Satellite orbit prediction method, device and equipment based on multi-features

Abstract

The present application discloses a satellite orbit prediction method, device and equipment based on multivariate features. The method includes: determining the velocity position parameters of the satellite in the ICRS coordinate system at the prediction time based on the orbital mechanics model to obtain the first velocity position parameter; converting the first velocity position parameter into the ECEF coordinate system to obtain the second velocity position parameter; obtaining the time characteristics, satellite state characteristics and space environment characteristics of the current moment, wherein the satellite state characteristics are in the ECEF coordinate system and the space environment characteristics are in the geodetic coordinate system; inputting the input characteristics obtained by feature fusion of the time characteristics, satellite state characteristics, space environment characteristics and the second velocity position parameter into the orbit correction model; evaluating the error of the second velocity position parameter based on the orbit correction model to obtain the target error prediction value; adjusting the second velocity position parameter based on the target error prediction value to obtain the target velocity position parameter of the satellite at the prediction time.

Description

Technical Field

[0001] The present application relates to the field of space environment detection technology, in particular to the field of satellite detection technology, and in particular to a satellite orbit prediction method, device and equipment based on multi-features. Background Art

[0002] Ultra-low-orbit satellites operate in orbits between 150km and 300km above the Earth's surface. Due to the large number of satellites and the increasing impact of the atmospheric environment, ultra-low-orbit satellites are now facing challenges such as rapid orbital decay, increased collision risks, and increased difficulty in attitude control. Therefore, it is crucial to build a complete ultra-low-orbit support system to ensure the stable operation of satellites and efficient mission execution.

[0003] At present, when predicting the future orbital state of a satellite, it is usually necessary to use orbital mechanics models to calculate the satellite position and velocity, such as the Simplified General Perturbation Model 4 (SGP4), High Precision Orbit Propagator (HPOP), etc. However, the calculation accuracy of the satellite position and velocity by the existing orbital mechanics models is not enough to guarantee the safe and stable operation of ultra-low-orbit satellites. Therefore, how to improve the prediction accuracy of satellite positions has become an urgent problem to be solved. Summary of the invention

[0004] In view of this, the embodiments of the present application provide a satellite orbit prediction method, device, equipment, medium and product based on multi-features, which can use the orbit correction model to correct the speed and position parameter prediction values ​​output by the orbital mechanics model, obtain more accurate target speed and position parameters, and improve the prediction accuracy of the satellite position.

[0005] In a first aspect, an embodiment of the present application provides a satellite orbit prediction method based on multivariate features, the method comprising: when the satellite orbit elements at the current moment are input into the orbital mechanics model, based on the orbital mechanics model, determining the velocity and position parameters of the satellite at the prediction moment in the geocentric inertial rectangular ICRS coordinate system to obtain a first velocity and position parameter; converting the first velocity and position parameter into the earth-fixed rectangular ECEF coordinate system to obtain a second velocity and position parameter at the prediction moment; acquiring the time characteristics, satellite state characteristics and space environment characteristics corresponding to the current moment, wherein the satellite state characteristics are in the ECEF coordinate system and the space environment characteristics are in the earth coordinate system; performing feature fusion on the time characteristics, satellite state characteristics, space environment characteristics and the second velocity and position parameters at the current moment, and obtaining input characteristics, and inputting the input characteristics into the orbit correction model, wherein the orbit correction model is used to evaluate the error of the predicted value of the velocity and position parameters output by the orbital mechanics model; evaluating the error of the second velocity and position parameters corresponding to the prediction moment based on the orbit correction model to obtain a target error prediction value; adjusting the second velocity and position parameters based on the target error prediction value to obtain the target velocity and position parameters of the satellite at the prediction moment.

[0006] In a second aspect, an embodiment of the present application provides a satellite orbit prediction device based on multivariate features, the device comprising: a determination module for determining the velocity and position parameters of the satellite at the prediction time in the Earth-centered inertial rectangular ICRS coordinate system based on the orbital mechanics model when the satellite orbit elements at the current time are input into the orbital mechanics model, and obtaining a first velocity and position parameter; a coordinate system conversion module for converting the first velocity and position parameter into the Earth-fixed rectangular ECEF coordinate system to obtain a second velocity and position parameter at the prediction time; an acquisition module for acquiring the time characteristics, satellite state characteristics and space environment characteristics corresponding to the current time, wherein the satellite state characteristics are the space environment characteristics in the ECEF coordinate system. The spatial environment characteristics are in the geodetic coordinate system; a feature fusion module is used to obtain input features after feature fusion of the time characteristics, satellite state characteristics, space environment characteristics, and second velocity position parameters at the current moment, and input the input features into the orbit correction model, wherein the orbit correction model is used to evaluate the error of the predicted value of the velocity position parameter output by the orbital mechanics model; an error evaluation module is used to evaluate the error of the second velocity position parameter corresponding to the predicted moment based on the orbit correction model to obtain a target error prediction value; an error correction module is used to adjust the second velocity position parameter based on the target error prediction value to obtain the target velocity position parameter of the satellite at the predicted moment.

[0007] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, the steps of the satellite orbit prediction method based on multivariate features as in the first aspect are implemented.

[0008] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having computer program instructions stored thereon. When the computer program instructions are executed by a processor, the steps of the satellite orbit prediction method based on multi-features as in the first aspect are implemented.

[0009] In a fifth aspect, an embodiment of the present application provides a computer program product, which is stored in a non-volatile storage medium. When the computer program product is executed by a processor, it implements the steps of the satellite orbit prediction method based on multivariate features as in the first aspect.

[0010] In a sixth aspect, an embodiment of the present application provides a chip, which includes a processor and a communication interface, wherein the communication interface and the processor are coupled, and the processor is used to run programs or instructions to implement the steps of the satellite orbit prediction method based on multivariate features as in the first aspect.

[0011] The present application provides a satellite orbit prediction method, device, equipment, medium and product based on multi-features. Considering the accuracy of the predicted value output by the orbital mechanics model, which is affected by the satellite state and the satellite space environment, the present application designs a multi-feature and reasonable learning feature for the orbit correction model, including time features, satellite state features and space environment features. By integrating the multi-features, the input information of the model is enhanced, and the generalization ability of the orbit correction model for orbit error prediction is improved. Specifically, the present application represents the satellite state features in the ECEF coordinate system and the space environment features in the earth coordinate system, so that the input features of the model are all represented in the earth-fixed coordinate system. Since the predicted value (i.e., the first velocity position parameter) output by the orbital mechanics model is represented in the ICRS coordinate system, the present application performs a coordinate system conversion on it, converts it into the ECEF coordinate system, and obtains the second velocity position parameter also represented in the earth-fixed coordinate system. Based on this, the error of the second velocity position parameter is evaluated, and the target error prediction value also represented in the earth-fixed coordinate system can be obtained. In this way, the input features and output labels of the orbit correction model are uniformly represented in the Earth-fixed coordinate system. Therefore, it is possible to fit the errors between the predicted values ​​and the true values ​​output by the orbital mechanics model for different satellite states and different space environment states, analyze the performance of the predicted values ​​output by the orbital mechanics model in different prediction scenarios, realize global analysis and modeling of the prediction errors of the orbital mechanics model, obtain the target error prediction value for the predicted value, and use the target error prediction value to correct the second velocity position parameter (i.e., the predicted value) output by the orbital mechanics model to obtain a target velocity position parameter with higher accuracy, thereby improving the prediction accuracy of the satellite position at the prediction time. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical application of the embodiments of the present application, the drawings in the embodiments of the present application are briefly introduced below.

[0013] Figure 1 It is a flowchart of a satellite orbit prediction method based on multi-features provided in an embodiment of the present application;

[0014] Figure 2 is a flowchart of a satellite orbit prediction method based on multivariate features provided in another embodiment of the present application;

[0015] Figure 3 It is a structural schematic diagram of a satellite orbit prediction device based on multi-features provided in an embodiment of the present application;

[0016] Figure 4 It is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] The principles and spirit of the present application will be described below with reference to several exemplary embodiments. It should be understood that the purpose of providing these embodiments is to make the principles and spirit of the present application clearer and more thorough, so that those skilled in the art can better understand and implement the principles and spirit of the present application. The exemplary embodiments provided herein are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments herein, all other embodiments obtained by ordinary technicians of the art without creative work are within the scope of protection of this application.

