CN113093225B - Wide-area and local-area fused high-precision ionospheric scintillation model establishment method - Google Patents
Wide-area and local-area fused high-precision ionospheric scintillation model establishment method Download PDFInfo
- Publication number
- CN113093225B CN113093225B CN202110189833.7A CN202110189833A CN113093225B CN 113093225 B CN113093225 B CN 113093225B CN 202110189833 A CN202110189833 A CN 202110189833A CN 113093225 B CN113093225 B CN 113093225B
- Authority
- CN
- China
- Prior art keywords
- model
- data
- ionospheric scintillation
- area
- flicker
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/01—Satellite radio beacon positioning systems transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
- G01S19/03—Cooperating elements; Interaction or communication between different cooperating elements or between cooperating elements and receivers
- G01S19/07—Cooperating elements; Interaction or communication between different cooperating elements or between cooperating elements and receivers providing data for correcting measured positioning data, e.g. DGPS [differential GPS] or ionosphere corrections
- G01S19/072—Ionosphere corrections
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
The invention provides a method for establishing a wide-area and local-area fused high-precision ionospheric scintillation model, which comprises the following steps: a) Acquiring original signal data of a space-based COSMIC satellite and a high-latitude area foundation CHAIN observation station, and preprocessing the acquired original signal data; b) Carrying out time-space characteristic preliminary analysis on the acquired ionospheric scintillation signal data; c) Establishing a wide area ionospheric scintillation ANN model, d) testing the wide area ionospheric scintillation ANN model, e) establishing a local ionospheric scintillation ANN model of a high-latitude polar region, f) and testing the local ionospheric scintillation ANN model of the high-latitude polar region. The ionospheric scintillation occurrence model is represented in the form of time rate of change index ROTI of total electron content by using two kinds of observation data of COSMIC and CHAIN, and the ionospheric scintillation model in a high latitude area is established by independently using CHAIN station data, so that wide-area and local area fusion is achieved, the performance of a receiver is improved, and the satellite navigation positioning accuracy is improved.
Description
Technical Field
The invention relates to the technical field of satellite navigation, in particular to a method for establishing a wide-area and local-area fused high-precision ionospheric scintillation model.
Background
Many
The ionosphere is a part of the earth's atmosphere and is located in an area above 60km above the earth's surface. Molecules and atoms in the atmosphere in this region are partially ionized by solar radiation (mainly ultraviolet radiation and X-rays), and there are a large number of positive ions and free electrons. The institute of electrical and electronics engineers standard (1969) defines the ionosphere in terms of the angle at which radio waves are affected as: "a portion of the earth's atmosphere in which ions and electrons are present in sufficient numbers to affect the propagation of radio waves".
With the rapid development of global satellite navigation systems, the ionosphere has become an essential research object for developing space exploration. Ionospheric interference is the largest source of error affecting satellite navigation positioning accuracy.
The occurrence of ionospheric flicker causes amplitude and phase fluctuations in the received GNSS signals, and these rapid fluctuations may degrade the positioning accuracy and even cause carrier phase cycle slip and loss of lock of the tracking loop. Based on the research on the physical process of electric wave propagation and the statistical analysis on a large amount of historical data, the physical structure and the activity rule of an ionosphere are known and mastered, the scintillation characteristic of the ionosphere is accurately simulated, and a high-precision ionosphere model is constructed, so that the method is an effective method for improving the precision of satellite navigation time service, speed measurement, positioning and the like.
The existing models can be divided into theoretical analysis models (namely a thin phase screen theory, a born approximation method, a huygens-fresnel integral theory method, a parabolic equation method, a multi-phase screen theory and the like), probability statistic models (typically, an AJ Stanford model and a Cornell model) and scintillation monitoring and forecasting systems (typically, a WBMOD model and a GISM model).
However, the physical structure of the ionized layer is complex, the influence factors of the ionized layer flicker are random and changeable, and the uncertainty of the relation between the irregular body and the ionized layer flicker, so that the theoretical analysis model has certain limitations. And two more typical probability statistical models, namely, an AJ Stanford model does not consider the limitation of flicker frequency, and a Cornell model does not consider the correlation between phase flicker and amplitude flicker. The foundation survey stations cannot be spread all over the world due to the limitation of geographical positions, the obtained ionospheric scintillation data is single, and the ionospheric scintillation research in a high latitude polar region is less. The flicker detection and prediction system mostly adopts data of a low latitude region in the modeling process, so that the error of the model in the prediction of a high latitude region is extremely large, and the ionosphere prediction can be only carried out in a local region.
Therefore, in order to solve the problem of large ionospheric scintillation prediction error in the prior art, a wide-area and local-area fused high-precision ionospheric scintillation model establishing method is needed.
