High-latitude ionosphere forecasting method based on multi-source heterogeneous data

Abstract

The invention belongs to the technical field of high-latitude ionosphere, and particularly relates to a high-latitude ionosphere forecasting method based on multi-source heterogeneous data. The method comprises the following steps: S1, establishing a time sequence data set of foundation ionosphere measurement data and a space weather index set; s2, establishing a time sequence data set of space-based high-energy particle sedimentation flux measurement data; s3, normalizing all the time sequence data sets, and establishing an input data set based on multi-source heterogeneous data; and S4, establishing an LSTM prediction neural network, and obtaining a prediction result of the high-latitude ionosphere parameters based on the input data set. According to the method, on the basis of using ground-based ionosphere measurement data and a space weather index, space-based high-energy particle sedimentation flux measurement data is introduced to obtain a normalized time sequence data set, so that an input data set based on multi-source heterogeneous data is established, and then an LSTM prediction neural network is used to obtain a prediction result of high-latitude ionosphere parameters.

Description

Technical Field

[0001] The present invention belongs to the technical field of high-latitude ionosphere, and in particular relates to a high-latitude ionosphere prediction method based on multi-source heterogeneous data. Background Art

[0002] Ionospheric parameter forecasting is crucial for the operation and maintenance of communication and navigation systems in high-latitude regions. This region is significantly affected by geomagnetic activity and polar particle precipitation. Especially during magnetic storms, high-energy particles penetrate the atmosphere, causing a sharp increase in ionospheric electron density, leading to problems such as attenuation of shortwave communications and a surge in GNSS positioning errors. Existing ionospheric parameter prediction models mostly rely on the solar radiation index (F10.7) and geomagnetic indices (Kp, Dst) as inputs, but fail to fully incorporate direct observations of high-energy particle precipitation flux, resulting in insufficient prediction capabilities for sudden ionospheric disturbances.

[0003] Current mainstream methods fall into two categories: the International Reference Ionospheric Model (IRI), based on empirical formulas, predicts electron density distribution through historical statistical relationships, but suffers from a significant lag in its dynamic response to particle precipitation events. Machine learning models based on time series algorithms such as LSTM can capture temporal changes in parameters, but because their training data is limited to geomagnetic indices and historical ionospheric values, they fail to account for the physical causal relationship between particle precipitation and ionospheric disturbances. For example, the real-time ionospheric correction model (STORM-E) used by NOAA, during the November 2022 strong magnetic storm, had a 52% deviation in its TEC peak prediction when precipitation flux data was ignored.

[0004] The limitations of existing technologies are mainly reflected in the following aspects: (1) the data of high-energy particle precipitation flux, as the direct driving factor of the sudden change of ionospheric electron density, has not been included in the input feature set of mainstream forecast models; (2) there is a mismatch between the spatiotemporal resolution of satellite-measured particle precipitation data (such as POES and Swarm satellite detection results) and ground-based ionospheric observations, which restricts the effectiveness of multi-source data fusion.

[0005] Therefore, the above-mentioned problems need to be solved urgently. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the present invention provides a high-latitude ionospheric forecasting method based on multi-source heterogeneous data. This method uses ground-based ionospheric measurement data and space weather indices, introduces space-based high-energy particle fallout flux measurement data, and obtains a normalized time-series dataset. This data is then used to establish an input dataset based on multi-source heterogeneous data. The method then uses an LSTM predictive neural network to obtain forecast results for high-latitude ionospheric parameters.

[0007] To solve the above technical problems, the technical solution adopted by the present invention is: a high-latitude ionosphere prediction method based on multi-source heterogeneous data, comprising the following steps:

[0008] S1, establish a time series dataset of ground-based ionospheric measurement data and space weather index set;

[0009] S10, ground-based ionospheric measurement data, refers to the TEC data measured by ground receivers, that is, the total number of free electrons in a unit cross-sectional area column perpendicular to the propagation path;

[0010] S11, space weather index set, including Dst index, IMF-Bz index, and F10.7 index;

[0011] TEC: total electron content, Dst index: geomagnetic storm ring current index, IMF-Bz index: interplanetary magnetic field north-south component index, F10.7 index: 10.7 cm solar radio flux index;

[0012] S2, establish a time series dataset of space-based high-energy particle fallout flux measurement data;

[0013] S20: The satellite is in constant motion and cannot make continuous measurements of a single point. However, energetic particle precipitation mainly occurs in the auroral belt area above 60° magnetic latitude. All satellite measurements of the energetic particle precipitation flux in this area can be recorded.