[0018] In this document, terms such as first, second, third, etc. are only used to distinguish one entity (or operation) from another entity (or operation), but not to require or imply any order or relationship between these entities (or operations).

[0019] The following is a brief description of the concepts and technical terms that may be involved in the embodiments of the present application.

[0020] Orbital mechanics model: A high-precision orbital propagation model is established by considering the various real physical forces (perturbations) acting on the satellite. It is used to predict the position and velocity of low-orbit satellites. Common orbital mechanics models include the Simplified General Perturbation Model 4 (SGP4), the High Precision Orbit Propagator (HPOP), etc.

[0021] Feature Fusion: It is the process of combining data features from different sources to enhance the input information of the model. This can be achieved through simple concatenation, weighted combination or more complex methods such as dimensionality reduction, aiming to improve the model's understanding and prediction ability of the task.

[0022] Model fusion technology: Model fusion technology is an integrated learning method that improves the overall prediction performance by combining the prediction results of multiple models. This method can make up for the shortcomings of a single model and improve the accuracy and robustness of the model.

[0023] Basic prediction model: referred to as base model, each basic prediction model is trained and predicted independently of other models, and generates its own prediction results based on different learning algorithms or different data subsets. Different basic prediction models have different learning biases and capabilities, and they capture different patterns or features in the data.

[0024] ECI: Earth-Centered Inertial (ECI) coordinate system, which does not rotate with the Earth.

[0025] ICRS coordinate system: International Celestial Reference System, is the celestial reference system standard adopted by the International Astronomical Union (IAU). It is a geocentric inertial rectangular coordinate system with coordinate axes defined based on the background of the distant starry sky. This coordinate system is not affected by the rotation or revolution of the earth and is mainly used in astronomy and satellite orbit calculations.

[0026] ECEF coordinate system: Earth-Centered, Earth-Fixed coordinate system, is an Earth-fixed rectangular coordinate system with the origin at the center of the Earth and the coordinate axes pointing to fixed positions on the Earth. The X-axis of the coordinate system usually points to the Greenwich astronomical meridian, the Y-axis points to 90 degrees east longitude, and the Z-axis is consistent with the Earth's rotation axis. This coordinate system changes with the Earth's rotation and is fixed relative to the Earth.

[0027] Geodetic Coordinate System (WGS84): It is a geocentric coordinate system widely used internationally. Its full name is WorldGeodetic System 1984 Coordinate System. It is a geocentric coordinate system with its origin at the center of mass of the earth. It uses longitude and latitude to represent the position and is defined based on the reference ellipsoid (WGS84 ellipsoid). It also changes with the rotation of the earth and is fixed relative to the earth.

[0028] Solar radiation parameters: Solar Indices, including F10.7, S10.7, M10.7, and Y10.7. F10.7 data is daily frequency, and its historical value can be obtained from NOAA (or the Atmospheric Density Forecast Center). The forecast value for the next 40 days can be obtained from NOAA, or it can be predicted using the F10.7 forecast model; S10.7, Y10.7, and M10.7 are daily frequency, and the forecast value for the next 40 days can be obtained using the corresponding forecast model; M10.7 is data with an update frequency of 3h, and its historical value can be obtained using an algorithm, and the forecast value can be obtained using the corresponding forecast model.

[0029] F10.7: is an index used to describe the intensity of solar radiation. It represents the solar radio flux at a wavelength of 10.7 cm, and therefore can reflect the intensity of solar activity. F10.7 is used as a good representative of solar ultraviolet radiation. The measurement of F10.7 is the average value of the intensity of solar radio radiation measured within one hour on the ground at a 100 MHz frequency band centered at 2800 MHz, and a value is updated daily. F10.7 is closely related to the activity on the surface of the sun, and also has a long and short periodic variation law of 11 years and 27 days. It is an important index to characterize solar activity, and there is a strong correlation between geomagnetic field disturbances and the activity level of the sun.

[0030] S10.7 (EUV): This is the solar radiation parameter in the extreme ultraviolet band, representing the solar radiation in the 26-34 nanometer band. This radiation mainly comes from solar active regions.

[0031] M10.7: This is the solar radiation parameter in the far ultraviolet band, representing the solar radiation around 280 nm, which includes the photospheric continuum and chromospheric line radiation. M10.7 is an index used to describe solar activity, which is calculated by the ratio of the core to the wing of the spectral line of the Mg II 280 nm band. This ratio reflects the emission of the solar active region in the chromosphere and is an index that is theoretically independent of changes in instrument sensitivity. The M10.7 index is used as a good representative of the solar Far Ultraviolet (FUV) and Extreme Ultraviolet (EUV) radiation, which have a significant impact on the Earth's atmosphere, especially the thermosphere (located between about 85 km and 600 km above the Earth's surface).

[0032] Y10.7: This is a mixed solar index representing mainly solar X-rays and Lyman-alpha radiation. X-rays come mainly from the solar canopy, while Lyman-alpha radiation is produced in the upper chromosphere and transition region of the sun.

[0033] Geomagnetic parameters: Geomagnetic parameters, including ap, Ap, Ap* and Dst indexes. The ap index is predicted by the deep learning model with 24 steps (each step is 3 days); Ap is the daily geomagnetic parameter, which is obtained by averaging the 8 ap values ​​of the day; Ap* is the sliding geomagnetic parameter, which is obtained by averaging the ap values ​​of the first 8 prediction periods of the target prediction period; Dst is a geomagnetic parameter with an hourly frequency.

[0034] ap index: also known as the equivalent planetary amplitude of geomagnetic disturbance, the unit is 2nT, a new ap value is generated every 3 hours, and 8 ap indexes are generated every day in chronological order. According to relevant standards, an ap value greater than 39 corresponds to a magnetic storm period. The maximum value of the ap index is 400.

[0035] Ap index: It is an index of the global daily geomagnetic disturbance intensity, also known as the planetary equivalent daily amplitude. It is an indicator to measure the level of geomagnetic activity. A new Ap value is generated every day, which is the average of 8 AP values ​​every day. Therefore, the Ap index can only reflect the overall situation of geomagnetic disturbance throughout the day.

[0036] Ap* index: It is an index of the intensity of geomagnetic disturbances in a predicted period. It is a real-time indicator for measuring the level of geomagnetic activity. Compared with the Ap index, it is more real-time and can therefore more accurately evaluate the overall situation of geomagnetic disturbances in a certain predicted period.

[0037] Dst Index: This is the Disturbance Storm Time (Dst) index, which is used to indicate the strength of the ring current during a magnetospheric storm. The Dst index is determined from hourly measurements of the magnetic field at four points around the Earth's equator and is used to simulate global density changes during geomagnetic storms.

[0038] The time and space parameters include altitude, longitude and latitude, universal time, DOY, month, and local time (or mean solar time, which cancels the whole hour restriction).

[0039] In the related art, when predicting the position and velocity of a satellite, it is usually based on orbital mechanics models or data-driven orbit prediction methods.

[0040] Existing orbital mechanics models, such as SGP4 and HPOP, have the following defects when using orbital mechanics models to predict the position and velocity of satellites:

[0041] The prediction results of high-precision orbital mechanics models are highly sensitive to certain input parameters (such as gravitational field model, air resistance, etc.), and the inaccuracy of input parameters may lead to significant prediction errors; although the model tries to take into account multiple factors, it may still not be able to fully capture certain complex orbital dynamics (such as atmospheric disturbances, solar radiation pressure, etc.).

[0042] Data-driven orbit prediction methods refer to the use of a large number of two-line elements (TLE) and machine learning algorithms to fit the orbit, rather than relying on traditional force analysis to predict satellite orbits. Pure data-driven orbit prediction methods are highly dependent on the quality and quantity of historical data for accuracy. Noise or incomplete data may cause model performance to degrade. At the same time, they lack the support of physical principles and may not be able to explain the reasons behind the prediction results. Their generalization ability is also not guaranteed. The model is trained under certain orbital or environmental conditions, which has great limitations and cannot adapt well to orbit prediction under different conditions.