Disclosure of Invention
One object of the present invention is to provide a method for establishing a wide-area and local-area fused high-precision ionospheric scintillation model, wherein the method comprises the following steps:
a) Acquiring original signal data of a space-based COSMIC satellite and a high latitude region foundation CHAIN observation station, preprocessing the acquired original signal data, and acquiring ionospheric scintillation signal data;
b) Carrying out primary analysis on the space-time characteristics of the acquired ionospheric scintillation signal data;
c) And establishing a wide-area ionospheric scintillation ANN model, which comprises the following steps:
dividing the data after the preliminary analysis of the space-time characteristics into first model construction data and first model test data, taking the first model construction data as the input signal of the neural network, training the neural network by means of an error back propagation algorithm,
obtaining an output result ROTI through an error back propagation algorithm;
d) Testing a wide-area ionospheric scintillation ANN model, including TSS index evaluation, and evaluation of inner coincidence accuracy and outer coincidence accuracy,
e) And establishing a local ionosphere scintillation ANN model in the high-latitude polar region, which comprises the following steps:
dividing data from a CHAIN observation station of a foundation in a high latitude area into second model construction data and second model test data after using the characteristic preliminary analysis, using the second model construction data as an input signal of a neural network, training the neural network by means of an error back propagation algorithm, obtaining an output result ROTI through the error back propagation algorithm,
f) And testing the local ionospheric scintillation ANN model in the high-latitude polar region, wherein the testing comprises TSS index evaluation, and evaluation of internal coincidence precision and external coincidence precision.
Preferably, the acquired original signal data is preprocessed in the step a), abnormal data is removed, missing data is supplemented, cycle slip detection is performed, and data smoothing is performed.
Preferably, the first model building data in step c) comprises:
latitude and longitude, local time, solar activity index, geomagnetic index, polar region characteristic index and solar activity cycle index.
Preferably, the step c) neural network training comprises the following method:
the input vector X is:
X＝(X 1 ，X 2 ，...X i ，...，X n ) T ，
the vector Y output by the hidden layer is:
Y＝(Y 1 ，Y 2 ，...，Y i ，..，Y m )，
output layer vector 0 is:
O＝(O 1 ，O 2 ，...，O k ，...，O l )＝{ROTI}，
the desired output vector d is:
d＝(d 1 ，d 2 ，....，d k ，...，d l )，
the weight matrix from the input layer to the hidden layer is represented by V:
V＝(V 1 ，V 2 ，...，V i ，...，V m )，
wherein, the column vector V j The weight vector corresponding to the jth neuron of the hidden layer is as follows:
the weight matrix from the hidden layer to the output layer is represented by W:
W＝(W 1 ，W 2 ，...，W k ，...，W l )，
wherein the column vector W k’ The weight vector corresponding to the kth neuron of the output layer is represented by an input vector by using a subscript i, and the output vector of the hidden layer by using a subscript j represents the output of the jth hidden layer;
for the output layer, there are:
o k ＝f(net k )，k＝1；
for the hidden layer, there are:
y j ＝f(net j )，j＝1，2，3，...，m；
the transfer functions f (x) of the output and hidden layers are both unipolar Sigmoid functions:
f (x) has the characteristic of being continuously conductive and has:
f′(x)＝f(x)[1-f(x)]；
in the error back-propagation algorithm, the source of the error E is the difference between the actual output ROTI and the desired output ROTI:
the error is spread out to the hidden layer,
and further expanding to obtain a cost function or a loss error function:
the weight values w and v are adjusted,
wherein, the weight adjustment amount of the output layer is:
and (3) adjusting the weight of the hidden layer:
wherein, the negative sign represents gradient decline, and eta is the learning coefficient.
Preferably, in step d), the wide-area ionospheric scintillation ANN model is evaluated using the TSS index:
wherein, true positive TP, false positive FP, false negative FN and true negative TN are four evaluation values,
a true positive TP indicates that flicker is occurring and the model predicts correct results; false negative FN indicates that flicker occurred but the model predicts no flicker; false positive FP indicates no flicker but the model predicts flicker; a true negative TN indicates no flicker occurred and the model predicted no flicker.
Preferably, in step d), the evaluation of the accuracy of the external fit comprises,
inputting untrained first model test data into a trained wide-area ionosphere scintillation ANN model, predicting the ROTI value at a certain moment, comparing the ROTI value predicted by the model with the real ROTI, and checking the external conformance accuracy of the model.
Preferably, in step d), the evaluation of the internal coincidence accuracy is: and (4) building data of the trained first model, inputting the data into the trained wide-area ionosphere scintillation ANN model again, and fitting the data.
Preferably, in step f), the local ionospheric scintillation ANN model of the high latitude polar region is evaluated using the TSS index:
wherein, the true positive TP, the false positive FP, the false negative FN and the true negative TN are four evaluation values,
a true positive TP indicates that flicker is occurring and the model predicts correct results; false negative FN indicates that flicker occurred but the model predicts no flicker; false positive FP indicates no flicker but the model predicts flicker; a true negative TN indicates no flicker occurred and the model predicted no flicker.
Preferably, in step f), the evaluation of the accuracy of the external fit comprises,
inputting untrained test data of the second model into a trained local ionosphere scintillation ANN model of the high latitude polar region, predicting the ROTI value at a certain moment, comparing the ROTI value predicted by the model with the real ROTI, and checking the external conformance accuracy of the model.
Preferably, in step f), the evaluation of the internal coincidence accuracy is: and (4) inputting the trained second model construction data into the trained local ionosphere scintillation ANN model of the high latitude polar region again for data fitting.
