Ionized layer accidental E layer identification method and system based on machine learning

Abstract

The invention discloses an ionosphere accidental E layer identification method and system based on machine learning. The method comprises the following steps: receiving occultation event data in an atmospheric profile file transmitted by a satellite, and calculating tangent point coordinate data of an occultation event according to the occultation event data; according to the tangent point coordinate data of the occultation event, screening coordinate data in an ionospheric accidental E layer occurrence area, calculating a signal-to-noise ratio disturbance sequence and an amplitude flicker index according to the coordinate data, and constructing an ionospheric accidental E layer identification model; and the ionosphere accidental E-layer identification model performs ionosphere accidental E-layer identification according to the signal-to-noise ratio disturbance sequence and the amplitude flicker index, and determines whether an ionosphere accidental E-layer event occurs according to an identification result. According to the method and the system, whether the ionospheric accidental E-layer event occurs or not is identified and judged by using a single threshold value can be effectively avoided, and the influence of the positioning error of the system is reduced, so that the identification accuracy of the ionospheric accidental E-layer event is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of ionosphere identification, and in particular to a method and system for identifying sporadic ionosphere E layers based on machine learning. Background Art

[0002] The ionosphere is a special region of the Earth's atmosphere, located between 60 kilometers above the ground and the top of the magnetosphere. Its main characteristic is that atmospheric molecules and atoms are ionized under the influence of solar ultraviolet rays, X-rays and high-energy particles, forming free electrons and ions, thus presenting an ionized state. The structure of the ionosphere is complex and is usually divided into the D layer, E layer and F layer (the F layer is further divided into the F1 layer and the F2 layer). The maximum electron density is about 10 per cubic centimeter. 6 electrons, mainly concentrated near an altitude of 300 kilometers.

[0003] The sporadic E layer (Es layer) is a unique inhomogeneous structure within the ionosphere's E layer, typically occurring at altitudes between 90 and 130 kilometers above Earth's surface. The electron density of the Es layer is significantly higher than that of the surrounding area, sometimes even exceeding the peak electron density of the F layer. This thin, high-electron-density layer, composed primarily of metal ions, typically ranges from 0.5 to 5 kilometers in thickness and extends horizontally over tens to hundreds of kilometers. The formation of the Es layer is associated with a variety of physical mechanisms, including neutral wind shear, the convergence of metal ions, and complex dynamical processes. Its appearance is random and unpredictable, occurring during the day or at night, and lasting from tens of minutes to hours. The characteristics of the Es layer vary depending on geographic location, appearing more often during the day at low latitudes and in the equatorial region and more frequently at night at high latitudes and in the polar regions. Furthermore, the Es layer's high electron density significantly affects radio wave propagation. It can reflect or scatter electromagnetic waves with frequencies as high as 50 to 80 MHz, enabling long-distance high-frequency or very high-frequency communications. However, the irregular structure of the Es layer can also cause signal flickering, satellite signal loss, or ranging errors, interfering with technical systems such as communications, navigation, and radar detection. Therefore, studying the spatiotemporal evolution of the Es layer and its impact on radio wave propagation is of great significance for optimizing communication systems and space weather forecasting.

[0004] Currently, most existing methods use only the S4 index (amplitude scintillation index) to identify sporadic ionospheric E-layer events. The S4 index is an important parameter used to quantify sporadic ionospheric E-layer events, mainly used to describe the rapid random fluctuations in signal amplitude. The value of the S4 index ranges from 0 to 1 and is often used to assess the impact of ionospheric scintillation on satellite communications and navigation systems (such as GPS). However, when using only the S4 index to identify sporadic ionospheric E-layer events, only a single threshold between 0 and 1 can be used for identification and judgment, which has significant limitations. For example, when the S4 index is greater than 0.4, the positioning error of the navigation system will increase significantly, resulting in a low accuracy rate in identifying sporadic ionospheric E-layer events. Summary of the Invention

[0005] In order to overcome the problem in the prior art that when a single threshold is used to identify sporadic ionospheric E-layer events, the positioning error of the navigation system will increase significantly when the threshold is greater than a certain value, resulting in a low accuracy rate in identifying sporadic ionospheric E-layer events, the purpose of the present invention is to propose an ionospheric E-layer identification method and system based on machine learning, which can avoid using a single threshold to identify and determine whether an ionospheric E-layer event has occurred, reduce the impact of the system's positioning error, and thus improve the accuracy rate of identifying sporadic ionospheric E-layer events.

[0006] To achieve the purpose of the present invention, the present invention adopts the following technical solutions:

[0007] A method for identifying sporadic ionospheric E-layers based on machine learning, the method comprising the following steps:

[0008] Receive occultation event data in the atmospheric profile file transmitted by the satellite, and calculate the tangent point coordinate data of the occultation event based on the occultation event data;

[0009] The coordinate data of the tangent point coordinate data of the occultation event are used to screen the coordinate data within the region where the sporadic ionospheric E layer occurs. The signal-to-noise ratio disturbance sequence and amplitude scintillation index are calculated based on the coordinate data, and the sporadic ionospheric E layer identification model is constructed.