[0014] S3 normalizes all time series data sets and establishes an input data set based on multi-source heterogeneous data;

[0015] S4, establishes an LSTM prediction neural network and obtains the forecast results of high-latitude ionospheric parameters based on the input data set.

[0016] Furthermore, the step S10 includes:

[0017] The ionospheric TEC data measured by the ground-based GNSS receiver in high latitudes is used as the ground-based measurement data. According to a certain time resolution, a time series data set of TEC data is established. TEC express:

[0018] S TEC ={TEC(t i )|t i =t0+(i-1)Δt,i=1,2,…,N i} Formula 1;

[0019] Among them, TEC(t i ) represents t i The value of the ionospheric TEC data at the moment, t is the time, t0 is the starting time, the subscript i is the time sequence number, Ni is the total number of data, Δt is the time resolution;

[0020] The S11 step includes:

[0021] According to a certain time resolution, the time series data sets of Dst index, IMF-Bz index and F10.7 index are established, respectively using S Dst 、S Bz and S F107 express:

[0022]

[0023] Establish a time series dataset of space weather index set {S Dst ,S Bz ,S F107};

[0024] Among them, Dst(t i )、Bz(t i )、F107(t i ) represent t i The values ​​of the Dst index, IMF-Bz index, and F10.7 index at the moment, t is the time, t0 is the starting time, the subscript i is the time sequence number, N i is the total number of data, and Δt is the time resolution.

[0025] Furthermore, the step S20 includes:

[0026] All the measured values ​​in S1 are averaged with time resolution to obtain a time series data set of high energy particle precipitation flux measurement data. P express:

[0027]

[0028] Among them, P k represents the measured value of high-energy particle precipitation flux, k is the measurement value number, φ is the magnetic latitude value corresponding to the high-energy particle precipitation flux, P(t i ) represents t i The value of the high-energy particle precipitation flux at time t is time, t0 is the starting time, subscript i is the time sequence number, N i is the total number of data, Δt is the time resolution, 1 条件 It is an indicator function, which takes 1 when the condition is met and 0 otherwise.

[0029] Furthermore, the S3 step includes:

[0030] S31, introduce local time series data, and use L(t i ) indicates that it is also normalized to obtain the local time sine value normalized time series data set and local time cosine value normalized time series dataset

[0031]

[0032] S32: Normalize the time series data sets of ground-based ionospheric measurement data, space weather index set, and space-based high-energy particle fallout flux measurement data; obtain the normalized time series data set of TEC data Normalized time series dataset of Dst index Normalized time series dataset of the IMF-Bz index Normalized time series dataset with F10.7 index and a normalized time series dataset of high-energy particle precipitation flux measurements

[0033] S33, take 5 consecutive time series data from the normalized time series data set, and build the neural network input data samples, respectively using TEC5(t i )、Dst5(t i )、Bz5(t i )、F1075(t i )、P5(t i )、Ls5(t i ) and Lc5(t i ) indicates that the input data set is represented by X TEC 、X Dst 、X Bz 、X F107 、X P 、X Ls and X Lc :

[0034]

[0035] Formula 6.

[0036] Furthermore, the S4 step includes:

[0037] S41, establish an LSTM prediction neural network, the input neuron uses the input data set, the number of output layer neurons is 1, expressed as Y TEC , that is, the ionospheric TEC value at the time corresponding to the forecast of Δt in advance;

[0038] S42, based on the trained prediction neural network, inputting the input samples 5 moments before the time to be predicted, and obtaining the high-latitude ionosphere prediction result;

[0039] S43, uses two parameters to evaluate the forecast error, including RMSE and PD:

[0040] RMSE is the root mean square error, calculated as:

[0041]

[0042] PD is the overall relative error, calculated as:

[0043]