The invention provides a method for establishing a wide-area and local-area fused high-precision ionospheric scintillation model, which is based on an ionospheric scintillation model under the global scale of space-based COSMIC data and ground-based CHAIN survey station data, simultaneously establishes an ionospheric scintillation model in a high-latitude region by independently using a CHAIN station, extracts the characteristics of factors related to the space-time characteristics of ionospheric scintillation by adopting a deep learning method, establishes the ionospheric scintillation model, improves the accuracy of ionospheric scintillation prediction, and solves the influence of the ionospheric scintillation on signal transmission and satellite navigation positioning precision.
The invention provides a wide-area and local-area fused high-precision ionospheric scintillation model establishing method, which combines the advantages of wide space-based global coverage and sustainable foundation observation, utilizes two kinds of observation data of COSMIC and CHAIN, adopts a deep learning method to extract and analyze the factor characteristics influencing the space-time distribution of ionospheric scintillation, represents the ionospheric scintillation occurrence condition in the form of time change rate index ROTI of the total electron content, constructs a model, gives the ionospheric scintillation distribution condition from the global scale, and independently uses CHAIN station data to establish an ionospheric scintillation model in a high-latitude area, thereby achieving wide-area and local-area fusion, further improving the performance of a receiver and improving the satellite navigation positioning precision.
The invention uses two kinds of observation data of the space foundation and the foundation, overcomes the defects of discontinuous space foundation observation and insufficient global coverage of the foundation, and solves the problem of single data caused by geographical position limitation to a certain extent.
The invention depends on long-term stable and accurate observation data of the ionospheric scintillation index, eliminates data which possibly has problems by using a processing method of locking time, altitude angle and azimuth angle limitation, and improves the accuracy of model establishment.
The invention uses the artificial neural network ANN, carries out feature extraction and training on the influence factors of ionospheric scintillation through an error back propagation algorithm (BP algorithm), and carries out model establishment from a global angle and a high latitude polar region respectively, thereby achieving wide area and local area fusion. The method utilizes the remaining data to evaluate the performance of the established model and verify the accuracy of the model.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention as claimed.
Drawings
Further objects, features and advantages of the present invention will become apparent from the following description of embodiments of the invention, with reference to the accompanying drawings, in which:
FIG. 1 schematically illustrates an ionospheric scintillation model building flow diagram in accordance with the present invention.
Figure 2 shows a schematic diagram of an artificial neural network.
FIG. 3 shows a schematic diagram of an artificial neural network for ionospheric scintillation modeling in accordance with the present invention.
Detailed Description
The objects and functions of the present invention and methods for accomplishing the same will be apparent by reference to the exemplary embodiments. However, the present invention is not limited to the exemplary embodiments disclosed below; it can be implemented in different forms. The nature of the description is merely to assist those skilled in the relevant art in a comprehensive understanding of the specific details of the invention.
Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the drawings, the same reference numerals denote the same or similar components, or the same or similar steps.
In order to solve the problems in the prior art, the ionospheric scintillation index is subjected to data preprocessing, and abnormal values are removed from a large amount of historical observation data; analyzing the correlation of CHAIN dual-frequency ionospheric scintillation data to preliminarily obtain the ionospheric scintillation condition; developing a model by using publicly available solar wind, geomagnetic activity, interplanetary magnetic field and ionosphere GNSS data and forming a machine learning database; training a large amount of influence factor data by using an Artificial Neural Network (ANN) and an error back propagation algorithm (BP algorithm) to extract features; and performing performance evaluation on the established model by using the residual data, and verifying the accuracy of the model.
The following describes the contents of the present invention in detail with reference to specific embodiments, and as shown in fig. 1, the present invention provides a flow chart for establishing an ionospheric scintillation model, and according to the embodiments of the present invention, a method for establishing a wide-area and local-area fused high-precision ionospheric scintillation model includes the following steps:
step a), acquiring original signal data of a space-based COSMIC satellite and a foundation CHAIN observation station in a high latitude area, preprocessing the acquired original signal data, and acquiring ionospheric scintillation signal data.
Source of observation data
Ionospheric scintillation data originated from space-based COSMIC satellites and stations of high latitude region ground foundation CHAIN, the used COSMIC data is 2007-2016, and the CHAIN data is all data observed from 2013 to 2019 by 15 stations (arcc, arvc, chuc, corc, adm, fsic, fsmc, gilc, gjoc, kugc, mcmc, rabc, ranc, repc, sacc) containing a polaxs receiver. The data obtained by the two systems are tested for years, the accuracy is high, and the ionospheric scintillation space-time distribution characteristic can be reflected more truly, so that the ionospheric scintillation model is established by adopting the data.
For ionospheric scintillation, the variety of spatial weather-affecting factors is large. Direct influencing factors such as the number of solar black seeds, the solar radiant flux F10.7cm, dst indexes for describing the annular current of the equator, a parameter Ap for describing the change of the magnetic field of the earth on all days, a parameter Kp for describing the change of the medium and low latitudes, an index AE for evaluating the energy of the magnetic field disturbance of a polar region by a planet magnetic field, and other factors such as particle deposition, irregular volume size and the like can be used in the modeling process.
Raw data pre-processing
According to the embodiment of the invention, the acquired original signal data is preprocessed, abnormal data is removed, missing data is supplemented, cycle slip detection is carried out, and data smoothing is carried out.