[0010] The ionospheric sporadic E-layer identification model identifies ionospheric sporadic E-layer events based on the signal-to-noise ratio disturbance sequence and the amplitude scintillation index, and confirms whether an ionospheric sporadic E-layer event occurs based on the identification results.

[0011] In the above technical solution, by calculating the tangent point coordinate data of the occultation event, the occurrence area of the ionospheric sporadic E layer can be confirmed, and then the coordinate data within the occurrence area of the ionospheric sporadic E layer can be screened according to the tangent point coordinate data of the occultation event, so as to calculate the signal-to-noise ratio disturbance sequence and the amplitude scintillation index according to the coordinate data, and construct an ionospheric sporadic E layer recognition model to identify the ionospheric sporadic E layer according to the calculated signal-to-noise ratio disturbance sequence and the amplitude scintillation index, so as to confirm whether the ionospheric sporadic E layer event occurs according to the recognition result; by calculating the signal-to-noise ratio disturbance sequence and the amplitude scintillation index, it is possible to avoid using a single threshold to identify and judge whether the ionospheric sporadic E layer event occurs. The constructed ionospheric sporadic E layer recognition model is a machine learning model, which can perform automatic machine learning according to the signal-to-noise ratio disturbance sequence and the amplitude scintillation index in the occultation event data to reduce the influence of the system's positioning error, thereby improving the accuracy of identifying the ionospheric sporadic E layer event.

[0012] Furthermore, the process of calculating the tangent point coordinate data of the occultation event based on the occultation event data includes:

[0013] The satellite's spatial position coordinates in the Earth-centered inertial coordinate system (ECI) and the velocity component data in the coordinate axis direction are extracted based on the occultation event data.

[0014] Calculate the ECI coordinates of the occultation tangent point when the occultation event occurs based on the spatial position coordinates and the velocity component data in the coordinate axis direction;

[0015] According to the calculated ECI coordinates of the occultation tangent point when the occultation event occurs, the Earth-centered inertial coordinate system ECI is converted into the Earth-centered Earth-fixed coordinate system ECEF, and the geodetic coordinates of the occultation tangent point are calculated according to the Earth-centered Earth-fixed coordinate system ECEF to obtain the tangent point coordinate data of the occultation event.

[0016] Furthermore, the expression for calculating the ECI coordinates of the occultation tangent point when the occultation event occurs is:

[0017]

[0018] The expression for transforming the Earth-centered inertial coordinate system ECI into the Earth-centered Earth-fixed coordinate system ECEF is:

[0019]

[0020] The expressions for calculating the geodetic coordinates of the occultation tangent point based on the Earth-centered Earth-fixed coordinate system ECEF include:

[0021]

[0022] in, The ECI coordinates of the occultation tangent point when the occultation event occurs in vector form; and represents the ECI coordinates of the LEO satellite and the GPS satellite respectively; || represents the modulus of the vector; X, Y, Z represent the spatial rectangular coordinates of the occultation tangent point in the ECEF coordinate frame; B, L, H represent the geodetic coordinates of the occultation tangent point in the ECEF coordinate frame, which are the geodetic latitude, geodetic longitude and altitude respectively; a, e and e′ represent the semi-major axis, the first eccentricity and the second eccentricity of the Earth reference ellipsoid respectively; N represents the radius of the occultation circle corresponding to the geodetic latitude of the occultation tangent point.

[0023] In the above technical solution, by calculating the ECI coordinates of the occultation tangent point when the occultation event occurs, converting the geocentric inertial coordinate system ECI into the geocentric Earth-fixed coordinate system ECEF, and calculating the geodetic coordinates of the occultation tangent point based on the geocentric Earth-fixed coordinate system ECEF, the tangent point coordinate data of the occultation event is obtained, thereby confirming the occurrence area of the sporadic E layer in the ionosphere.

[0024] Furthermore, the process of calculating the signal-to-noise ratio disturbance sequence based on the coordinate data in the ionospheric sporadic E-layer occurrence area includes:

[0025] Extract the satellite's moisture-free atmospheric profile data based on the coordinate data within the ionospheric sporadic E layer occurrence area, and extract the signal-to-noise ratio data in the occultation event data based on the satellite's moisture-free atmospheric profile data. Use the sliding average method to average a continuous section of the signal-to-noise ratio data to obtain the background value of the signal-to-noise ratio data.

[0026] The background value is subtracted from the original signal-to-noise ratio data to obtain a signal-to-noise ratio perturbation sequence, and the signal-to-noise ratio normalized standard deviation is calculated based on the signal-to-noise ratio perturbation sequence.

[0027] Furthermore, the process of calculating the normalized standard deviation of the signal-to-noise ratio according to the signal-to-noise ratio disturbance sequence includes:

[0028] The signal-to-noise ratio disturbance sequence is normalized to extract the relevant characteristic parameters of the sporadic E layer in the ionosphere. The expression is:

[0029]

[0030] According to the relevant characteristic parameters of the sporadic E layer in the ionosphere, the normalized standard deviation of the signal-to-noise ratio is calculated using the sliding average method. The expression is:

[0031]

[0032] Among them, SNR nstd represents the normalized standard deviation of the signal-to-noise ratio, X i represents the i-th data after normalization of the signal-to-noise ratio disturbance sequence, represents the i-th data after the signal-to-noise ratio disturbance sequence is smoothed, and k represents the size of the sliding average window.