[0043] Therefore, no matter whether the orbital mechanics model or the data-driven orbit prediction method is used in the prior art, the defects existing therein have affected the calculation accuracy of the satellite position and velocity to varying degrees. The use of the satellite position and satellite velocity information with low accuracy can no longer guarantee the safe and stable on-orbit operation of ultra-low orbit satellites, and cannot meet the needs of the ultra-low orbit support system.

[0044] Based on this, the relevant technology proposes a hybrid prediction scheme that uses machine learning algorithms to correct the prediction error of the orbital mechanics model. For example, an orbital mechanics model (such as SGP4) is used as the basic model, and an artificial neural network (ANN) is used to predict the orbit prediction error of the orbital mechanics model, that is, the error of the predicted value output by the orbital mechanics model.

[0045] Among them, the input features (learning variables) of ANN include the following five items:

[0046] ① Prediction duration: the time interval from the current moment to the predicted moment.

[0047] ② Estimated state: The estimated state at the current moment, which is expressed in the satellite orbital elements (Orbital Elements, COE) and Earth-Centered Inertial (Earth-Centered Inertial, ECI) coordinate system.

[0048] ③Estimated drag coefficient: the estimated drag coefficient at the current moment.

[0049] ④ Ground station observation status: the maximum measured elevation angle η in the current trajectory, and the corresponding range ρ and azimuth α at that moment.

[0050] ⑤ Predicted state based on the estimated state at the current moment: The predicted state at the future moment based on the estimated state at the current moment, expressed in COE and ECI coordinates.

[0051] The output feature (target variable) of ANN is ⑤ the prediction error at the future time, expressed in the radial solar direction / inertial velocity / inertial direction RSW coordinate system.

[0052] The selection of input and output features in the above hybrid prediction scheme is determined through a trial-and-error process in order to capture and learn the potential relationship between the prediction error and the available parameters, thereby improving the accuracy of orbit prediction. However, this scheme has the following problems:

[0053] ① As an error correction model, the input features of ANN are not comprehensive enough and do not consider the impact of the space environment and major celestial bodies on the satellite;

[0054] ② The expression of ANN input features and output features is not uniform: the input satellite current position and the predicted position at the next moment are expressed in the ECI coordinate system, the ground station observation status is expressed in a specific ground coordinate system, and the predicted error is expressed in the RSW coordinate system. Therefore, the inconsistent representation of input features and output features in different coordinate systems will increase the difficulty of model training and reduce the interpretability of the model. The use of specific ground station observation indicators as input features reduces the versatility of the algorithm in business.

[0055] It can be seen from this that the above-mentioned hybrid prediction scheme that uses machine learning algorithms to correct the prediction errors of the orbital mechanics model is still unable to accurately calculate the satellite position and speed, and thus cannot guarantee the safe and stable on-orbit operation of ultra-low orbit satellites, and cannot meet the needs of the ultra-low orbit support system.

[0056] Based on this, the present application designs multiple and reasonable learning features for the orbit correction model, including time features, satellite state features, and space environment features. By integrating multiple features, the input information of the model is enhanced, and the generalization ability of the orbit correction model for orbit error prediction is improved. In addition, the input features and output labels of the orbit correction model are uniformly represented in the earth-fixed coordinate system, and the error between the predicted value and the true value output by the orbital mechanics model for different satellite states and different space environment states can be fitted, and the performance of the predicted value output by the orbital mechanics model in different prediction scenarios can be analyzed, so as to realize global analysis and modeling of the prediction error of the orbital mechanics model, obtain the target error prediction value for the predicted value, and use the target error prediction value to correct the second velocity position parameter (i.e., the predicted value) output by the orbital mechanics model, so as to obtain a target velocity position parameter with higher accuracy, thereby improving the prediction accuracy of the satellite position at the prediction time, so as to solve at least one of the above technical problems.

[0057] The satellite orbit prediction method based on multi-features provided in the embodiment of the present application is described in detail below through specific embodiments and their application scenarios in conjunction with the accompanying drawings.

[0058] Figure 1 It is a flow chart of a satellite orbit prediction method based on multi-features provided in one embodiment of the present application. The executor of the satellite orbit prediction method based on multi-features may be a satellite orbit prediction device in an ultra-low orbit support system.

[0059] The following takes the execution subject of the satellite orbit prediction method based on multiple features as a satellite orbit prediction device as an example to illustrate the satellite orbit prediction method based on multiple features of the present application. It should be noted that the above execution subjects and application scenarios do not constitute a limitation on the present application.

[0060] like Figure 1 As shown, the satellite orbit prediction method based on multivariate features provided in the embodiment of the present application may include steps 110 to 160.

[0061] Step 110, when the satellite orbital elements at the current moment are input into the orbital mechanics model, the velocity and position parameters of the satellite at the prediction moment in the geocentric inertial rectangular ICRS coordinate system are determined based on the orbital mechanics model to obtain the first velocity and position parameters;

[0062] Step 120, converting the first velocity position parameter into an earth-fixed rectangular ECEF coordinate system to obtain a second velocity position parameter at the prediction time;

[0063] Step 130, obtaining the time feature, satellite state feature and space environment feature corresponding to the current moment, wherein the satellite state feature is in the ECEF coordinate system and the space environment feature is in the geodetic coordinate system;

[0064] Step 140, fusing the time feature, satellite state feature, space environment feature, and second velocity position parameter at the current moment to obtain input features, and inputting the input features into the orbit correction model;

[0065] Step 150, evaluating the error of the second speed position parameter corresponding to the prediction time based on the track correction model to obtain a target error prediction value;

[0066] Step 160, adjusting the second velocity position parameter based on the target error prediction value to obtain the target velocity position parameter of the satellite at the prediction time.

[0067] The satellite orbit prediction method based on multivariate features provided in the embodiment of the present application designs multivariate and reasonable learning features for the orbit correction model, including time features, satellite state features, and space environment features. By integrating the multivariate features, the input information of the model is enhanced, and the generalization ability of the orbit correction model for orbit error prediction is improved. In addition, the input features and output labels of the orbit correction model are uniformly represented in the earth-fixed coordinate system, and the error between the predicted value and the true value output by the orbital mechanics model for different satellite states and different space environment states can be fitted, and the performance of the predicted value output by the orbital mechanics model in different prediction scenarios can be analyzed, so as to realize global analysis and modeling of the prediction error of the orbital mechanics model, obtain the target error prediction value for the predicted value, and use the target error prediction value to correct the second velocity position parameter (i.e., the predicted value) output by the orbital mechanics model, so as to obtain a target velocity position parameter with higher accuracy, thereby improving the prediction accuracy of the satellite position at the prediction time.

[0068] The specific implementation of the above steps will be described in detail below in conjunction with specific embodiments.

[0069] Involving step 110, when the satellite orbital elements at the current moment are input into the orbital mechanics model, the velocity and position parameters of the satellite at the predicted moment in the geocentric inertial rectangular ICRS coordinate system are determined based on the orbital mechanics model to obtain the first velocity and position parameters.

[0070] In step 110, the orbital mechanics model may be SGP4, HPOP, SGT4 or other models, which are not specifically limited in this application. The prediction time is a future time, after the current time, and there may be a preset period between the current time and the prediction time. The preset period can be preset according to specific needs, for example, 10s, 12s or other durations, which are not specifically limited in this application. The prediction frequency of the orbital mechanics model and the orbital correction model is the same, and both are predicted once every preset period.

[0071] In some embodiments, since the ultra-low orbit satellite can fly 1 degree in the speed direction every 15 seconds, in order to ensure the calculation accuracy, a preset cycle length can be set. Less than or equal to 15s. The world is divided into many regions, and different regions have different performances. Limit the size of the area. The smaller it is, the finer the particle size.

[0072] Satellite orbit elements are a set of six key parameters needed to describe the motion of a celestial body or spacecraft in its Kepler orbit, including semi-major axis, eccentricity, orbital inclination, right ascension of the ascending node, argument of perigee, and true anomaly. These parameters can accurately define the orbital position and velocity of an object in three-dimensional space.