[0044] Among them, j is the serial number of the sample used for error statistics, N j The total number of samples used for the error statistics.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] This paper presents a high-latitude ionospheric forecasting method based on multi-source heterogeneous data. By combining high-energy particle fallout flux data with a traditional LSTM model, the method improves the accuracy of existing ionospheric forecasting methods. High-energy particle fallout flux data from POES / MetOp satellites is incorporated into the LSTM model, along with historical ionospheric parameters and geomagnetic indices, to form multi-dimensional input features. Field measurements have shown that this method reduces the root mean square error (RMS) of the forecast results from 4.25 to 2.59, and the overall relative error from 0.45 to 0.17. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 Schematic diagram of a high-latitude ionosphere prediction method based on multi-source heterogeneous data according to the present invention;

[0049] Figure 2 Schematic diagram of the changes and correlations of mep0e1, AE index, and TEC data in Example 1 of the present invention;

[0050] Figure 3 This is a schematic diagram of the statistical results of the correlation between the TEC predicted value and the actual value obtained in Example 1 of the present invention;

[0051] Figure 4 This is a schematic diagram of the statistical results of the correlation between the TEC predicted value and the actual value obtained by the present invention using the existing method. DETAILED DESCRIPTION

[0052] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0053] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention but is merely representative of selected embodiments of the present invention.

[0054] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field under the description before making any creative work shall fall within the scope of protection of the present invention.

[0055] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0056] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0057] like Figure 1 As shown:

[0058] A high-latitude ionosphere prediction method based on multi-source heterogeneous data comprises the following steps:

[0059] S1, establish a time series dataset of ground-based ionospheric measurement data and space weather index set;

[0060] TEC: total electron content, Dst index: geomagnetic storm ring current index, IMF-Bz index: interplanetary magnetic field north-south component index, F10.7 index: 10.7 cm solar radio flux index;

[0061] S10, ground-based ionospheric measurement data, refers to the TEC data measured by ground receivers, that is, the total number of free electrons in a unit cross-sectional area column perpendicular to the propagation path (usually refers to the path from the satellite to the receiver with a cross section of 1 square meter);

[0062] The step S10 includes:

[0063] The ionospheric TEC data measured by the ground-based GNSS receiver in high latitudes is used as the ground-based measurement data. According to a certain time resolution, a time series data set of TEC data is established. TEC express:

[0064] S TEC ={TEC(t i )|t i =t0+(i-1)Δt,i=1,2,…,N i} Formula 1;

[0065] Among them, TEC(t i ) represents t i The value of the ionospheric TEC data at the moment, t is the time, t0 is the starting time, the subscript i is the time sequence number, N i is the total number of data, Δt is the time resolution;

[0066] S11, space weather index set, including Dst index, IMF-Bz index, and F10.7 index;

[0067] The S11 step includes:

[0068] According to a certain time resolution, the time series data sets of Dst index, IMF-Bz index and F10.7 index are established, respectively using S Dst 、S Bz and S F107 express:

[0069]

[0070] Establish a time series dataset of space weather index set {S Dst ,S Bz ,S F107};

[0071] Among them, Dst(t i )、Bz(t i )、F107(t i ) represent t i The values ​​of the Dst index, IMF-Bz index, and F10.7 index at the moment, t is the time, t0 is the starting time, the subscript i is the time sequence number, N i is the total number of data, Δt is the time resolution;

[0072] Among them, the Dst index can be downloaded from websites such as the World Geomagnetic Data Center, the IMF-Bz index can be downloaded from websites such as OMNIWeb, and the F10.7 index can be downloaded from websites such as the Canadian Space Weather Agency.

[0073] S2, establish a time series dataset of space-based high-energy particle fallout flux measurement data;

[0074] S20: The satellite is in constant motion and cannot make continuous measurements of a single point. However, energetic particle precipitation mainly occurs in the auroral belt area above 60° magnetic latitude. All satellite measurements of the energetic particle precipitation flux in this area can be recorded.

[0075] The S20 step includes:

[0076] All the measured values ​​in S1 are averaged with time resolution to obtain a time series data set of high energy particle precipitation flux measurement data. P express:

[0077]

[0078]

[0079] Among them, P k represents the measured value of high-energy particle precipitation flux, k is the measurement value number, φ is the magnetic latitude value corresponding to the high-energy particle precipitation flux, P(t i ) represents t i The value of the high-energy particle precipitation flux at time t is time, t0 is the starting time, subscript i is the time sequence number, N i is the total number of data, Δt is the time resolution, 1 条件 It is an indicator function, which takes 1 when the condition is met and 0 otherwise.