[0033] In the above technical solution, by performing normalization operation on the calculated signal-to-noise ratio disturbance sequence, data such as short arc segments in the sequence data can be effectively proposed, thereby improving the practicality and reliability of the data, and effectively extracting the relevant characteristic parameters of the sporadic E layer of the ionosphere, thereby calculating the normalized standard deviation of the signal-to-noise ratio of the signal-to-noise ratio disturbance sequence. The calculated normalized standard deviation of the signal-to-noise ratio can effectively improve the accuracy and effectiveness of the signal-to-noise ratio data in detecting Es events.

[0034] Furthermore, the process of calculating the amplitude scintillation index based on the coordinate data within the ionospheric sporadic E-layer occurrence area includes:

[0035] According to the coordinate data of the ionospheric sporadic E layer occurrence area, the signal intensity fluctuation data in the occultation event data is calculated, and the expression is:

[0036]

[0037] The amplitude flicker index S4 is calculated based on the signal intensity fluctuation data. The expression is:

[0038]

[0039] Among them, σ I Indicates the root mean square value of signal strength fluctuation, I represents the signal strength, Indicates the mean value of the signal strength, and <> indicates averaging the physical quantity within the preset number of samples.

[0040] Furthermore, the ionospheric sporadic E-layer identification model performs the following process on the ionospheric sporadic E-layer identification based on the signal-to-noise ratio disturbance sequence and the amplitude scintillation index:

[0041] The ionospheric sporadic E-layer identification model includes a forward long short-term memory unit and a reverse long short-term memory unit;

[0042] Normalize the signal-to-noise ratio to the standard deviation SNR nstd The ionospheric sporadic E-layer recognition model is trained for several rounds using the training set, wherein:

[0043] The forward long short-term memory unit is normalized according to the input signal-to-noise ratio (SNR). nstd Produce a positive hidden state with the amplitude flicker index S4;

[0044] The reverse long short-term memory unit is normalized according to the input signal-to-noise ratio (SNR) nstd Producing an inverse hidden state with the amplitude flicker index S4;

[0045] Merging the forward hidden state and the reverse hidden state to obtain a bidirectional hidden state, and performing ionospheric sporadic E-layer identification based on the bidirectional hidden state;

[0046] The test set is used to test the results of the ionospheric sporadic E-layer recognition. A loss function is set to optimize the parameters of the ionospheric sporadic E-layer recognition model during the training process. When the training round ends or the loss function converges, the trained ionospheric sporadic E-layer recognition model is obtained.

[0047] The trained ionospheric sporadic E layer recognition model is used to normalize the standard deviation of the signal-to-noise ratio (SNR). nstd The ionospheric sporadic E-layer event is identified by combining the amplitude scintillation index S4, and whether an ionospheric sporadic E-layer event occurs is confirmed based on the identification results.

[0048] Furthermore, the expression for merging the forward hidden state and the reverse hidden state is:

[0049]

[0050] in, represents the forward hidden state, Represents the inverse hidden state.

[0051] In the above technical solution, the signal-to-noise ratio normalized standard deviation SNR is used nstd The ionospheric sporadic E-layer recognition model constructed by combining the amplitude scintillation index S4 is trained for several rounds. This enables the ionospheric sporadic E-layer recognition model to automatically learn the characteristics of ionospheric sporadic E-layer events in occultation event data during the training process, thereby achieving efficient and accurate recognition of ionospheric sporadic E-layer events.

[0052] A system for identifying sporadic ionospheric E-layers based on machine learning, the system comprising:

[0053] A data receiving module is used to receive occultation event data in the atmospheric profile file transmitted by the satellite;

[0054] A data processing module, used for calculating the tangent point coordinate data of the occultation event based on the occultation event data;

[0055] The data processing module is further configured to filter coordinate data within the ionospheric sporadic E layer occurrence region based on the tangent point coordinate data of the occultation event, so as to calculate the signal-to-noise ratio disturbance sequence and the amplitude scintillation index based on the coordinate data;

[0056] Model building module, used to build the ionospheric sporadic E layer identification model;

[0057] The ionospheric sporadic E-layer identification module is used to identify the ionospheric sporadic E-layer according to the signal-to-noise ratio disturbance sequence and the amplitude scintillation index of the ionospheric sporadic E-layer identification model, and confirm whether the ionospheric sporadic E-layer event occurs based on the identification results.

[0058] An electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, the steps of a method for identifying sporadic ionospheric E layers based on machine learning are implemented.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] The present invention proposes a method and system for identifying sporadic ionospheric E-layer events based on machine learning. By calculating the tangent point coordinate data of an occultation event, the occurrence area of the sporadic ionospheric E-layer event can be confirmed. Then, the coordinate data within the occurrence area of the sporadic ionospheric E-layer event is screened according to the tangent point coordinate data of the occultation event, so as to calculate the signal-to-noise ratio disturbance sequence and the amplitude scintillation index according to the coordinate data, and construct an ionospheric sporadic E-layer identification model. The ionospheric sporadic E-layer event is identified according to the calculated signal-to-noise ratio disturbance sequence and the amplitude scintillation index, so as to confirm whether the sporadic ionospheric E-layer event has occurred according to the identification result. By calculating the signal-to-noise ratio disturbance sequence and the amplitude scintillation index, the use of a single threshold to identify and judge whether the sporadic ionospheric E-layer event has occurred can be avoided. The constructed ionospheric sporadic E-layer identification model is a machine learning model, which can automatically perform machine learning according to the signal-to-noise ratio disturbance sequence and the amplitude scintillation index in the occultation event data, so as to reduce the influence of the positioning error of the system, thereby improving the accuracy of identifying the sporadic ionospheric E-layer event. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 A flowchart of the steps of a method for identifying sporadic ionospheric E-layers based on machine learning provided in an embodiment of the present application;