When data modeling is carried out, due to the fact that observation data are wide in time span and large in data quantity, in order to improve the accuracy of the established model, corresponding preprocessing needs to be carried out on original observation data. In the original data, COSMIC obtains the amplitude flicker index S 4 . The scintillation index obtained by CHAIN is refined into an amplitude scintillation index, a corrected amplitude scintillation index, a phase scintillation index within 1s, a phase scintillation index within 3s, a phase scintillation index within 10s, a phase scintillation index within 30s and a phase scintillation index within 60 s.
Through analysis, the CHAIN data is subjected to amplitude flicker index selection and phase flicker index selection within 60s as a research object in the implementation. There are many factors causing ionospheric scintillation abnormality, which may be the longitude and latitude setting of the observation station or the activity change of the irregular body, and these abnormal values are atypical and rare, and when they participate in the model training, the result will be deviated, and the model performance will be reduced. Therefore, the initial data preprocessing is needed when the ionospheric scintillation model is constructed, and the steps comprise abnormal data elimination, missing data supplement, cycle slip detection and data smoothing.
And eliminating abnormal data, wherein the effective value of the ionospheric scintillation index is 0-1.4, and abnormal data can be regarded as if the effective value exceeds 1.4. Although the quantity of abnormal data is small, the abnormal data should be removed in preprocessing in order to ensure the accuracy of the model.
The cycle slip detection aims at that under the influence of serious ionospheric flicker, a GPS receiver is unlocked and cycle slips occur, and in the case, the obtained data is unavailable. And performing cycle slip detection, wherein a MW combination method is used, and the method is obtained by subtracting the narrow lane combination of the pseudo-range observation value from the wide lane combination of the phase observation value of the same epoch, is suitable for cycle slip detection of a real-time observation value, and can ensure the data quality and the usability.
And step b), carrying out primary analysis on the space-time characteristics of the acquired ionospheric scintillation signal data.
The obtained original ionospheric scintillation signal data is analyzed from an ionospheric scintillation annual change angle, a ionospheric scintillation month change angle and a ionospheric scintillation place-to-place time change angle in a time angle.
And analyzing angles such as ionospheric scintillation changing with the altitude angle of the satellite and changing with the azimuth angle of the satellite in a space angle.
In the embodiment, the space-time distribution of ionospheric scintillation is described by calculating ionospheric puncture points and drawing a grid diagram, so that a more comprehensive and specific theoretical basis is provided for model establishment.
Step c), establishing a wide area ionospheric scintillation ANN model, comprising:
dividing data subjected to the preliminary analysis of the spatio-temporal characteristics into first model construction data and first model test data, wherein the first model construction data are used as input signals of a neural network, training the neural network by means of an error back propagation algorithm, and obtaining an output result ROTI through the error back propagation algorithm.
According to an embodiment of the present invention, the first model build data comprises 80% of the entire data set and the first model test data comprises 20% of the entire data set.
According to an embodiment of the present invention, the first model construction data includes: latitude and longitude, local time, solar activity index, geomagnetic index, polar characteristic index, and solar activity cycle index.
Specifically, the main factors influencing ionospheric scintillation are the number of solar black seeds, the solar radiation flux f10.7cm (solar activity index), dst index describing the annular current of the equator, parameter Ap (geomagnetic index) describing the change of the earth's whole-day magnetic field and parameter Kp (latitude and longitude) describing the change of the middle and low latitude, and interplanetary magnetic field as index AE (polar region characteristic index) for evaluating the energy of magnetic field disturbance in the polar region.
Other factors that affect ionospheric scintillation are many, such as particle deposition and irregular volume size, and cycle index.
The influence factors of ionospheric scintillation formation and distribution are used to obtain a 'test' sample for model construction. Considering that the raw data is world time UT and the analysis needs to be converted into local time LT, in the embodiment, the input value includes local time LT (local time) in addition to the earth motion and the solar motion cycle (cycle index).
After ionosphere scintillation space-time characteristic analysis, modeling is carried out on the global scale by combining space-based data and foundation data (COSMIC data and CHAIN data). Before model building, data is standardized, and the purpose is to normalize the data into data in a [0,1] interval, so that the training time is short, and the iterative convergence is fast. In some embodiments, the method used for normalization is min-max normalization.
As shown in fig. 2, the network result of the neural network is from the input layer to the hidden layer to the output layer, and the neurons between the layers are completely connected, but the neurons in each layer are not connected. The input layer receives the signal "x" from the outside world. The output layer produces a result "Y" that is processed by the neural network system. The hidden layer between the input layer and the output layer is not visible outside the system. The weight of the connection between neurons reflects the strength of the connection between neurons.
During the training process, the output value Y gradually approaches the target value by means of an error back propagation algorithm (BP algorithm). The basic idea of the error back propagation algorithm is that the learning process consists of two processes, namely the forward propagation of signals and the back propagation of errors. In forward propagation, input samples are transmitted from the input layer, processed layer by each hidden layer, and transmitted to the output layer. If the actual output of the output layer does not match the expected output, the error is propagated back to the error propagation stage. The error back propagation is to reversely transmit the output error to the input layer by layer through the hidden layer in a certain form, and distribute the error to all units of each layer, thereby obtaining an error signal of each layer, wherein the error signal is used as a basis for correcting the weight of the unit. The weight value adjustment process of each layer of signal forward propagation and error backward propagation is carried out repeatedly, the process of continuously adjusting the weight value, namely the process of network learning training, is carried out until the error of network output is reduced to an acceptable degree or is carried out to preset learning times.