[0073] The speed and position parameters of the present application include speed parameters and position parameters. By inputting the satellite orbit elements at the current moment into the orbital mechanics model, the orbital mechanics model can predict the speed and position parameters at future moments based on the satellite orbit elements, and represent the predicted speed and position parameters in the ICRS coordinate system to obtain the first speed and position parameters.

[0074] Step 120, converting the first velocity position parameter into an earth-fixed rectangular ECEF coordinate system to obtain a second velocity position parameter at the prediction time.

[0075] In step 120, the present application may use a professional mathematical library or Geographic Information System (GIS) software (such as PROJ, GDAL or MATLAB's geographic toolbox) to implement the coordinate system conversion of the velocity position parameters. The conversion process may also need to consider parameters such as the earth's flattening and reference ellipsoid model. Among them, PROJ is applied to the field of geographic information systems (GIS) and map projections, and generally refers to an open source library for coordinate conversion and map projection, supporting conversion between multiple map projection methods and coordinate reference systems (CRS). GDAL (Geospatial Data Abstraction Library) is an open source library for reading and writing raster and vector geospatial data formats. MATLAB (Matrix Laboratory) is a high-level programming language and environment for numerical and scientific computing.

[0076] Step 130, obtaining the time characteristics, satellite status characteristics and space environment characteristics corresponding to the current moment, wherein the satellite status characteristics are in the ECEF coordinate system and the space environment characteristics are in the geodetic coordinate system.

[0077] In step 130, the satellite status characteristics are represented in the ECEF coordinate system, and the space environment characteristics are represented in the geodetic coordinate system wgs84. Both the wgs84 and ECEF coordinate systems change with the rotation of the earth and are fixed relative to the earth. Therefore, the two are relatively stationary and both belong to earth-fixed coordinate systems.

[0078] The current time is , the prediction time is , The time features corresponding to the current time may include the current month, DOY (day of year), and the mean solar time of the longitude of the satellite at the current time. DOY refers to the day of the year at the current time.

[0079] In some embodiments of the present application, satellite status characteristics may include the mass of the satellite, the windward area, and the velocity and position parameters of the satellite at the current moment in the ECEF coordinate system, and space environment characteristics may include the satellite position at the current moment in the geodetic coordinate system, the relative position of major celestial bodies and the satellite, and the satellite's solar radiation parameters and geomagnetic parameters.

[0080] Specifically, the satellite state characteristics may include the mass m of the satellite, the frontal area s, and the velocity position parameters of the satellite in the ECEF coordinate system at the current moment. For the convenience of distinction, the velocity position may be the third velocity position parameter, which may specifically include three position component parameters ( , , ) and three velocity component parameters ( , , ).

[0081] In an embodiment of the present application, by converting the velocity position parameters of the satellite at the current moment in the input features into the ECEF coordinate system, the input features and output features of the orbit correction model can be unified to the earth-fixed coordinate system relatively fixed to the earth, reducing the difficulty of training the orbit correction model and enhancing the interpretability of the orbit correction model. In addition to time features and satellite state features, the input features can also include the relative positions of major celestial bodies (earth, sun, moon), solar radiation parameters and geomagnetic parameters, and all kinds of position features, time features and model prediction results are uniformly represented in the earth-fixed system, so that the orbit correction model can directly learn the relationship between the earth position, sun position, moon position and the prediction error of the orbital mechanics model. In this way, by inputting multivariate features into the orbit correction model, the orbit correction model can analyze and evaluate the prediction error of the orbital mechanics model under the influence of the current satellite state and the satellite space environment, and obtain the target error prediction value that can accurately correct the output prediction value of the orbital mechanics model.

[0082] In some embodiments of the present application, in addition to being represented in ECEF (Earth-fixed rectangular coordinate system), there is an alternative scheme for the second velocity position parameter, the third velocity position parameter, the predicted result of the orbital mechanics model, and the prediction error of the orbital correction model in the present application, that is, performing a coordinate transformation on the above parameters and converting them into the Earth-fixed polar coordinate system, so that they are all represented in the Earth-fixed polar coordinate system. The position velocity expression in the Earth-fixed polar coordinate system has a more direct correlation with other features represented in the geodetic coordinate system.

[0083] In some embodiments of the present application, the satellite position may include the longitude, latitude and altitude of the satellite, and the relative position of the main celestial body and the satellite includes the longitude and latitude of the point where the sun is directly above the earth, the relative longitude and latitude difference between the satellite and the sun, the longitude and latitude of the point where the moon is directly above the earth, and the relative longitude and latitude difference between the satellite and the moon.

[0084] Specifically, the space environment characteristics can include the following items: the longitude of the satellite ,latitude and altitude; the longitude of the point where the sun shines directly on the earth and latitude , and the relative longitude and latitude difference between the satellite and the sun and ; The longitude of the point where the moon shines directly on the earth and latitude , and the relative longitude and latitude differences between the satellite and the moon and ; Solar radiation parameters F10.7, M10.7, S10.7, Y10.7; Geomagnetic parameters ap, Ap*.

[0085] The longitude and latitude of the satellite can be calculated from the satellite orbital roots. The longitude and latitude of the sun and moon at a specific time can also be directly calculated using professional astronomical software. The relative longitude and latitude difference calculation formulas (1), (2), (3), and (4) can be shown as follows:

[0086] (1)

[0087] (2)

[0088] (3)

[0089] (4)

[0090] In some embodiments, the astronomical software may be, for example, Astropy, which is a Python software package for astronomical data analysis and processing. It is an open source project jointly promoted by the community and aims to provide a powerful astronomy and astrophysics core toolkit for the Python programming environment.

[0091] Step 140, the time feature, satellite state feature, space environment feature, and second velocity position parameter of the current moment are integrated to obtain input feature, and the input feature is input into the orbit correction model.

[0092] In step 140, the time features, satellite state features, space environment features, and second velocity position parameters at the current moment can be spliced ​​by feature fusion technology to obtain input features. The orbit correction model is pre-trained, and the orbit correction model is used to evaluate the error of the predicted value of the velocity position parameter output by the orbital mechanics model. The error is relative to the true value. During the training of the orbit correction model based on historical training data, the true value can be obtained by actually detecting the satellite through the satellite detection system. The above-mentioned first velocity position parameter and second velocity position parameter are both output by the orbital mechanics model, and are the predicted values ​​of the velocity position parameters at the predicted moment. The difference between the two is that they are represented in different coordinate systems.

[0093] It should be noted that the present application can input the input features after feature fusion into the orbit correction model, or can input the time features, satellite status features, space environment features, and second velocity position parameters of the current moment into the orbit correction model respectively, and perform feature fusion of various features through the orbit correction model.

[0094] In some embodiments of the present application, in order to train an orbit correction model that can accurately evaluate the prediction error of the orbital mechanics model based on the current satellite state and the satellite space environment state, before inputting the input features into the orbit correction model in the above step 140, the following steps may also be included:

[0095] Acquire multiple first-moment time characteristics, satellite state characteristics, space environment characteristics, and satellite orbit elements;

[0096] Inputting the satellite orbital elements at the first moment into the orbital mechanics model to obtain the predicted values ​​of the velocity and position parameters at the second moment output by the orbital mechanics model, wherein the first moment and the second moment are separated by a preset period of time;

[0097] Combining the time characteristics, satellite state characteristics, space environment characteristics of each first moment, and the predicted value and true value of the velocity position parameters at the second moment, historical training data is constructed;

[0098] The preset model is trained based on historical training data to obtain an orbit correction model associated with the orbital mechanics model.

[0099] Specifically, the first moment and the second moment are two adjacent historical prediction moments, and the second moment is after the first moment. The true value of the velocity position parameter at the second moment can be obtained by actually detecting the satellite through the satellite detection system. The mean square error can be used as the loss function during model training, and the back propagation algorithm is used for training. The orbit correction model is associated with the orbital mechanics model. The preset model can be set according to specific needs, and this application does not make specific limitations on this.