[0062] Figure 2 A basic data flow chart for identifying sporadic ionospheric E-layer events provided in an embodiment of the present application;

[0063] Figure 3 A schematic diagram of a confusion matrix for identifying sporadic ionospheric E-layer events provided in an embodiment of the present application;

[0064] Figure 4 A comparison chart of the effects of identifying sporadic ionospheric E-layer events provided in an embodiment of the present application;

[0065] Figure 5 A structural diagram of an ionospheric sporadic E-layer identification system based on machine learning is provided in an embodiment of the present application. DETAILED DESCRIPTION

[0066] To facilitate understanding of the present invention, the present invention will be described more fully below with reference to the accompanying drawings. Preferred embodiments of the present invention are shown in the accompanying drawings. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the present disclosure.

[0067] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The terms used in this specification of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0068] Example 1:

[0069] This embodiment provides a method for identifying sporadic ionospheric E layers based on machine learning. Figure 1 , the method comprises the following steps:

[0070] Step S1: receiving occultation event data in an atmospheric profile file transmitted by a satellite, and calculating tangent point coordinate data of the occultation event based on the occultation event data;

[0071] Step 2: Based on the tangent point coordinate data of the occultation event, the coordinate data within the ionospheric sporadic E layer occurrence area are screened to calculate the signal-to-noise ratio disturbance sequence and amplitude scintillation index based on the coordinate data, and to construct an ionospheric sporadic E layer identification model;

[0072] Step S3: The ionospheric sporadic E-layer identification model identifies the ionospheric sporadic E-layer according to the signal-to-noise ratio disturbance sequence and the amplitude scintillation index, and confirms whether the ionospheric sporadic E-layer event occurs based on the identification result.

[0073] As a preferred embodiment, in step S1, see Figure 2 The process of calculating the tangent point coordinate data of the occultation event based on the occultation event data includes:

[0074] The satellite's spatial position coordinates in the Earth-centered inertial coordinate system (ECI) and the velocity component data in the coordinate axis direction are extracted based on the occultation event data.

[0075] Calculate the ECI coordinates of the occultation tangent point when the occultation event occurs based on the spatial position coordinates and the velocity component data in the coordinate axis direction;

[0076] According to the calculated ECI coordinates of the occultation tangent point when the occultation event occurs, the Earth-centered inertial coordinate system ECI is converted into the Earth-centered Earth-fixed coordinate system ECEF, and the geodetic coordinates of the occultation tangent point are calculated according to the Earth-centered Earth-fixed coordinate system ECEF to obtain the tangent point coordinate data of the occultation event.

[0077] Specifically, the water-free atmospheric profile atmPhs file of Level 1b was downloaded from the COSMIC-1 satellite, and the information provided in the file on the occurrence of many occultation events was obtained, such as universal time, the spatial position coordinates of GPS and LEO satellites in the Earth Centered Inertial (ECI) coordinate system, and the velocity components in the coordinate axis direction.

[0078] As a preferred embodiment, the expression for calculating the ECI coordinates of the occultation tangent point when an occultation event occurs is:

[0079]

[0080] The above coordinates are all in vector form. They include the coordinate components along the X-axis, Y-axis, and Z-axis in the ECI coordinate system. The expression for converting the Earth-centered inertial coordinate system ECI to the Earth-centered Earth-fixed coordinate system ECEF is:

[0081]

[0082] After the coordinate system conversion, the coordinates of the occultation tangent point are still in the form of spatial rectangular coordinates, and it is impossible to intuitively understand the altitude of the occultation tangent point. Therefore, the expression for calculating the geodetic coordinates of the occultation tangent point based on the Earth-centered Earth-fixed coordinate system ECEF includes:

[0083]

[0084] in, The ECI coordinates of the occultation tangent point when the occultation event occurs in vector form; and represents the ECI coordinates of the LEO satellite and the GPS satellite respectively; || represents the modulus of the vector; X, Y, Z represent the spatial rectangular coordinates of the occultation tangent point in the ECEF coordinate frame; B, L, H represent the geodetic coordinates of the occultation tangent point in the ECEF coordinate frame, which are the geodetic latitude, geodetic longitude and altitude respectively; a, e and e′ represent the semi-major axis, the first eccentricity and the second eccentricity of the Earth reference ellipsoid respectively; N represents the radius of the occultation circle corresponding to the geodetic latitude of the occultation tangent point.