Wide-area ionospheric scintillation ANN model establishment
As shown in fig. 3, the ionospheric scintillation modeling artificial neural network of the present invention is schematically illustrated, and according to an embodiment of the present invention, the neural network training includes the following methods:
the input vector X is:
X＝(X 1 ，X 2 ，...X i ，...，X n ) T ，
the vector Y output by the hidden layer is:
Y＝(Y 1 ，Y 2 ，...，Y j ，...，Y m )，
output layer vector 0 is:
0＝(O 1 ，O 2 ，...，O k ，...，O l )＝{ROTI}，
the desired output vector d is:
d＝(d 1 ，d 2 ，...，d k ，...，d l )，
the weight matrix from the input layer to the hidden layer is represented by V:
V＝(V 1 ，V 2 ，...，V i ，...，V m )，
wherein the column vector V j A weight vector corresponding to the jth neuron of the hidden layer:
the weight matrix from the hidden layer to the output layer is represented by W:
W＝(W 1 ，W 2 ，...，W k ，...，W l )，
wherein the column vector W k’ The weight vector corresponding to the kth neuron of the output layer is represented by an input vector by using a subscript i, and the output vector of the hidden layer by using a subscript j represents the output of the jth hidden layer;
for the output layer, there are:
O k ＝f(net k )，k＝1；
for the hidden layer, there are:
y j ＝f(net j )，j＝1，2，3，...，m；
the transfer functions f (x) of the output layer and the hidden layer are both unipolar Sigmoid functions:
f (x) is characterized by continuous conductivity and has:
f′(x)＝f(x)[1-f(x)]；
in the error back-propagation algorithm, the source of the error E is the difference between the actual output ROTI and the desired output ROTI:
the error is spread out to the hidden layer,
and further expanding to obtain a cost function or a loss error function:
a cost function or a loss error function, in which weights w and v are both present, E can be minimized as long as w and v take appropriate values. And the weight value can be updated by using a gradient descent method to achieve the optimal value, and because the gradient of a certain point is the first-order partial derivative of the gradient, the partial derivatives of w and v can be obtained.
The weight values w and v are adjusted,
wherein, the weight adjustment amount of the output layer is:
and (3) adjusting the weight of the hidden layer:
wherein, the negative sign represents gradient decline, and eta is the learning coefficient.
The obtained output result, ROTI (time rate of change index of total electron content), is expected after the error back propagation algorithm.
And d), testing the wide-area ionospheric scintillation ANN model, wherein the wide-area ionospheric scintillation ANN model comprises TSS index evaluation and evaluation of inner coincidence precision and outer coincidence precision.
The evaluation of the model can adopt the total skill score TSS index, the accuracy, the root mean square error of the predicted value and the true value in each data set, and the evaluation of the inner coincidence precision and the outer coincidence precision of the model.
According to an embodiment of the invention, a wide-area ionospheric scintillation ANN model is evaluated using a TSS index:
wherein, the true positive TP, the false positive FP, the false negative FN and the true negative TN are four evaluation values,
a true positive TP indicates that flicker is occurring and the model predicts correct results; false negative FN indicates that flicker occurred but the model predicts no flicker; false positive FP indicates no flicker but the model predicts flicker; a true negative TN indicates no flicker occurred and the model predicted no flicker.
False negative FN reflects the false negative rate of the model, and the larger FN means that the false negative rate of the model is larger, so that the false negative rate of the model is more, and the false negative rate of the model is serious for the regions with frequent ionospheric scintillation events, such as the low latitude equatorial region and the high latitude polar region.
The false positive FP reflects the false report rate of the model, and the false report has a certain influence on the performance of the model but the brought consequences are not as serious as the false report rate. The smaller the FN value in the performance evaluation of the model constructed by the present invention, the better.
The TSS index can well reflect the number of false negative FNs and can better evaluate the overall performance of the model, the larger the TSS index is, the larger the true positive TP is, the smaller the false negative FN is, the smaller the false positive FP is, the larger the true negative TN is, and the ionospheric scintillation prediction is more accurate.
Conversely, the smaller the TSS index is, the smaller the true positive TP is, the larger the false negative FN is, the larger the false positive FP is, and the smaller the true negative TN is, which indicates that the ionospheric scintillation prediction is not accurate enough.
Therefore, the TSS index can reflect the comprehensive performance of the wide-area ionospheric scintillation ANN model provided by the invention. The value range of the TSS index is between [0 and 1], the closer the TSS value is to 0, the poorer the prediction effect is, and the closer the TSS value is to 1, the better the prediction effect is.
According to an embodiment of the present invention, the evaluation of the accuracy of the outer fit comprises,
inputting untrained first model test data into a trained wide-area ionosphere scintillation ANN model, predicting the ROTI value at a certain moment, comparing the ROTI value predicted by the model with the real ROTI, and checking the external conformance accuracy of the model.
According to an embodiment of the present invention, the evaluation of the intra-coincidence accuracy is: and (4) building data of the trained first model, inputting the data into the trained wide-area ionosphere scintillation ANN model again, and fitting the data.
According to the embodiment of the invention, the accuracy is judged by the internal coincidence accuracy, the external coincidence accuracy and the root mean square error of the predicted value and the true value in each data set, and the accuracy of the model is higher under the condition that the internal coincidence accuracy is smaller, the external coincidence accuracy is smaller and the root mean square error of the predicted value and the true value is smaller.
