【作者】 徐磊；

【导师】 高井祥；

【作者基本信息】 中国矿业大学 ， 大地测量学与测量工程， 2023， 博士

【摘要】 电离层是地球大气最为活跃的区域之一,由大量的自由电子和离子组成。电离层粒子活动驱动力来自于太阳活动、地磁活动和行星扰动,其异常活动会对人类的生存环境带来灾难性的影响。因此,通过监测电离层活动可以预测极端天气和地质灾害的产生,为人类应对灾难提供决策依据。从信号传播的角度来看,电离层误差是无线电传播主要的误差源之一,严重影响着信号的传播功率,更甚者完全阻断信号传播。在信息化时代,信息传播的终止意味着生活的停滞,因此研究电离层的精细变化具有重要的现实意义。反过来,利用电离层对信号的作用机制,基于全球导航卫星系统(Global Navigation Satellite System,GNSS)技术,采用被动方式获取电离层电子含量是目前最行之有效的监测手段之一。随着GNSS全球化服务进程的推进,以及各种电离层监测与建模技术的不断发展,加之人类对电离层高精度模型需求的日益提高,基于GNSS的电离层监测与建模技术将引来新的机遇和挑战。本文紧跟高精度定位的需求,针对区域电离层建模过程中原始信息精度、空间模型假设和数学函数模型的不足,提出电离层建模数据处理和模型建立的精细化方法,并完成了模型精度的内外部验证。主要研究内容和结论如下:(1)提出了一种伪距多路径误差修正的电离层原始信息提取模型伪距多路径误差是制约传统载波平滑伪距(Carrier-to-Code leveling,CCL)技术精度的主要误差源,CCL是目前使用最为普遍的电离层提取技术。本文基于非差非组合精密单点定位(Undifferenced and Uncombined Precise Point Positioning,UD-UC-PPP)双频伪距残差,采用吉洪诺夫正则化对伪距残差进行去噪处理,还原多路径误差的系统特性。鉴于约束吉洪诺夫一次差分正则化法方程的特殊结构,采用托马斯快速算法对问题进行求解。采用自举法确定问题求解过程中引入的超参数。基于恒星日滤波算法,实现了双频伪距多路径误差实时改正。以PPP提取的电离层作为参考,相比CCL方法,多路径误差模型改正的电离层原始信息提取精度提高达到80%;GPS、Galileo和BDS电离层共站单差弧间差精度提升较大,约为50%,GLONASS精度提提升不明显;电离层模型精度提高约为10%。(2)提出了一种接收机频间偏差短时变化建模方法本文基于改进的载波平滑伪距(Modified CCL,MCCL)估计得到的接收机频间偏差(Receiver Differential Code Bias,RDCB)短时变化序列,采用核函数将其表示为以时间为自变量的多个核函数的和。为了解决函数的不适定问题,采用L1正则化约束保证问题的稀疏性和凸优化性质。为了获得函数的稀疏解,采用快速迭代收敛阈值算法(Fast Iterative Shrinkage Thresholding Algorithm,FISTA)求解法方程。采用赤池信息量准则(Akaike Information Criterion,AIC)或广义交叉验证(Generalized cross validation,GCV)法确定问题求解过程中引入的超参数。实验结果表明,稀疏核函数能准确地描述RDCB短时变化特征,相比传统CCL方法,本文方法电离层提取精度提高约30%～50%;GNSS卫星DCB估计稳定性增益10%～30%;在高频数据建模方面,本文方法的计算效率是MCCL方法的4倍,相比传统CCL方法,RDCB估计精度提高达50%。(3)建立了一种可变的电离层薄层高度模型由于电离层电子密度的时空变化特征,固定的薄层有效高度(Ionospheric Effective Height,IEH)模型近似表示全球或者区域电离层分布不合理。本文借助神经网络多层感知的隐性表达能力,将多个标定电离层电子变化的参数作为模型输入,输出电离层薄层高度。在训练过程中,将IGS GIM的VTEC和实测STEC通过投影函数解算的IEH当作期望输出,全天分6个时段训练神经网络。本文测试了两种神经网络——基于遗传算法优化的后向传播神经网络(BackPropagation Neural Network Optimized by a Genetic Algorithm,BP-NN-GA)和径向基神经网络(Radial Basis Function Neural Network,RBF-NN)的应用效果。计算可变高度IEH模型在检核站上获得的VTEC与GIM VTEC RMS值,比先前的研究方法具有更小的RMS值;在电离层建模上,可变IEH模型比固定IEH模型具有更高的电离层拟合精度,基于RBF-NN和BP-NN-GA高度模型精度提高约为18%和14%。(4)构建了一种分辨率可变的区域电离层数学模型基于Ne Quick广播电离层模型,分析电离层的时空变化特征,并据此探索区域电离层模型在时空分辨率上的具体形式。采用AIC准则确定不同时段Ne Quick区域电离层多项式数学模型在经纬度上的阶数,并以此数学模型对区域电离层进行建模。在电离层相对平静的美国地区,可变分辨率模型和固定多项式模型表现差别不大。在电离层活跃的香港地区,相比固定多项式模型,可变分辨率模型残差精度提升20%,拟合精度提升11%;在一致性分析上,可变分辨率模型更能表现电离层的变化特征,在测站上空的一致性提升约为7%;在d STEC精度分析上,相比固定多项式模型,可变分辨率模型在测区所有测站上的精度提升约为12%。该论文包含图104幅,表32个,参考文献256篇。更多还原