[0085] It can be understood that by calculating the ECI coordinates of the occultation tangent point when the occultation event occurs, converting the geocentric inertial coordinate system ECI into the geocentric earth-fixed coordinate system ECEF, and calculating the geodetic coordinates of the occultation tangent point based on the geocentric earth-fixed coordinate system ECEF, the tangent point coordinate data of the occultation event is obtained, thereby confirming the occurrence area of the sporadic E layer in the ionosphere.

[0086] As a preferred embodiment, in step S2, see Figure 2 The process of calculating the signal-to-noise ratio disturbance sequence based on the coordinate data in the ionospheric sporadic E layer occurrence area includes:

[0087] Extract the satellite's moisture-free atmospheric profile data based on the coordinate data within the ionospheric sporadic E layer occurrence area, and extract the signal-to-noise ratio data in the occultation event data based on the satellite's moisture-free atmospheric profile data. Use the sliding average method to average a continuous section of the signal-to-noise ratio data to obtain the background value of the signal-to-noise ratio data.

[0088] The background value is subtracted from the original signal-to-noise ratio data to obtain a signal-to-noise ratio perturbation sequence, and the signal-to-noise ratio normalized standard deviation is calculated based on the signal-to-noise ratio perturbation sequence.

[0089] Specifically, since the signal-to-noise ratio at altitudes above 35 kilometers is relatively stable, reflecting that the ionospheric activity is less or more uniform, the sliding average method can be used to calculate the background value. This method takes the average of a continuous period of signal-to-noise ratio data and uses the average value as the background value for that period of data. To ensure the accuracy of the calculation, the sliding window is usually set to an odd number, which ensures that the data points in the center of the window are fully included, avoiding deviations caused by window offsets, and also improving the symmetry and stability of the data. After obtaining the background value sequence, the background value is subtracted from the original signal-to-noise ratio data to obtain the signal-to-noise ratio disturbance sequence caused by the Es layer. Although the normalized data can preliminarily show the disturbance caused by Es on the data, it still has limitations in determining whether Es occurs, its specific location, intensity, thickness, and number of layers. In order to more accurately identify and evaluate Es layer events, we need to further process the signal-to-noise ratio normalized data to extract characteristic parameters related to the Es layer. Doing so will help improve the accuracy and effectiveness of signal-to-noise ratio data in detecting Es events.

[0090] As a preferred embodiment, the process of calculating the normalized standard deviation of the signal-to-noise ratio according to the signal-to-noise ratio disturbance sequence includes:

[0091] The signal-to-noise ratio disturbance sequence is normalized to extract the relevant characteristic parameters of the sporadic E layer in the ionosphere. The expression is:

[0092]

[0093] According to the relevant characteristic parameters of the sporadic E layer in the ionosphere, the normalized standard deviation of the signal-to-noise ratio is calculated using the sliding average method. The expression is:

[0094]

[0095] Among them, SNR nstd represents the normalized standard deviation of the signal-to-noise ratio, X i represents the i-th data after normalization of the signal-to-noise ratio disturbance sequence, represents the i-th data after the signal-to-noise ratio disturbance sequence is smoothed, and k represents the size of the sliding average window.

[0096] In this example, the normalized standard deviation of the signal-to-noise ratio is calculated, and the background value of the signal-to-noise ratio is calculated using a sliding average method with a window length of 2 km to obtain the disturbance sequence. The data study altitude range is 80-130 km, which is sufficient to cope with the data reduction brought about by two sliding averages. Each signal-to-noise ratio disturbance sequence is then normalized to avoid the influence of different basic signal power values on further data processing:

[0097]

[0098] Here, X represents the normalized value; || represents the absolute value. To more accurately identify and evaluate Es layer events, we further calculated the normalized standard deviation of the signal-to-noise ratio to determine whether Es occurred, its specific location, intensity, thickness, and number of layers. 0.1 was then used as the threshold for determining Es occurrence.

[0099] It can be understood that by normalizing the calculated signal-to-noise ratio disturbance sequence, data such as short arc segments in the sequence data can be effectively proposed, the practicality and reliability of the data can be improved, and the relevant characteristic parameters of the sporadic E layer of the ionosphere can be effectively extracted, thereby calculating the signal-to-noise ratio normalized standard deviation of the signal-to-noise ratio disturbance sequence. The calculated signal-to-noise ratio normalized standard deviation can effectively improve the accuracy and effectiveness of the signal-to-noise ratio data in detecting Es events.

[0100] As a preferred embodiment, in step S2, see Figure 2 The process of calculating the amplitude scintillation index based on the coordinate data of the ionospheric sporadic E layer occurrence area includes:

[0101] Specifically, unlike the inverted quantities provided for the atmosphere and ionosphere, such as temperature, water vapor, and electron density, the S4 index is calculated after the inversion process. The onboard algorithm of the GPS occultation receiver does not measure the S4 index directly, but rather measures the signal strength (the square of the signal-to-noise ratio) fluctuations from the raw 50Hz L1 amplitude measurements, which are recorded in the data stream at a rate of 1Hz and minimized. Therefore, the signal strength fluctuation data from the occultation event data is calculated based on the coordinate data within the region where the ionospheric E layer occurs, using the expression:

[0102]

[0103] Assuming this is a Gaussian distribution:

[0104]

[0105] In order to obtain the average intensity per second, which is considered to be more accurate , a low-pass filter needs to be applied to the original This can reduce the impact of received signal power changes due to satellite motion, and thus calculate the amplitude scintillation index S4 based on the signal strength fluctuation data, as shown in the following expression:

[0106]

[0107] Among them, σ I Indicates the root mean square value of signal strength fluctuation, I represents the signal strength, Indicates the mean value of the signal strength, and <> indicates averaging the physical quantity within the preset number of samples.