Step e), establishing a local ionosphere scintillation ANN model in the high latitude polar region, comprising the following steps:
dividing data from a foundation CHAIN observation station in a high latitude area into second model construction data and second model test data after using the characteristic preliminary analysis, wherein the second model construction data is used as an input signal of a neural network, training the neural network by means of an error back propagation algorithm, and obtaining an output result ROTI through the error back propagation algorithm.
According to an embodiment of the present invention, the second model build data comprises 80% of the entire data set and the second model test data comprises 20% of the entire data set.
The high latitude polar region local ionospheric scintillation ANN model uses data from the high latitude region ground based CHAIN observatory as input data. And (4) training a neural network by an error back propagation algorithm, wherein the process of establishing the local ionospheric scintillation ANN model in the high latitude polar region is the same as the process of the step c), and the details are not repeated here.
Step f), testing a local ionosphere scintillation ANN model in the high latitude polar region, wherein the testing comprises TSS index evaluation and evaluation of internal coincidence precision and external coincidence precision.
The evaluation of the model can adopt the total skill score TSS index, the accuracy, the root mean square error of the predicted value and the true value in each data set, and the evaluation of the internal coincidence precision and the external coincidence precision of the model.
According to the embodiment of the invention, a local ionospheric scintillation ANN model in a high latitude polar region is evaluated by using a TSS index:
wherein, true positive TP, false positive FP, false negative FN and true negative TN are four evaluation values,
a true positive TP indicates that flicker is occurring and the model predicts correct results; false negative FN indicates that flicker occurred but the model predicted no flicker; false positive FP indicates no flicker but the model predicts flicker; a true negative TN indicates no flicker occurred and the model predicted no flicker.
According to an embodiment of the present invention, the evaluation of the accuracy of the outer fit comprises,
inputting untrained second model test data into a trained local ionosphere scintillation ANN model of the high latitude polar region, predicting the ROTI value at a certain moment, comparing the ROTI value predicted by the model with the real ROTI, and checking the external conformance accuracy of the model.
According to an embodiment of the present invention, the evaluation of the intra-coincidence accuracy is: and (4) inputting the trained second model construction data into the trained local ionosphere scintillation ANN model of the high latitude polar region again for data fitting.
According to the embodiment of the invention, the accuracy is judged by the internal coincidence accuracy, the external coincidence accuracy and the root mean square error of the predicted value and the true value in each data set, and the accuracy of the model is higher under the condition that the internal coincidence accuracy is smaller, the external coincidence accuracy is smaller and the root mean square error of the predicted value and the true value is smaller.
The invention provides a method for establishing a wide-area and local-area fused high-precision ionospheric scintillation model, which is based on an ionospheric scintillation model under the global scale of space-based COSMIC data and ground-based CHAIN survey station data, simultaneously establishes an ionospheric scintillation model in a high-latitude region by independently using a CHAIN station, extracts the characteristics of factors related to the space-time characteristics of ionospheric scintillation by adopting a deep learning method, establishes the ionospheric scintillation model, improves the accuracy of ionospheric scintillation prediction, and solves the influence of the ionospheric scintillation on signal transmission and satellite navigation positioning precision.
The invention provides a wide-area and local-area fused high-precision ionospheric scintillation model establishing method, which combines the advantages of wide space-based global coverage and sustainable foundation observation, utilizes two kinds of observation data of COSMIC and CHAIN, adopts a deep learning method to extract and analyze the factor characteristics influencing the space-time distribution of ionospheric scintillation, represents the ionospheric scintillation occurrence condition in the form of time change rate index ROTI of the total electron content, constructs a model, gives the ionospheric scintillation distribution condition from the global scale, and independently uses CHAIN station data to establish an ionospheric scintillation model in a high-latitude area, thereby achieving wide-area and local-area fusion, further improving the performance of a receiver and improving the satellite navigation positioning precision.
The invention uses two kinds of observation data of the space foundation and the foundation, overcomes the defects of discontinuous space foundation observation and insufficient global coverage of the foundation, and solves the problem of single data caused by geographical position limitation to a certain extent.
The method relies on long-term stable and accurate observation data of the ionospheric scintillation index, eliminates data which possibly have problems by using a processing method of locking time, altitude angle and azimuth angle limitation, and improves the accuracy of model establishment.
The invention uses the artificial neural network ANN, carries out feature extraction and training on the influence factors of ionospheric scintillation through an error back propagation algorithm (BP algorithm), and carries out model establishment from a global angle and a high latitude polar region respectively, thereby achieving wide area and local area fusion. The method utilizes the remaining data to evaluate the performance of the established model and verify the accuracy of the model.
Other embodiments of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims.