[0100] The above-mentioned training of the preset model based on historical training data may specifically include:

[0101] The time characteristics of the first moment, the satellite state characteristics, the space environment characteristics, and the predicted values ​​of the speed and position parameters at the second moment are used as input training data, and the difference between the predicted value and the true value of the speed and position parameters at the second moment is used as output training data (i.e., output label) for training.

[0102] Among them, the difference between the predicted value and the true value of the speed position parameter at the second moment can be the true error value. The mean square error between the training error prediction value output by the preset model and the true error value is calculated to obtain the loss function value.

[0103] It should be noted that in the historical training data, the satellite state characteristics are in the ECEF coordinate system, the space environment characteristics are in the geodetic coordinate system, and the predicted value and true value of the velocity and position parameters at the second moment are both in the ECEF coordinate system, ensuring that the historical training data are uniformly represented in the earth-fixed coordinate system.

[0104] In the embodiment of the present application, during the training process of the orbit correction model, the training data used is derived from the predicted value actually output by the orbital mechanics model. Based on this, during the training process, the orbit correction model can learn the deviation between the predicted value output by the orbital mechanics model under different time characteristics, different satellite state characteristics, and different space environment characteristics and the true value, so that the trained orbit correction model can have the ability to reasonably and accurately evaluate the error of the predicted value output by the orbital mechanics model in different satellite orbit prediction scenarios by combining time characteristics, satellite state characteristics and space environment characteristics, and obtain an orbit correction model adapted to the orbital mechanics model, so that the orbit correction model can have a better prediction error evaluation capability for the associated orbital mechanics model. In addition, for different orbital mechanics models, an associated orbit correction model can be trained for each orbital mechanics model. The correction capability of the orbit correction model trained for a specific orbital mechanics model is only effective for the orbital mechanics model. For example, the correction model of HPOP is basically ineffective for the error correction of SGP4. This can make the adaptability and correlation between the orbital mechanics model and the orbit correction model stronger, and further enhance the prediction error evaluation capability of the orbit correction model for the associated orbital mechanics model.

[0105] Step 150 , evaluating the error of the second speed position parameter corresponding to the prediction time based on the orbit correction model to obtain a target error prediction value.

[0106] In step 150, the orbit correction model can evaluate the error of the second velocity position parameter corresponding to the predicted time based on the time characteristics, satellite state characteristics and space environment characteristics corresponding to the current time, and obtain a target error prediction value.

[0107] Step 160, adjusting the second velocity position parameter based on the target error prediction value to obtain the target velocity position parameter of the satellite at the prediction time.

[0108] In step 160, the target error prediction value is also represented in the ECEF coordinate system, and the target error prediction value can be added to the second speed position parameter to obtain the target speed position parameter at the prediction time. The target error prediction value can include a speed error prediction value and a position error prediction value.

[0109] Exemplarily, the second speed position parameter includes a speed parameter and positional parameters The orbit correction model outputs the predicted value of the satellite's velocity error in the ECEF coordinate system and the predicted position error , then the second speed position parameter and the prediction result of the correction model are combined to calculate the corrected target speed position parameter according to formulas (5) and (6). The target speed position parameter includes the speed parameter and positional parameters :

[0110] (5)

[0111] (6)

[0112] Finally and Converted to the final forecast value in ICRS coordinate system and Make a report.

[0113] In some embodiments of the present application, the second speed position parameter and the target speed position parameter both include position parameters and speed parameters, the position parameter includes three position component parameters on the x, y, and z axes, and the speed parameter includes three speed component parameters on the x, y, and z axes, so the second speed position parameter and the target speed position parameter both include three position component parameters and three speed component parameters. Accordingly, the speed error prediction value includes three speed component error prediction values ​​on the x, y, and z axes, and the position error prediction value includes three position component error prediction values ​​on the x, y, and z axes.

[0114] Based on this, in order to improve the accuracy of the orbit correction model's error assessment of the predicted values ​​output by the orbital mechanics model, Figure 2It is a flowchart of a satellite orbit prediction method based on multi-features provided in another embodiment of the present application. Before the input features are input into the orbit correction model in the above step 140, the method may further include steps 210 and 220. The above step 140 may specifically include step 230.

[0115] Step 210, obtaining six sub-fusion models corresponding to the six component parameters, wherein different sub-fusion models are used to evaluate the errors of the corresponding component parameters;

[0116] Step 220, constructing a track correction model by combining the six sub-fusion models;

[0117] Step 230 , based on the six sub-fusion models, respectively evaluate the errors of the six component parameters in the second velocity position parameter to obtain a target error prediction value including the six component error prediction values.

[0118] Among them, the six component error prediction values ​​include three speed component error prediction values ​​and three position component error prediction values, and the six component error prediction values ​​are used to adjust the six component parameters in the second speed position parameter respectively to obtain the target speed position parameter.

[0119] For example, the prediction time is , the current time is , the satellite is The third position parameter of the moment contains three position component parameters ( , , ) and three velocity component parameters ( , , ); Satellite in The second velocity position parameter at the moment contains three position component parameters ( , , ), and three velocity component parameters ( , , .

[0120] Based on this, the input features of the orbit correction model can include: satellite state features: m, s, , , , , , ; Second speed position parameters: , , , , , ; Non-temporal spatial environment characteristics: , , , , , , , , , , ap and Ap* at the moment (one value each); spatial environmental characteristics of the time series: daily frequencies F10.7, M10.7, S10.7, Y10.7, Ap (25 values) of the previous 5 days; time characteristics: month, DOY, and mean solar time of the longitude of the satellite.

[0121] The predicted values ​​of the three velocity components output by the orbit correction model are ( , , ), the three position component error prediction values ​​are ( , , ), so the six component parameters in the second velocity position parameter can be adjusted by these six component error prediction values, specifically by adding each component parameter to the corresponding component error prediction value, , , , , , , the six component parameters of the target speed position parameters can be obtained, and the correction of the second speed position parameters can be completed.

[0122] In this way, the present application trains a sub-fusion model for each component parameter in a targeted manner, so that the sub-fusion model has the ability to perform error evaluation on a single component parameter, and is more targeted and adaptable. Therefore, the accuracy of the orbit correction model's error evaluation of the predicted value output by the orbit mechanics model can be improved.

[0123] In some embodiments of the present application, before the above step 210 acquires the six sub-fusion models corresponding to the six component parameters, the following steps may also be included:

[0124] Obtain historical training data, wherein the historical training data includes multiple sets of data sets, each set of data sets includes time characteristics, satellite state characteristics, space environment characteristics at a first moment, and predicted values ​​and true values ​​of velocity and position parameters at a second moment, the predicted values ​​are output by an orbital mechanics model, and the true values ​​are obtained by actual detection;

[0125] Train the sub-fusion model corresponding to each component parameter based on historical training data;

[0126] Specifically, the satellite state characteristics in the historical training data are in the ECEF coordinate system, the space environment characteristics are in the geodetic coordinate system, and the predicted value and true value of the velocity and position parameters at the second moment are both in the ECEF coordinate system.

[0127] The training of the sub-fusion model corresponding to each component parameter based on the historical training data may specifically include:

[0128] The time characteristics, satellite state characteristics, space environment characteristics of the first moment, and the predicted values ​​of the velocity and position parameters at the second moment are used as input training data, and the difference between the predicted value and the true value of the corresponding component parameters of the velocity and position parameters at the second moment is used as output training data (i.e., output label) for training.

[0129] It should be noted that the present application has already described the historical training data, the first moment and the second moment in detail in the above embodiments, and for the sake of brevity, they will not be repeated here.

[0130] In some embodiments of the present application, the training of the sub-fusion model corresponding to each component parameter based on the historical training data may specifically include:

[0131] Divide multiple data sets into training data sets and test data sets according to multi-fold cross validation;

[0132] Combined with the training data set and the test data set, the sub-fusion model corresponding to each component parameter is trained separately.

[0133] Exemplarily, multiple data sets can be divided into two parts: a training data set and a test data set, with a ratio of 5:2 or 5:3. The training data set is divided into five parts of training data and validation data with a ratio of 4:1 according to a 5-fold cross validation.