[0108] As a preferred embodiment, in step S3, see Figure 2 The process of identifying sporadic ionospheric E-layers based on the signal-to-noise ratio disturbance sequence and the amplitude scintillation index includes:

[0109] The ionospheric sporadic E-layer identification model includes a forward long short-term memory unit and a reverse long short-term memory unit;

[0110] Normalize the signal-to-noise ratio to the standard deviation SNR nstd The ionospheric sporadic E-layer recognition model is trained for several rounds using the training set, wherein:

[0111] The forward long short-term memory unit is normalized according to the input signal-to-noise ratio (SNR). nstd Produce a positive hidden state with the amplitude flicker index S4;

[0112] The reverse long short-term memory unit is normalized according to the input signal-to-noise ratio (SNR) nstd Producing an inverse hidden state with the amplitude flicker index S4;

[0113] Merging the forward hidden state and the reverse hidden state to obtain a bidirectional hidden state, and performing ionospheric sporadic E-layer identification based on the bidirectional hidden state;

[0114] The test set is used to test the results of the ionospheric sporadic E-layer recognition. A loss function is set to optimize the parameters of the ionospheric sporadic E-layer recognition model during the training process. When the training round ends or the loss function converges, the trained ionospheric sporadic E-layer recognition model is obtained.

[0115] The trained ionospheric sporadic E layer recognition model is used to normalize the standard deviation of the signal-to-noise ratio (SNR). nstd The ionospheric sporadic E layer is identified with the amplitude scintillation index S4, and whether an ionospheric sporadic E layer event has occurred is confirmed based on the identification result. The ionospheric sporadic E layer identification model is a bidirectional long short-term memory model (BiLSTM model).

[0116] As a preferred embodiment, the expression for merging the forward hidden state and the reverse hidden state is:

[0117]

[0118] in, represents the forward hidden state, Represents the inverse hidden state.

[0119] Specifically, the core of the BiLSTM model is that it contains LSTM units in two directions at the same time:

[0120] Forward LSTM: processes data one by one from the beginning to the end of the sequence (i.e., from time step t=1 to t=T).

[0121] Backward LSTM: processes data one by one from the end of the sequence to the beginning (i.e., from time step t=T to t=1).

[0122] The LSTM unit in each direction generates a hidden state, which represents the context information of the sequence in that direction.

[0123] Merging of two-way information:

[0124] At each time step t, the forward LSTM and the backward LSTM will output a hidden state, respectively and BiLSTM merges these two hidden states to generate the final bidirectional hidden state h t There are usually two ways to merge:

[0125] ①Concatenation:

[0126]

[0127] This approach preserves all information in both directions, but increases the dimensionality of the hidden state.

[0128] ②Summation:

[0129]

[0130] This approach does not increase the dimensionality, but may lose some information.

[0131] Mathematical representation of BiLSTM:

[0132] Assume that the input sequence is X={x1,x2,…,x T }, the calculation process of BiLSTM can be expressed as:

[0133] Forward LSTM:

[0134]

[0135] in, represents the hidden state of the forward LSTM at time step t.

[0136] Backward LSTM:

[0137]

[0138] in, represents the hidden state of the reverse LSTM at time step t.

[0139] Bidirectional hidden state:

[0140]

[0141] Finally, the bidirectional hidden states {h1,h2,…,ht} at all time steps can be used for subsequent tasks such as classification, regression or sequence generation.

[0142] In this embodiment, the normalized standard deviation sequence of the signal-to-noise ratio is interpolated every 500 meters from 90 km to 120 km to obtain sequence data of the same length, eliminating the problem of being unable to train due to different lengths. The interpolated data are trained using the Bilstm, Lstm, and SVM learning models respectively, and the ratio of the training set to the test set is set to 0.8:0.2; the prediction accuracy of the three models is shown in Table 1. It can be seen from Table 1 that the prediction performance of the Bilstm model is the best, and it performs best in the three evaluation indicators of precision, accuracy, and regression rate. Therefore, this example uses the Bilstm model as an ionospheric sporadic E-layer recognition model, and draws the confusion matrix and recognition example. Figure 3 , Figure 4 As shown in the figure, the prediction effect is more intuitive. Figure 4 The black circles represent manually annotated labels, and the gray crosses represent predicted labels. 0 and 1 correspond to no and recognized Es, respectively. The results show that the actual and predicted labels are very close, demonstrating the feasibility of using the Bilstm model for identifying sporadic Es in the ionosphere.

[0143] Table 1:

[0144] Bilstm Lstm Support Vector Machine Accuracy 0.83955 0.80051 0.71055 Accuracy 0.91272 0.85934 0.74016 Regression rate 0.74676 0.71901 0.65603

[0145] It can be understood that the standard deviation SNR is normalized using the signal-to-noise ratio nstd The ionospheric sporadic E-layer recognition model constructed by combining the amplitude scintillation index S4 is trained for several rounds. This enables the ionospheric sporadic E-layer recognition model to automatically learn the characteristics of ionospheric sporadic E-layer events in occultation event data during the training process, thereby achieving efficient and accurate recognition of ionospheric sporadic E-layer events.