[0080] S3 normalizes all time series data sets and establishes an input data set based on multi-source heterogeneous data;

[0081] The S3 step includes:

[0082] S31, introduce local time series data, and use L(t i ) indicates that it is also normalized to obtain the local time sine value normalized time series data set and local time cosine value normalized time series dataset

[0083]

[0084] S32: Normalize the time series data sets of ground-based ionospheric measurement data, space weather index set, and space-based high-energy particle fallout flux measurement data; obtain the normalized time series data set of TEC data Normalized time series dataset of Dst index Normalized time series dataset of the IMF-Bz index Normalized time series dataset with F10.7 index and a normalized time series dataset of high-energy particle precipitation flux measurements

[0085] S33, slide and take 5 consecutive time series data from the normalized time series data set to establish the neural network input data samples, respectively using TEC5(t i )、Dst5(t i )、Bz5(t i )、F1075(t i )、P5(t i )、Ls5(t i ) and Lc5(t i ) indicates that the input data set is represented by X TEC 、X Dst 、X Bz 、X F107 、X P 、X Ls and X Lc :

[0086]

[0087] S4, establishes an LSTM prediction neural network and obtains the forecast results of high-latitude ionospheric parameters based on the input data set;

[0088] The S4 step includes:

[0089] S41, establish an LSTM prediction neural network, the input neuron uses the input data set, the number of output layer neurons is 1, expressed as Y TEC , that is, the ionospheric TEC value at the time corresponding to the forecast of Δt in advance;

[0090] S42, based on the trained prediction neural network, inputting the input samples 5 moments before the time to be predicted, and obtaining the high-latitude ionosphere prediction result;

[0091] S43, uses two parameters to evaluate the forecast error, including RMSE and PD:

[0092] RMSE is the root mean square error, calculated as:

[0093]

[0094] PD (Percentage Deviation) is the overall relative error, calculated as follows:

[0095]

[0096] Among them, j is the serial number of the sample used for error statistics, N j The total number of samples used for the error statistics.

[0097] Example 1:

[0098] The forecast method of the present invention uses ionospheric TEC data (2014) measured by a ground-based GNSS receiver at the EURC (Eureka Weather Station) in Canada, at a geomagnetic latitude of 88.5°N. The Dst index, IMF-Bz index, and F10.7 index data for the corresponding time period are obtained. High-energy particle fallout flux measurements are also obtained, using the mep0e1 data, which represents the fallout electron flux in the >30 keV channel of the MEPED instrument on the POES / MetOp satellite.

[0099] Figure 2 Figure 2 shows the changes in mep0e1, the AE index, and TEC data for January 4, 2014. It shows a moderate correlation between the energetic particle fallout flux measurement data (mep0e1) and TEC. Furthermore, the correlation coefficient between mep0e1 and TEC for 2014 was calculated to be 0.45, indicating a moderate correlation between mep0e1 and TEC.

[0100] Figure 3 The statistical results of the correlation between the TEC forecast value and the actual value obtained by the forecast method of the present invention are shown in Table 1. Figure 4 The statistical results of the correlation between the TEC predicted values ​​and actual values ​​obtained by the existing method (without introducing the high-energy particle precipitation flux) are given.

[0101] From the above comparison, it can be seen that the parameter prediction results obtained by the prediction method of the present invention are reduced from 4.25 to 2.59, and the overall relative error PD is reduced from 0.45 to 0.17 compared with the existing method, which fully demonstrates that the introduction of high-energy particle precipitation flux in the prediction method of the present invention can improve the accuracy of TEC prediction results.

[0102] The present invention has been described above by way of example with reference to the accompanying drawings. It is apparent that the specific implementation of the present invention is not limited to the above-described embodiments. Those skilled in the art may make various modifications or variations to the present invention without departing from the technical concept of the present invention, and such modifications or variations shall naturally fall within the scope of protection of the present invention.