Claims (10)
1. A method for establishing a wide-area and local-area fused high-precision ionospheric scintillation model is characterized by comprising the following steps:
a) Acquiring original signal data of a space-based COSMIC satellite and a high latitude region foundation CHAIN observation station, preprocessing the acquired original signal data, and acquiring ionospheric scintillation signal data;
eliminating abnormal data, wherein the effective value of the ionospheric scintillation index is 0-1.4, and abnormal data can be regarded as if the effective value exceeds 1.4;
the COSMIC obtains original data which are amplitude flicker indexes S4, and the CHAIN data select the amplitude flicker indexes and phase flicker indexes within 60S;
b) Carrying out time-space characteristic preliminary analysis on the acquired ionospheric scintillation signal data;
performing primary analysis on the space-time characteristics of the acquired ionospheric scintillation signal data; analyzing the obtained original ionospheric scintillation signal data from ionospheric scintillation year-to-year change, month-to-month change and place-to-place change angles in a time angle; analyzing angles such as ionospheric scintillation, satellite altitude angle change and satellite azimuth angle change in a space angle; describing the time-space distribution of ionospheric scintillation by calculating ionospheric puncture points and drawing a grid diagram;
c) And establishing a wide-area ionospheric scintillation ANN model, which comprises the following steps:
dividing data subjected to the preliminary analysis of the spatio-temporal characteristics into first model construction data and first model test data, wherein the first model construction data are used as input signals of a neural network, performing neural network training by means of an error back propagation algorithm, and obtaining an output result ROTI through the error back propagation algorithm;
d) Testing a wide-area ionospheric scintillation ANN model, including TSS index evaluation, and evaluation of inner coincidence accuracy and outer coincidence accuracy,
e) And establishing a local ionosphere scintillation ANN model in the high-latitude polar region, which comprises the following steps:
dividing data from a CHAIN observation station of a foundation in a high latitude area into second model construction data and second model test data after using the characteristic preliminary analysis, using the second model construction data as an input signal of a neural network, training the neural network by means of an error back propagation algorithm, obtaining an output result ROTI through the error back propagation algorithm,
f) And testing the local ionospheric scintillation ANN model in the high-latitude polar region, wherein the testing comprises TSS index evaluation, and evaluation of internal coincidence precision and external coincidence precision.
2. The method according to claim 1, wherein the acquired original signal data is preprocessed in step a), abnormal data is removed, missing data is supplemented, cycle slip detection is performed, and data smoothing is performed.
3. The method of claim 1, wherein the first model building data in step c) comprises:
latitude and longitude, local time, solar activity index, geomagnetic index, polar characteristic index, and solar activity cycle index.
4. The method of claim 1, wherein the neural network training of steps c) and e) comprises the following method:
the input vector X is:
X＝(X 1 ，X 2 ，...，X i ，...，X n ) T ，
the vector Y output by the hidden layer is:
Y＝(Y 1 ，Y 2 ，...，Y i ，...，Y m )，
the output layer vector O is:
O＝(O 1 ，O 2 ，...，O k ，...，O 1 )＝{ROTI}，
the desired output vector d is:
d＝(d 1 ，d 2 ，...，d k ，...，d 1 )，
the weight matrix from the input layer to the hidden layer is represented by V:
V＝(V 1 ，V 2 ，...，V，...，V m )，
wherein the column vector V j For the jth of the hidden layerWeight vector corresponding to neuron:
the weight matrix from the hidden layer to the output layer is represented by W:
W＝(W 1 ，W 2 ，...，W k ，...，W l )，
wherein the column vector W k， The weight vector corresponding to the kth neuron of the output layer is represented by an input vector by using a subscript i, and the output vector of the hidden layer by using a subscript j represents the output of the jth hidden layer;
for the output layer, there are:
o k ＝f(netk ) ，k＝1；
for the hidden layer, there are:
y j ＝f(net j )，j＝1，2，3，...，m；
the transfer functions f (x) of the output layer and the hidden layer are both unipolar Sigmoid functions:
f (x) has the characteristic of being continuously conductive and has:
f′(x)＝f(x)[1-f(x)]；
in the error back-propagation algorithm, the source of the error E is the difference between the actual output ROTI and the desired output ROTI:
the error is spread out to a hidden layer,
and further expanding to obtain a cost function or a loss error function:
the weight values w and v are adjusted,
wherein, the weight adjustment amount of the output layer is as follows:
and (3) adjusting the weight of the hidden layer:
wherein, the negative sign represents gradient decline, and eta is the learning coefficient.
5. The method of claim 1, wherein in step d), the wide-area ionospheric scintillation ANN model is evaluated using TSS indices:
wherein, true positive TP, false positive FP, false negative FN and true negative TN are four evaluation values,
a true positive TP indicates that flicker is occurring and the model predicts correct results; false negative FN indicates that flicker occurred but the model predicted no flicker; false positive FP indicates no flicker but the model predicts flicker; a true negative TN indicates no flicker occurred and the model predicted no flicker.
6. The method according to claim 1, wherein in step d), the evaluation of the accuracy of the external fit comprises,
inputting untrained first model test data into a trained wide-area ionosphere scintillation ANN model, predicting the ROTI value at a certain moment, comparing the ROTI value predicted by the model with the real ROTI, and checking the external conformance accuracy of the model.
7. The method according to claim 1, wherein in step d), the evaluation of the internal coincidence accuracy is: and (4) inputting the trained first model construction data into the trained wide-area ionosphere scintillation ANN model again for data fitting.
8. The method according to claim 1, wherein in step f), the local ionospheric scintillation ANN model of the high latitude polar region is evaluated using TSS index:
wherein, true positive TP, false positive FP, false negative FN and true negative TN are four evaluation values,
a true positive TP indicates that flicker is occurring and the model predicts correct results; false negative FN indicates that flicker occurred but the model predicts no flicker; false positive FP indicates no flicker but the model predicts flicker; a true negative TN indicates no flicker occurred and the model predicted no flicker.