[0134] In some embodiments of the present application, in order to further improve the accuracy of the error assessment of the predicted values ​​output by the orbital mechanics model by the orbit correction model, each sub-fusion model is composed of multiple base models and a target fusion model. Each base model is used to evaluate the deviation between the predicted value and the true value of the corresponding component parameter in the speed position parameter at the prediction moment, and obtain the predicted component error associated with the corresponding component parameter. The target fusion model is used to fuse multiple predicted component errors output by multiple base models to obtain the component error prediction value output by the sub-fusion model.

[0135] Specifically, the six sub-fusion models have the same structure, and the multiple base models may include at least two of lightGBM, random forest algorithm, support vector machine SVM algorithm, long short-term memory network LSTM, and encoder transformer encoder.

[0136] Multiple base models are trained based on historical training data. Specifically, the time characteristics, satellite state characteristics, space environment characteristics of the first moment, and the predicted values ​​of the velocity and position parameters at the second moment are used as input training data, and the difference between the predicted value and the true value of the corresponding component parameters of the velocity and position parameters at the second moment is used as output training data (i.e., output label) for training.

[0137] The target fusion model is used to fuse the output results of multiple base models. For example, a linear regression algorithm can be used to perform weighted summation on the output results of multiple base models to obtain target speed position parameters.

[0138] Thus, compared with other hybrid prediction schemes, this application uses ensemble learning technology, and simultaneously performs multi-feature fusion and model fusion, amplifying the advantages of various machine learning models and reducing their disadvantages. Each sub-fusion model uses multiple basic models to fuse multiple input features (space environment features, satellite state features, and time features) to predict the prediction error under a certain component parameter of the orbital mechanics model, and uses a meta-model (i.e., target fusion model) to further fuse the prediction results of multiple basic models and multi-input features to obtain the final prediction error prediction. In this way, a variety of machine learning algorithms can be used to fit the error between the prediction results and the measured values ​​of a certain orbital mechanics model for satellites in different states in different environments, making up for the shortcomings of a single model, so that the overall model has stronger generalization ability and robustness. The prediction results of multiple machine learning models are combined to improve the overall error evaluation performance of the orbit correction model, and when facing different satellites in different space environment conditions, the overall prediction error when predicting their orbits is reduced.

[0139] Moreover, in the present application, the degree of coupling between the various models is very low (modular), so the models can be flexibly replaced, increased or decreased according to actual needs. For example, the orbital mechanics model can be replaced, and the machine learning model (base model) involved in error prediction can be replaced, increased or decreased. It has high scalability and high practicality.

[0140] In order to explain the sub-fusion model in this application in detail, HPOP is taken as the orbital mechanics model, lightGBM, random forest algorithm, SVM, and multi-layer fully connected network are selected as base models, and the sub-fusion model corresponding component parameters are taken as the velocity component parameters on the x-axis as an example to explain the sub-fusion model in detail.

[0141] 1. The first base model trained using lightGBM

[0142] The input features of the model include (input features do not need to be scaled and normalized):

[0143] ① Satellite status characteristics: m, s, , , , , , ;

[0144] ② Orbital mechanics model prediction results: , , , , , ;

[0145] ③Non-temporal environmental characteristics: , , , , , , , , , , ap and Ap* at the moment (one value each);

[0146] ④ Environmental characteristics of the time series: daily frequencies of F10.7, M10.7, S10.7, Y107, and Ap (25 values) in the first 5 days;

[0147] ⑤ Time characteristics: month, DOY, and mean solar time of the satellite’s longitude.

[0148] When training the model, the hyperparameter objective is set to regression and boosting_type is set to gbdt. Other hyperparameters can be used to obtain the best combination through grid search, which may include but are not limited to: n_estimators, max_depth, feature_fraction, min_data_in_leaf, bagging_fraction.

[0149] The output characteristics of the model are the velocity error prediction values ​​of the velocity component parameters of the satellite on the x-axis predicted by the first base model for HPOP .

[0150] 2. The second base model trained using the random forest algorithm

[0151] The input features of the model are the same as those in lightGBM. The hyperparameters usually involved in tuning are n_estimators, max_depth, min_sample_split, min_sample_leaf, max_leaf_nodes, etc. The output features of the model are the velocity error prediction values ​​of the velocity component parameters of the satellite on the x-axis predicted by the second base model for HPOP .

[0152] 3. The third base model obtained by SVM training

[0153] The input features of the model are the same as those in lightGBM, but the input features need to be scaled and normalized before entering the model. The Standard Scaling method can be used to standardize the features to a mean of 0 and a standard deviation of 1.

[0154] The kernel function of the model is rbf. During model training, the regularization parameter C, epsilon, and kernel function coefficient gamma can be used to find the best combination through hyperparameter search. The loss function uses mean square error.

[0155] The model output is the velocity error prediction value of the satellite velocity component parameters on the x-axis predicted by the third base model for HPOP .

[0156] 4. The fourth base model trained using LSTM

[0157] The input features of the model are similar to those in lightGBM, with two differences: first, the temporal environmental features ④ are changed to the 3h frequency F10.7, M10.7, S10.7, Y10.7, and ap value of the previous 5 days (40 steps in total); second, the ap and Ap* of the day are removed from the non-temporal environmental features ③. The input features need to be scaled and normalized before entering the model.

[0158] The 40-step time series environment characteristics (using Represents) loop input lstm, the output of the last hidden layer after the 40th round is H, , where the last round of output H can be regarded as a high-level feature extracted from the original temporal environment features. H is combined with the non-temporal environment features ( ) and input a fully connected layer. The output dimension of the fully connected layer is 1, and the output value is the velocity error prediction value of the velocity component parameter of the satellite on the x-axis predicted by the fourth basis model for HPOP , .

[0159] The mean square error is used as the loss function during model training, and the meta-model is trained using the back propagation algorithm. Some examples of hyperparameters of the model are as follows: the length of the LSTM sequence is 40, the hidden layer depth is 1 or 2, and the hidden layer dimension is 32 or 64.

[0160] 5. The fifth base model trained using transformer encoder

[0161] The input features and overall structure of the model are the same as those of the fourth basic model, except that the neural network for extracting time series features is replaced by a transformer encoder instead of an LSTM. Similarly, 40 steps of continuous time series features with a step length of 3 hours are taken as 40 tokens and input into the transformer encoder model. The original vector of each token is composed of the corresponding time After the normalization operation, the original vector is converted into an embedding vector of dimension d through a fully connected network. After the embedding vectors of all tokens pass through several layers of encoder, 40 output vectors of dimension d are output. Each output vector is connected to a fully connected layer and the dimension is converted to 1, and 40 high-level time series features are extracted. The 40 high-level time series features are merged with the non-time series features and input into a fully connected layer (FC). The output dimension of the fully connected layer is 1, and the output value is the velocity error prediction value of the velocity component parameters of the satellite on the x-axis predicted by the fifth base model for HPOP .

[0162] The meta-model is trained using the back propagation algorithm and the mean square error as the loss function. The following examples are given for some of the model's hyperparameters: the transformer encoder time series length is 40, the number of encoder layers is 1 or 2, the embedding vector dimension is 32 or 64, and the number of multi-head attention heads is 4 or 8.

[0163] 6. Target Fusion Model

[0164] A linear regression algorithm is used to integrate the prediction results of the five base models, and the final prediction result is obtained by weighted summation based on formula (7): :

[0165] (7)

[0166] Among them, w and d are parameters of the linear regression model, w corresponds to the weight of each base model, and d is the overall deviation term.

[0167] The training method uses the prediction errors of various models from a large amount of historical data. As input features, the true prediction error is used as the label. The mean square error (MSE) is used as the error, and the gradient descent method is used to update the model parameters.

[0168] In this way, the prediction results of the five base models can be effectively fused through the above five base models and a target fusion model, thereby achieving an accurate and reasonable evaluation of the velocity error prediction value of the velocity component parameters of the satellite on the x-axis predicted by HPOP.

[0169] Corresponding to the method embodiment of the present application, the present application also provides a satellite orbit prediction device based on multi-features.

[0170] Figure 3 is a schematic diagram of the structure of a satellite orbit prediction device based on multi-features provided in an embodiment of the present application. Figure 3 As shown, the satellite orbit prediction device 300 based on multivariate features may include: a determination module 310 , a coordinate system conversion module 320 , an acquisition module 330 , a feature fusion module 340 , an error assessment module 350 , and an error correction module 360 ​​.