Claims

1. A high-latitude ionospheric prediction method based on multi-source heterogeneous data, characterized by: The following steps are involved: S1, establish a time series dataset of ground-based ionospheric measurement data and space weather index set; S10, ground-based ionospheric measurement data, refers to the TEC data measured by ground receivers, that is, the total number of free electrons in a unit cross-sectional area column perpendicular to the propagation path; S11, space weather index set, including Dst index, IMF-Bz index, and F10.7 index; TEC: total electron content, Dst index: geomagnetic storm ring current index, IMF-Bz index: interplanetary magnetic field north-south component index, F10.7 index: 10.7 cm solar radio flux index; S2, establish a time series dataset of space-based high-energy particle fallout flux measurement data; S20: The satellite is in constant motion and cannot make continuous measurements of a single point. However, energetic particle precipitation mainly occurs in the auroral belt area above 60° magnetic latitude. All satellite measurements of the energetic particle precipitation flux in this area can be recorded. S3 normalizes all time series data sets and establishes an input data set based on multi-source heterogeneous data; S4, establishes an LSTM prediction neural network and obtains the forecast results of high-latitude ionospheric parameters based on the input data set.

2. The high-latitude ionospheric prediction method based on multi-source heterogeneous data according to claim 1, characterized in that: The step S10 includes: The ionospheric TEC data measured by the ground-based GNSS receiver in high latitudes is used as the ground-based measurement data. According to a certain time resolution, a time series data set of TEC data is established. TEC express: S TEC = {TEC(t i ) | t i = t0 + (i - 1)Δt, i = 1, 2, …, N i}} Equation 1; Among them, TEC(t i ) represents t i The value of the ionospheric TEC data at the moment, t is the time, t0 is the starting time, the subscript i is the time sequence number, N i is the total number of data, Δt is the time resolution; The step S11 includes: According to a certain time resolution, the time series data sets of Dst index, IMF-Bz index and F10.7 index are established, respectively using S Dst 、S Bz and S F107 express: Establish a time series dataset of space weather index set {S Dst ,S Bz ,S F107 }; Among them, Dst(t i )、Bz(t i )、F107(t i ) represent t i The values ​​of the Dst index, IMF-Bz index, and F10.7 index at the moment, t is the time, t0 is the starting time, the subscript i is the time sequence number, N i is the total number of data, and Δt is the time resolution.

3. The high-latitude ionosphere prediction method based on multi-source heterogeneous data according to claim 2, characterized in that: The S20 step includes: All the measured values ​​in S1 are averaged with time resolution to obtain a time series data set of high energy particle precipitation flux measurement data. P express: Among them, P k represents the measured value of high-energy particle precipitation flux, k is the measurement value number, φ is the magnetic latitude value corresponding to the high-energy particle precipitation flux, P(t i ) represents t i The value of the high-energy particle precipitation flux at time t is time, t0 is the starting time, subscript i is the time sequence number, N i is the total number of data, Δt is the time resolution, 1 条件 It is an indicator function, which takes 1 when the condition is met and 0 otherwise.

4. The high-latitude ionosphere prediction method based on multi-source heterogeneous data according to claim 3, characterized in that: The S3 step includes: S31, introduce local time series data, and use L(t i ) indicates that it is also normalized to obtain the local time sine value normalized time series data set and local time cosine value normalized time series dataset S32: Normalize the time series data sets of ground-based ionospheric measurement data, space weather index set, and space-based high-energy particle fallout flux measurement data; obtain the normalized time series data set of TEC data Normalized time series dataset of Dst index Normalized time series dataset of the IMF-Bz index Normalized time series dataset with F10.7 index and a normalized time series dataset of high-energy particle precipitation flux measurements S33, slide and take 5 consecutive time series data from the normalized time series data set to establish the neural network input data samples, respectively using TEC5(t i )、Dst5(t i )、Bz5(t i )、F1075(t i )、P5(t i )、Ls5(t i ) and Lc5(t i ) indicates that the input data set is represented by X TEC 、X Dst 、X Bz 、X F107 、X P 、X Ls and X Lc : Formula 6.

5. The high-latitude ionosphere prediction method based on multi-source heterogeneous data according to claim 4, characterized in that: The S4 step includes: S41, establish an LSTM prediction neural network, the input neuron uses the input data set, the number of output layer neurons is 1, expressed as Y TEC , that is, the ionospheric TEC value at the time corresponding to the forecast of Δt in advance; S42, based on the trained prediction neural network, inputting the input samples 5 moments before the time to be predicted, and obtaining the high-latitude ionosphere prediction result; S43, uses two parameters to evaluate the forecast error, including RMSE and PD: RMSE is the root mean square error, calculated as: PD is the overall relative error, calculated as: Among them, j is the serial number of the sample used for error statistics, N j The total number of samples used for the error statistics.

Images