【Abstract】 The ionosphere is one of the most active regions of the Earth’s atmosphere,composed of numbers of free electrons and ions.The driving force of ionospheric activity comes from solar activity,geomagnetic activity,and planetary disturbances,which can have catastrophic impacts on the human living environment.Therefore,monitoring the activity of the ionosphere can predict the occurrence of extreme weather and geological disaster,providing decision-making basis for human disaster prevention.From the perspective of signal propagation,the ionosphere is one of the main sources of error in radio propagation,seriously affecting the propagation power of signals,and even completely blocking signal propagation.In this information age,the termination of information dissemination means the stagnation of life,so studying the fine changes in the ionosphere has important practical significance.In turn,based on the mechanism of the ionospheric effect on signals,monitoring ionospheric changes with GNSS technologies in passive form is one of the effective means.With the advancement of the GNSS global service process,the continuous improvement of various ionospheric monitoring and modeling technologies,and the increasing demand for high-accuracy ionospheric model of human beings,ionospheric monitoring and modeling technologies based on GNSS will be confronted with new opportunities and challenges.To closely meet the demand of high-precision positioning,refined methods for ionospheric modeling data processing and model construction are proposed deal with the shortcomings of ionospheric information accuracy,spatial model assumption,and mathematical function model,and the internal and external verification of model accuracy are completed.The main research content and conclusions are as follows:(1)Proposing a new extraction model for ionospheric raw information with pseudo range multipath error correction.Pseudorange multipath error is the main error source that restricts ionospheric retrieval accuracy of the traditional Carrier to Code Leveling(CCL)technique that is currently the most commonly used ionospheric extraction technique.Based on the double frequency pseudo range residuals of Undifferenced and Uncombined Precise Point Positioning(UD-UC-PPP),The thesis uses the Tikhonov regularization method to denoise the pseudo range residuals,and to restore the system characteristics of multipath.Given the special structure of the normal equation with constrained by Tikhonov’s first-order deviation,the Thomas fast algorithm is used to solve the problem and then the bootstrap method is applied to determine the hyperparameters introduced during the solving process.Based on the sidereal filtering algorithm,real-time correction of dual frequency pseudo range multipath errors has been achieved.Compare to the traditional CCL,ionospheric information extracted by PPP having been taken as reference,the accuracy of the ionospheric information corrected by the multipath error model is improved up to 80%;ionospheric inter-arc differences between co-location stations of GPS,Galileo,and BDS observables retrieved by the proposed method are smaller,with an accuracy increase of about 50%,while the improvement accuracy on GLONASS observables is not significant;GIM VTEC regarded as reference,the accuracy of the ionospheric model using the proposed method has been improved by about 10%.(2)Proposing modeling method of short-term variations of receiver differential code bias.The study of the short-term variations of receiver differential code bias(RDCB)can not only understand the performance of the receiver,but also reduce the smoothing error of the ionosphere retrieval.Based on the RDCB offsets relative to the reference epoch derived from the modified Carrier-to-Code leveling(MCCL),the RDCB offsets is represented as the sum of multiple kernel functions with time as the independent variable.To solve the problem of ill-posed function,L1 regularization constraint is used to realize sparsity and convex optimization properties of the problem.To obtain the sparse solution,fast iterative shrinkage thresholding algorithm(FISTA)is used to obtain the resulting solution iteratively.The akaike information criterion(AIC)or generalized cross validation(GCV)method is used to determine the hyperparameters introduced during problem solving.The experimental results show that sparse kernel functions can effectively describe the short-term variations of RDCB,and compared to traditional CCL method,the accuracy of ionospheric extraction is improved by 30%～50%and GNSS satellite DCB estimation have improvement of 10% to 30% in stability.In terms of high-frequency data modeling,the computational efficiency of this method is four times higher than that of the MCCL method,and compared to the traditional CCL method,the RDCB estimation accuracy is improved by up to 50%.(3)Establishing a flexible and variable ionospheric thin layer height modelDue to the spatiotemporal variation characteristics of ionospheric electron density,it is unreasonable to establish a fixed ionospheric effective height(IEH)model for a global or regional ionospheric distribution.This thesis utilizes the implicit expression ability of neural network multi-layer perception in which multiple parameters calibrated ionospheric electronic changes are treated as model inputs and IEH of the ionospheric thin layer is the model output.During the training process,the desired IEH is calculated by the mapping function mapping VTEC of IGS GIM into measured STEC,and the neural network is trained 6 times corresponding to 6 periods during one day.Two neural networks,Back Propagation Neural Network Optimized by Genetic Algorithm(BPNN-GA)and Radial Basis Function Neural Network(RBF-NN),are trained in this thesis The RMS of VTECs calculated by the flexible IEH model and GIM at checking stations has smaller values than that of previous study.In terms of ionospheric modeling,the flexible IEH model based on RBF-NN and BP-NN-GA have higher ionospheric fitting accuracies than the fixed IEH model with accuracies improvements of approximately 18% and 14%.(4)Constructing a variable resolution regional ionospheric mathematical modelBased on the Ne Quick broadcast ionospheric model,the spatiotemporal variation characteristics of the ionosphere are analyzed to explore the specific forms of regional ionospheric models in terms of spatiotemporal resolution.The AIC criterion is applied to determine the order of the polynomial model on Ne Quick TEC in different periods and the final model form is used to model the regional ionosphere.In the United States with relatively calm ionospheric changes,there is little difference in performance between the flexible model and the fixed polynomial model.In Hong Kong with active ionospheric changes,compared to the fixed polynomial model,the flexible model has20% improvement in residual accuracy and an 11% improvement in fitting accuracy;In terms of consistency analysis,the flexible model is more capable of representing the characteristics of ionospheric changes,with a consistency improvement of about 7%over all stations;In terms of d STEC accuracy analysis,compared to the fixed polynomial model,the flexible model has an accuracy improvement of about 12% at all stations in the study area.This thesis includes 104 figures,32 tables,and 256 references.更多还原