[0171] Among them, the determination module 310 is used to determine the speed and position parameters of the satellite at the predicted time in the Earth-centered inertial rectangular ICRS coordinate system based on the orbital mechanics model when the satellite orbital elements at the current time are input into the orbital mechanics model, and obtain the first speed and position parameters; the coordinate system conversion module 320 is used to convert the first speed and position parameters into the Earth-fixed rectangular ECEF coordinate system to obtain the second speed and position parameters at the predicted time; the acquisition module 330 is used to obtain the time characteristics, satellite state characteristics and space environment characteristics corresponding to the current time, wherein the satellite state characteristics are in the ECEF coordinate system, and the space environment characteristics are in the earth coordinate system; feature fusion Module 340 is used to obtain input features after feature fusion of the time features, satellite status features, space environment features, and second velocity position parameters at the current moment, and input the input features into the orbit correction model, wherein the orbit correction model is used to evaluate the error of the predicted value of the velocity position parameter output by the orbital mechanics model; the error evaluation module 350 is used to evaluate the error of the second velocity position parameter corresponding to the predicted moment based on the orbit correction model to obtain the target error prediction value; the error correction module 360 ​​is used to adjust the second velocity position parameter based on the target error prediction value to obtain the target velocity position parameter of the satellite at the predicted moment.

[0172] The satellite orbit prediction device based on multivariate features provided in the embodiment of the present application designs multivariate and reasonable learning features for the orbit correction model, including time features, satellite state features, and space environment features. By integrating the multivariate features, the input information of the model is enhanced, and the generalization ability of the orbit correction model for orbit error prediction is improved. In addition, the input features and output labels of the orbit correction model are uniformly represented in the earth-fixed coordinate system, and the error between the predicted value and the true value output by the orbital mechanics model for different satellite states and different space environment states can be fitted, and the performance of the predicted value output by the orbital mechanics model in different prediction scenarios can be analyzed, so as to realize global analysis and modeling of the prediction error of the orbital mechanics model, obtain the target error prediction value for the predicted value, and use the target error prediction value to correct the second velocity position parameter (i.e., the predicted value) output by the orbital mechanics model, so as to obtain a target velocity position parameter with higher accuracy, thereby improving the prediction accuracy of the satellite position at the prediction time.

[0173] In some embodiments of the present application, satellite status characteristics include the mass of the satellite, the windward area, and the velocity and position parameters of the satellite at the current moment in the ECEF coordinate system, and space environment characteristics include the satellite position at the current moment in the geodetic coordinate system, the relative position of major celestial bodies and the satellite, and the satellite's solar radiation parameters and geomagnetic parameters.

[0174] In some embodiments of the present application, the satellite position includes the longitude, latitude and altitude of the satellite, and the relative position of the main celestial body and the satellite includes the longitude and latitude of the point where the sun is directly above the earth, the relative longitude and latitude difference between the satellite and the sun, the longitude and latitude of the point where the moon is directly above the earth, and the relative longitude and latitude difference between the satellite and the moon.

[0175] In some embodiments of the present application, the second speed position parameter and the target speed position parameter both include three position component parameters and three speed component parameters, and also include: an acquisition module, used to acquire six sub-fusion models corresponding to the six component parameters before inputting the input features into the track correction model, wherein different sub-fusion models are used to evaluate the errors of the corresponding component parameters; a model construction module, used to construct a track correction model in combination with the six sub-fusion models; the error evaluation module is specifically used to: based on the six sub-fusion models, respectively evaluate the errors of the six component parameters in the second speed position parameter, and obtain a target error prediction value containing the six component error prediction values, wherein the six component error prediction values ​​are used to adjust the six component parameters in the second speed position parameter respectively to obtain the target speed position parameter.

[0176] In some embodiments of the present application, it also includes: an acquisition module, which is also used to acquire historical training data before acquiring the six sub-fusion models corresponding to the six component parameters, wherein the historical training data includes multiple sets of data sets, each set of data sets includes time characteristics, satellite state characteristics, space environment characteristics at a first moment, and predicted values ​​and true values ​​of speed and position parameters at a second moment, the predicted values ​​are output by the orbital mechanics model, and the true values ​​are obtained by actual detection; a model training module, which is used to train the sub-fusion model corresponding to each component parameter based on the historical training data; wherein the satellite state characteristics in the historical training data are in the ECEF coordinate system, the space environment characteristics are in the geodetic coordinate system, and the predicted values ​​and true values ​​of the speed and position parameters at the second moment are both in the ECEF coordinate system.

[0177] In some embodiments of the present application, each sub-fusion model is composed of multiple base models and a target fusion model, each base model is used to evaluate the deviation between the predicted value and the true value of the corresponding component parameter in the speed position parameter at the prediction moment, and obtain the predicted component error associated with the corresponding component parameter, and the target fusion model is used to fuse multiple predicted component errors output by multiple base models to obtain the component error prediction value output by the sub-fusion model; the multiple base models include at least two of lightGBM, random forest algorithm, support vector machine SVM algorithm, long short-term memory network LSTM, and encoder transformer encoder.

[0178] In some embodiments of the present application, the model training module is specifically used to: divide multiple groups of data sets into training data sets and test data sets according to multi-fold cross-validation; combine the training data sets and the test data sets to train the sub-fusion model corresponding to each component parameter respectively.

[0179] In some embodiments of the present application, the orbit correction model is associated with the orbital mechanics model and also includes: an acquisition module, which is used to acquire multiple time characteristics, satellite state characteristics, space environment characteristics and satellite orbit elements at the first moment before inputting the input characteristics into the orbit correction model; an input-output module, which is used to input the satellite orbit elements at the first moment into the orbital mechanics model to obtain the predicted value of the velocity and position parameters for the second moment output by the orbital mechanics model, wherein the first moment and the second moment are separated by a preset period of time; a data construction module, which is used to construct historical training data by combining the time characteristics, satellite state characteristics, space environment characteristics of each first moment, and the predicted value and true value of the velocity and position parameters for the second moment; a model training module, which is used to train the preset model based on the historical training data to obtain an orbit correction model associated with the orbital mechanics model.

[0180] The satellite orbit prediction device based on multi-features provided in the embodiment of the present application can achieve Figure 1-2 The various processes implemented by the service platform in the method embodiment can achieve the same technical effect. To avoid repetition, they will not be described here.

[0181] Figure 4 It is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application.

[0182] like Figure 4 As shown, the electronic device 400 includes a memory 401 , a processor 402 , and a computer program stored in the memory 401 and executable on the processor 402 .

[0183] In an example, the processor 402 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0184] The memory 401 may include a read-only memory (ROM), a random access memory (RAM), a disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Therefore, generally, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the satellite orbit prediction method based on multivariate features in the embodiment of the first aspect of the present application.

[0185] The processor 402 runs a computer program corresponding to the executable program code by reading the executable program code stored in the memory 401, so as to implement the satellite orbit prediction method based on multivariate features in the embodiment of the first aspect mentioned above.

[0186] In some examples, the electronic device 400 may further include a communication interface 403 and a bus 410. Figure 4 As shown, the memory 401, the processor 402, and the communication interface 403 are connected via a bus 410 and communicate with each other.

[0187] The communication interface 403 is mainly used to realize the communication between the modules, devices, units and / or equipment in the embodiment of the present application. The input device and / or output device can also be accessed through the communication interface 403.

[0188] The bus 410 includes hardware, software, or both, coupling the components of the electronic device 400 to each other. By way of example and not limitation, the bus 410 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a Memory Bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-E) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses or a combination of two or more of these. Where appropriate, the bus 410 may include one or more buses. Although embodiments of the present application describe and illustrate a particular bus, the present application contemplates any suitable bus or interconnect.

[0189] The electronic device provided in the embodiment of the present application can realize Figure 1-2 The various processes implemented by the electronic device in the method embodiment can achieve the same technical effect, and to avoid repetition, they will not be described here.