[0146] In this embodiment, by calculating the tangent point coordinate data of the occultation event, the occurrence area of the ionospheric sporadic E layer can be confirmed, and then the coordinate data within the occurrence area of the ionospheric sporadic E layer is screened according to the tangent point coordinate data of the occultation event, so as to calculate the signal-to-noise ratio disturbance sequence and the amplitude scintillation index according to the coordinate data, and construct an ionospheric sporadic E layer identification model to identify the ionospheric sporadic E layer according to the calculated signal-to-noise ratio disturbance sequence and the amplitude scintillation index, so as to confirm whether the ionospheric sporadic E layer event occurs according to the identification result; by calculating the signal-to-noise ratio disturbance sequence and the amplitude scintillation index, it is possible to avoid using a single threshold to identify and judge whether the ionospheric sporadic E layer event occurs. The constructed ionospheric sporadic E layer identification model is a machine learning model, which can perform automatic machine learning according to the signal-to-noise ratio disturbance sequence and the amplitude scintillation index in the occultation event data to reduce the influence of the positioning error of the system, thereby improving the accuracy of identifying the ionospheric sporadic E layer event.

[0147] Example 2:

[0148] This embodiment provides an ionospheric sporadic E-layer identification system based on machine learning, see Figure 5 , the system comprising:

[0149] A data receiving module is used to receive occultation event data in the atmospheric profile file transmitted by the satellite;

[0150] A data processing module, used for calculating the tangent point coordinate data of the occultation event based on the occultation event data;

[0151] The data processing module is further configured to filter coordinate data within the ionospheric sporadic E layer occurrence region based on the tangent point coordinate data of the occultation event, so as to calculate the signal-to-noise ratio disturbance sequence and the amplitude scintillation index based on the coordinate data;

[0152] Model building module, used to build the ionospheric sporadic E layer identification model;

[0153] The ionospheric sporadic E-layer identification module is used to identify the ionospheric sporadic E-layer according to the signal-to-noise ratio disturbance sequence and the amplitude scintillation index of the ionospheric sporadic E-layer identification model, and confirm whether the ionospheric sporadic E-layer event occurs based on the identification results.

[0154] Example 3:

[0155] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, the steps of a method for identifying sporadic ionospheric E layers based on machine learning are implemented.

[0156] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention's description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for identifying sporadic ionospheric E layers based on machine learning, characterized in that: The method comprises the following steps: Receive occultation event data in the atmospheric profile file transmitted by the satellite, and calculate the tangent point coordinate data of the occultation event based on the occultation event data; The coordinate data of the tangent point coordinate data of the occultation event are used to screen the coordinate data within the region where the sporadic ionospheric E layer occurs. The signal-to-noise ratio disturbance sequence and amplitude scintillation index are calculated based on the coordinate data, and the sporadic ionospheric E layer identification model is constructed. The ionospheric sporadic E-layer identification model identifies ionospheric sporadic E-layer events based on the signal-to-noise ratio disturbance sequence and the amplitude scintillation index, and confirms whether an ionospheric sporadic E-layer event occurs based on the identification results.

2. The method for identifying sporadic ionospheric E-layers based on machine learning according to claim 1, characterized in that: The process of calculating the tangent point coordinate data of the occultation event based on the occultation event data includes: The satellite's spatial position coordinates in the Earth-centered inertial coordinate system (ECI) and the velocity component data in the coordinate axis direction are extracted based on the occultation event data. Calculate the ECI coordinates of the occultation tangent point when the occultation event occurs based on the spatial position coordinates and the velocity component data in the coordinate axis direction; According to the calculated ECI coordinates of the occultation tangent point when the occultation event occurs, the Earth-centered inertial coordinate system ECI is converted into the Earth-centered Earth-fixed coordinate system ECEF, and the geodetic coordinates of the occultation tangent point are calculated according to the Earth-centered Earth-fixed coordinate system ECEF to obtain the tangent point coordinate data of the occultation event.

3. The method for identifying sporadic ionospheric E-layers based on machine learning according to claim 2, wherein: The expression for calculating the ECI coordinates of the occultation tangent point when an occultation event occurs is: The expression for transforming the Earth-centered inertial coordinate system ECI into the Earth-centered Earth-fixed coordinate system ECEF is: The expressions for calculating the geodetic coordinates of the occultation tangent point based on the Earth-centered Earth-fixed coordinate system ECEF include: in, The ECI coordinates of the occultation tangent point when the occultation event occurs in vector form; and represents the ECI coordinates of the LEO satellite and the GPS satellite respectively; || represents the modulus of the vector; X, Y, Z represent the spatial rectangular coordinates of the occultation tangent point in the ECEF coordinate frame; B, L, H represent the geodetic coordinates of the occultation tangent point in the ECEF coordinate frame, which are the geodetic latitude, geodetic longitude, and altitude, respectively; a, e, and e′ represent the semi-major axis, the first eccentricity, and the second eccentricity of the Earth reference ellipsoid, respectively; N represents the radius of the occultation circle corresponding to the geodetic latitude of the occultation tangent point, and δ represents the Euler angle difference between the ECI coordinate system and the ECEF coordinate system caused by the Earth's rotation.

