h预测策略的全球平均RMSE为1.0 NEU(最大值的0.4%),1h-to-24h和24h-to-24h预测策略的全球平均RMSE为6.3 NEU(2.6%)。由实验结果得出以下结论,一是Nadam优化算法更适合电离层Ne的深度学习,二是1h预测策略的性能与之前类似的电离层TEC预测工作(RMSE高于1.5 TECU,最大值的1%)相比有竞争力,但预测时间太短且对数据的实时性要求较高,三是两种24h预测策略虽能实现长期预测但性能不理想,要实现三维空间时间序列的长期高精度预测需要进一步改善神经网络、模型结构和预测策略。
基金supported by the National Basic Research Project of China (Grant No.2009CB72400205)the National Natural Science Foundation of China (Grant No.40804005)the National High Technology Research and Development Program of China (Grant No.2009AA121401)
文摘Ionospheric delay is one of the major error sources in GNSS navigation and positioning.Nowadays,the dual-frequency technique is the most widely used in ionospheric refraction correction.However,dual-frequency measurements can only eliminate the first-order term of ionospheric delay,while the effect of the second-order term on GNSS observations may be several centimeters.In this paper,two models,the International Reference Ionosphere (IRI) 2007 and International Geomagnetic Reference Field (IGRF) 11 are used to estimate the second-order term through the integral calculation method.Besides,the simplified single layer ionosphere model in a dipole moment approximation for the earth magnetic field is used.Since the traditional integral calculation method requires large calculation load and takes much time,it is not convenient for practical use.Additionally,although the simplified single layer ionosphere model is simple to implement,it results in larger errors.In this study,second-order term ionospheric correction formula proposed by Hoque (2007) is improved for estimating the second-order term at a global scale.Thus,it is more practicable to estimate the second-order term.More importantly,its results have a higher precision of the sub-millimeter level for a global scale in normal conditions.Compared with Hoque's original regional correction model,which calculates coefficients through polynomial fitting of elevation and latitudes,this study proposes a piece-wise look-up table and interpolation technique to modify Hoque model.Through utilizing a table file,the modified Hoque model can be conveniently implemented in an engineering software package,like as PANDA in this study.Through applying the proposed scheme for the second-order ionospheric correction into GNSS precise positioning in both PPP daily and epoch solutions,the results have shown south-shift characteristics in daily solution at a global scale and periodic change with VTEC daily variation in epoch positioning solution.