[0190] In combination with the satellite orbit prediction method based on multi-features in the above embodiments, the present application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when the computer program instructions are executed by a processor, the steps of any one of the satellite orbit prediction methods based on multi-features in the above embodiments are implemented.

[0191] In combination with the satellite orbit prediction method based on multivariate features in the above embodiment, the embodiment of the present application can provide a computer program product for implementation. The (computer) program product is stored in a non-volatile storage medium, and when the program product is executed by at least one processor, it implements the steps of any one of the satellite orbit prediction methods based on multivariate features in the above embodiment.

[0192] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface and the processor are coupled, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned satellite orbit prediction method embodiment based on multiple features, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0193] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0194] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0195] The functional blocks shown in the structural block diagram described above can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier. "Machine-readable medium" may include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROMs, flash memories, erasable ROMs (EROMs), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0196] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0197] Aspects of the present disclosure are described above with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the functions / actions specified in one or more boxes of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It can also be understood that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can also be implemented by dedicated hardware that performs the specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0198] The above is only a specific implementation of the present application. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. It should be understood that the protection scope of the present application is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in this application, and these modifications or replacements should be included in the protection scope of this application.

Claims

1. A satellite orbit prediction method based on multivariate features, characterized in that: include: When the satellite orbital elements at the current moment are input into the orbital mechanics model, the velocity and position parameters of the satellite at the predicted moment in the geocentric inertial rectangular ICRS coordinate system are determined based on the orbital mechanics model to obtain the first velocity and position parameters; Converting the first velocity position parameter into an earth-fixed rectangular ECEF coordinate system to obtain a second velocity position parameter at a prediction time; Obtain the time characteristics, satellite status characteristics and space environment characteristics corresponding to the current moment, wherein the satellite status characteristics are represented in the ECEF coordinate system, and the space environment characteristics are represented in the geodetic coordinate system. The time characteristics include the month and DOY at the current moment, and the mean solar time of the longitude of the satellite at the current moment. DOY refers to the day of the year at the current moment. The satellite status characteristics include the mass and windward area of ​​the satellite, and the speed and position parameters of the satellite at the current moment in the ECEF coordinate system. The space environment characteristics include the satellite position at the current moment in the geodetic coordinate system, the relative position of the main celestial body and the satellite, and the solar radiation parameters and geomagnetic parameters of the satellite; Performing feature fusion on the time feature at the current moment, the satellite state feature, the space environment feature, and the second velocity position parameter to obtain input features, and inputting the input features into an orbit correction model, wherein the orbit correction model is used to evaluate the error of the predicted value of the velocity position parameter output by the orbital mechanics model; Evaluate the error of the second speed position parameter corresponding to the prediction time based on the trajectory correction model to obtain a target error prediction value; The second speed position parameter is adjusted based on the target error prediction value to obtain the target speed position parameter of the satellite at the prediction time.

2. The method according to claim 1, characterized in that The satellite position includes the longitude, latitude and altitude of the satellite, and the relative position of the main celestial body and the satellite includes the longitude and latitude of the point where the sun directly shines on the earth, the relative longitude and latitude difference between the satellite and the sun, the longitude and latitude of the point where the moon directly shines on the earth, and the relative longitude and latitude difference between the satellite and the moon.

3. The method according to claim 1, characterized in that The second speed position parameter and the target speed position parameter both include three position component parameters and three speed component parameters. Before the input feature is input into the track correction model, the method further includes: Obtain six sub-fusion models corresponding to the six component parameters, wherein different sub-fusion models are used to evaluate errors of corresponding component parameters; Combining the six sub-fusion models to construct the orbit correction model; The error of the second speed position parameter corresponding to the prediction time is evaluated based on the track correction model to obtain a target error prediction value, including: Based on the six sub-fusion models, the errors of the six component parameters in the second speed position parameters are evaluated respectively to obtain a target error prediction value including six component error prediction values, wherein the six component error prediction values ​​are used to adjust the six component parameters in the second speed position parameters respectively to obtain the target speed position parameters.

4. The method according to claim 3, characterized in that Before obtaining the six sub-fusion models corresponding to the six component parameters, it also includes: Acquire historical training data, wherein the historical training data includes multiple sets of data sets, each set of data sets includes time characteristics, satellite state characteristics, space environment characteristics at a first moment, and predicted values ​​and true values ​​of velocity and position parameters at a second moment, the predicted values ​​are output by an orbital mechanics model, and the true values ​​are obtained by actual detection; Training the sub-fusion model corresponding to each component parameter based on the historical training data; Among them, the satellite state characteristics in the historical training data are in the ECEF coordinate system, the space environment characteristics are in the geodetic coordinate system, and the predicted value and true value of the velocity and position parameters at the second moment are both in the ECEF coordinate system.

5. The method according to claim 3 or 4, characterized in that: Each sub-fusion model is composed of multiple base models and a target fusion model. Each base model is used to evaluate the deviation between the predicted value and the true value of the corresponding component parameter in the speed position parameter at the prediction moment to obtain the predicted component error associated with the corresponding component parameter. The target fusion model is used to fuse multiple predicted component errors output by multiple base models to obtain the component error prediction value output by the sub-fusion model. The multiple base models include at least two of lightGBM, random forest algorithm, support vector machine SVM algorithm, long short-term memory network LSTM, and encoder transformer encoder.

6. The method according to claim 4, characterized in that The training of the sub-fusion model corresponding to each component parameter based on the historical training data includes: Dividing the multiple data sets into training data sets and test data sets according to multi-fold cross validation; The sub-fusion model corresponding to each component parameter is trained respectively in combination with the training data set and the test data set.

7. The method according to claim 1, characterized in that The track correction model is associated with the track mechanics model, and before the input feature is input into the track correction model, the method further includes: Acquire multiple first-moment time characteristics, satellite state characteristics, space environment characteristics, and satellite orbit elements; Inputting the satellite orbit elements at the first moment into the orbital mechanics model to obtain the predicted value of the velocity position parameter output by the orbital mechanics model for the second moment, wherein the first moment and the second moment are separated by a preset period of time; Combining the time characteristics, satellite state characteristics, space environment characteristics of each first moment, and the predicted value and true value of the velocity position parameters at the second moment, historical training data is constructed; The preset model is trained based on the historical training data to obtain a track correction model associated with the track mechanics model.

8. A satellite orbit prediction device based on multivariate features, characterized in that: include: A determination module is used to determine the velocity position parameters of the satellite at the prediction time in the geocentric inertial rectangular ICRS coordinate system based on the orbital mechanics model when the satellite orbit elements at the current time are input into the orbital mechanics model, so as to obtain the first velocity position parameters; A coordinate system conversion module, used for converting the first velocity position parameter into an earth-fixed rectangular ECEF coordinate system to obtain a second velocity position parameter at a prediction time; An acquisition module, used to acquire the time characteristics, satellite status characteristics and space environment characteristics corresponding to the current moment, wherein the satellite status characteristics are in the ECEF coordinate system, and the space environment characteristics are in the geodetic coordinate system. The time characteristics include the month and DOY at the current moment, and the mean solar time of the longitude of the satellite at the current moment. DOY refers to the day of the year at the current moment. The satellite status characteristics include the mass and windward area of ​​the satellite, and the speed and position parameters of the satellite at the current moment in the ECEF coordinate system. The space environment characteristics include the satellite position at the current moment in the geodetic coordinate system, the relative position of the main celestial body and the satellite, and the solar radiation parameters and geomagnetic parameters of the satellite; A feature fusion module is used to obtain input features after feature fusion of the time features, satellite state features, space environment features, and the second velocity position parameters at the current moment, and input the input features into an orbit correction model, wherein the orbit correction model is used to evaluate the error of the predicted value of the velocity position parameters output by the orbital mechanics model; An error evaluation module, used to evaluate the error of the second speed position parameter corresponding to the prediction time based on the track correction model to obtain a target error prediction value; The error correction module is used to adjust the second speed position parameter based on the target error prediction value to obtain the target speed position parameter of the satellite at the prediction time.

9. An electronic device, characterized in that: The electronic device comprises: a processor and a memory storing computer program instructions; when the electronic device executes the computer program instructions, the method according to any one of claims 1 to 7 is implemented.

Images