4. The method for identifying sporadic ionospheric E-layers based on machine learning according to claim 1, wherein: The process of calculating the signal-to-noise ratio disturbance sequence based on the coordinate data in the ionospheric sporadic E layer occurrence area includes: Extract the satellite's moisture-free atmospheric profile data based on the coordinate data within the ionospheric sporadic E layer occurrence area, and extract the signal-to-noise ratio data in the occultation event data based on the satellite's moisture-free atmospheric profile data. Use the sliding average method to average a continuous section of the signal-to-noise ratio data to obtain the background value of the signal-to-noise ratio data. The background value is subtracted from the original signal-to-noise ratio data to obtain a signal-to-noise ratio perturbation sequence, and the signal-to-noise ratio normalized standard deviation is calculated based on the signal-to-noise ratio perturbation sequence.

5. The method for identifying sporadic ionospheric E-layers based on machine learning according to claim 4, characterized in that: The process of calculating the normalized standard deviation of the signal-to-noise ratio according to the signal-to-noise ratio disturbance sequence includes: The signal-to-noise ratio disturbance sequence is normalized to extract the relevant characteristic parameters of the sporadic E layer in the ionosphere. The expression is: According to the relevant characteristic parameters of the sporadic E layer in the ionosphere, the normalized standard deviation of the signal-to-noise ratio is calculated using the sliding average method. The expression is: Among them, SNR nstd represents the normalized standard deviation of the signal-to-noise ratio, X i represents the i-th data after normalization of the signal-to-noise ratio disturbance sequence, represents the i-th data after the signal-to-noise ratio disturbance sequence is smoothed, and k represents the size of the sliding average window.

6. The method for identifying sporadic ionospheric E-layers based on machine learning according to claim 5, characterized in that: The process of calculating the amplitude scintillation index based on the coordinate data of the ionospheric sporadic E layer occurrence area includes: According to the coordinate data of the ionospheric sporadic E layer occurrence area, the signal intensity fluctuation data in the occultation event data is calculated, and the expression is: The amplitude flicker index S4 is calculated based on the signal intensity fluctuation data. The expression is: Among them, σ I Indicates the root mean square value of signal strength fluctuation, I represents the signal strength, Indicates the mean value of the signal strength, and <> indicates averaging the physical quantity within the preset number of samples.

7. The method for identifying sporadic ionospheric E-layers based on machine learning according to claim 6, characterized in that: The ionospheric sporadic E-layer identification model uses the signal-to-noise ratio disturbance sequence and amplitude scintillation index to identify the ionospheric sporadic E-layer. The process includes: The ionospheric sporadic E-layer identification model includes a forward long short-term memory unit and a reverse long short-term memory unit; Normalize the signal-to-noise ratio to the standard deviation SNR nstd The ionospheric sporadic E-layer recognition model is trained for several rounds using the training set, wherein: The forward long short-term memory unit is normalized according to the input signal-to-noise ratio (SNR). nstd Produce a positive hidden state with the amplitude flicker index S4; The reverse long short-term memory unit is normalized according to the input signal-to-noise ratio (SNR) nstd Producing an inverse hidden state with the amplitude flicker index S4; Merging the forward hidden state and the reverse hidden state to obtain a bidirectional hidden state, and performing ionospheric sporadic E-layer identification based on the bidirectional hidden state; The test set is used to test the results of the ionospheric sporadic E-layer recognition. A loss function is set to optimize the parameters of the ionospheric sporadic E-layer recognition model during the training process. When the training round ends or the loss function converges, the trained ionospheric sporadic E-layer recognition model is obtained. The trained ionospheric sporadic E layer recognition model is used to normalize the standard deviation of the signal-to-noise ratio (SNR). nstd The ionospheric sporadic E-layer event is identified by combining the amplitude scintillation index S4, and whether an ionospheric sporadic E-layer event occurs is confirmed based on the identification results.

8. The method for identifying sporadic ionospheric E-layers based on machine learning according to claim 7, characterized in that: The expression for merging the forward hidden state and the reverse hidden state is: in, represents the forward hidden state, Represents the inverse hidden state.

9. A machine learning-based ionospheric sporadic E-layer identification system, characterized by: The system comprises: A data receiving module is used to receive occultation event data in the atmospheric profile file transmitted by the satellite; A data processing module, used for calculating the tangent point coordinate data of the occultation event based on the occultation event data; The data processing module is further used to screen coordinate data within the ionospheric sporadic E layer occurrence area based on the tangent point coordinate data of the occultation event, so as to calculate the signal-to-noise ratio disturbance sequence and the amplitude scintillation index based on the coordinate data; Model building module, used to build the ionospheric sporadic E layer identification model; The ionospheric sporadic E-layer identification module is used to identify the ionospheric sporadic E-layer according to the signal-to-noise ratio disturbance sequence and the amplitude scintillation index of the ionospheric sporadic E-layer identification model, and confirm whether the ionospheric sporadic E-layer event occurs based on the identification results.

10. An electronic device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

Images