9. The method according to claim 1, wherein in step f), the evaluation of the accuracy of the external fit comprises,
inputting untrained second model test data into a trained local ionosphere scintillation ANN model of the high latitude polar region, predicting the ROTI value at a certain moment, comparing the ROTI value predicted by the model with the real ROTI, and checking the external conformance accuracy of the model.
10. The method according to claim 1, wherein in step f), the evaluation of the internal fit accuracy is: and (4) inputting the trained second model construction data into the trained local ionosphere scintillation ANN model of the high latitude polar region again for data fitting.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110189833.7A CN113093225B (en) | 2021-02-18 | 2021-02-18 | Wide-area and local-area fused high-precision ionospheric scintillation model establishment method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110189833.7A CN113093225B (en) | 2021-02-18 | 2021-02-18 | Wide-area and local-area fused high-precision ionospheric scintillation model establishment method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113093225A CN113093225A (en) | 2021-07-09 |
| CN113093225B true CN113093225B (en) | 2023-04-18 |
Family
ID=76663888
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110189833.7A Active CN113093225B (en) | 2021-02-18 | 2021-02-18 | Wide-area and local-area fused high-precision ionospheric scintillation model establishment method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113093225B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113569433B (en) * | 2021-09-24 | 2022-03-25 | 北京航空航天大学 | Deep space electromagnetic wave channel propagation modeling method under solar scintillation |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| SG11201701730WA (en) * | 2014-09-05 | 2017-04-27 | Fugro Intersite B V | Ionospheric scintillation prediction |
| CN107991690A (en) * | 2017-06-28 | 2018-05-04 | 江苏省测绘工程院 | A kind of computational methods of the scintillation index based on Beidou navigation |
| CN111581803B (en) * | 2020-04-30 | 2022-04-26 | 北京航空航天大学 | Method for constructing Krigin proxy model of global ionized layer electron content |
| CN111651941B (en) * | 2020-04-30 | 2022-05-17 | 北京航空航天大学 | Global ionosphere electron total content prediction algorithm |
| CN112070203A (en) * | 2020-07-20 | 2020-12-11 | 中国科学院空天信息创新研究院 | Deep learning-based northern hemisphere high-latitude region ROTI prediction method |
| CN111983654B (en) * | 2020-08-24 | 2021-03-19 | 中国矿业大学 | Method for constructing ionosphere phase scintillation factor in arctic region based on GNSS |
-
2021
- 2021-02-18 CN CN202110189833.7A patent/CN113093225B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113093225A (en) | 2021-07-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10962651B2 (en) | Ionospheric delay correction method for LEO satellite augmented navigation systems | |
| Wang et al. | Development of the global assimilative ionospheric model | |
| Huang et al. | Ionospheric single-station TEC short-term forecast using RBF neural network | |
| CN111639747B (en) | GNSS-R sea surface wind speed inversion method and system based on BP neural network | |
| Tebabal et al. | Local TEC modelling and forecasting using neural networks | |
| CN113936142A (en) | Rainfall approach forecasting method and device based on deep learning | |
| CN110990505B (en) | Loran-C ASF correction method based on neural network | |
| CN114265836A (en) | All-weather assimilation method and device of satellite microwave hygrothermograph based on cloud area temperature and humidity profile inversion | |
| Guo et al. | GNSS-R wind speed retrieval of sea surface based on particle swarm optimization algorithm | |
| CN114417580A (en) | Method for evaluating influence of observation system on assimilation performance of global ionosphere data | |
| CN113093225B (en) | Wide-area and local-area fused high-precision ionospheric scintillation model establishment method | |
| Xu et al. | An investigation of optimal machine learning methods for the prediction of ROTI | |
| Xu et al. | A cloud-dependent 1DVAR precipitation retrieval algorithm for FengYun-3D microwave soundings: a case study in tropical cyclone mekkhala | |
| Yu et al. | Using the GPS observations to reconstruct the ionosphere three-dimensionally with an ionospheric data assimilation and analysis system (IDAAS) | |
| Hirooka et al. | Neural network based tomographic approach to detect earthquake-related ionospheric anomalies | |
| Jiang et al. | Comparison of the Kriging and neural network methods for modeling foF2 maps over North China region | |
| Aranzulla et al. | Volcanic ash detection by GPS signal | |
| Stamper et al. | Nowcasting, forecasting and warning for ionospheric propagation: tools and methods | |
| Santosa et al. | One day prediction of nighttime VLF amplitudes using nonlinear autoregression and neural network modeling | |
| Guoyan et al. | Prediction of ionospheric TEC based on the NARX neural network | |
| Priyadarshi et al. | Machine Learning-based ionospheric modelling performance during high ionospheric activity | |
| Pietrella et al. | A comparative study of ionospheric IRIEup and ISP assimilative models during some intense and severe geomagnetic storms | |
| Tafazoli et al. | Performance improvement of GPS GDOP approximation using recurrent wavelet neural network | |
| Forootan et al. | Empirical data assimilation for merging total electron content data with empirical and physical models | |
| CN113625356A (en) | Real-time anomaly monitoring method suitable for single-station ionized layer TEC |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |