Orbit prediction method, prediction device, electronic equipment and medium for space target

Abstract

Embodiments of the present application provide a space target orbit prediction method, a prediction device, an electronic device and a medium. The method comprises: obtaining two-line orbital elements of a space target at a current time, space environment data and perturbation coefficients, the perturbation coefficients being coefficients in a space target orbit prediction model; correcting the two-line orbital elements and / or the perturbation coefficients based on a pre-trained parameter correction model, the two-line orbital elements, the perturbation coefficients and the space environment data, to obtain corrected data, the corrected data comprising corrected two-line orbital elements and / or corrected perturbation coefficients; and predicting orbit information of the space target at a preset future time based on the space target orbit prediction model and the corrected data. The method makes the predicted orbit information more accurate.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and in particular to a method, device, electronic equipment, and medium for predicting the orbit of a space target. Background Technology

[0002] With the rapid development of artificial intelligence technology, its application in predicting the orbits of space targets is of great significance.

[0003] In related technologies, when predicting the orbit of a space target, the detected current two-line elements (TLE) are typically input into the Simplified General Perturbations 4 (SGP4) model. The predicted orbit information of the space target is then obtained through SGP4. Furthermore, an error compensation model is used to compensate for errors in the predicted orbit information output by SGP4, yielding the final orbit information.

[0004] However, the orbital information of space targets obtained by the above-mentioned technologies is not accurate enough and contains a large error. Summary of the Invention

[0005] This application provides a method, device, electronic device, and medium for predicting the orbit of a space target, thereby improving the accuracy of predicting the orbit information of a space target.

[0006] In a first aspect, embodiments of this application provide a method for predicting the orbit of a space target, comprising:

[0007] The system obtains the two orbital elements, space environment data, and perturbation coefficients of the space target at the current moment, wherein the perturbation coefficients are coefficients in the space target orbit prediction model.

[0008] Based on the pre-trained parameter correction model, the two rows of orbital elements, the perturbation coefficients, and the space environment data, the two rows of orbital elements and / or the perturbation coefficients are corrected to obtain corrected data, which includes the corrected two rows of orbital elements and / or the corrected perturbation coefficients.

[0009] Based on the space target orbit prediction model and the corrected data, the orbit information of the space target at a preset future time is predicted.

[0010] In one possible implementation, the pre-trained parameter correction model includes a pre-trained first neural network model;

[0011] The method involves correcting the two rows of orbital elements and / or the perturbation coefficients based on a pre-trained parameter correction model, the two rows of orbital elements, the perturbation coefficients, and the spatial environment data to obtain corrected data, including:

[0012] From the two rows of orbital elements and / or the perturbation coefficient, at least one first parameter to be corrected and at least one first auxiliary parameter are determined, wherein the first auxiliary parameter is the remaining parameter among the multiple parameters excluding the first parameter to be corrected.

[0013] The first parameter to be corrected, the first auxiliary parameter, and the space environment data are input into the first neural network model to obtain the first corrected parameter;

[0014] Replace the first parameter to be corrected with the first corrected parameter to obtain the first corrected data.

[0015] In one possible implementation, the pre-trained parameter correction model includes a pre-trained second neural network model;

[0016] The method involves correcting the two rows of orbital elements and / or the perturbation coefficients based on a pre-trained parameter correction model, the two rows of orbital elements, the perturbation coefficients, and the spatial environment data to obtain corrected data, including:

[0017] From the two rows of track elements and / or the preset perturbation coefficient, at least one second parameter to be corrected and at least one second auxiliary parameter are determined. The second auxiliary parameter is the remaining parameter among the multiple parameters other than the second parameter to be corrected.

[0018] The second parameter to be corrected, the second auxiliary parameter, and the space environment data are input into the second neural network model to obtain the residual value of the second parameter to be corrected.

[0019] The residual value is added to the original value of the second parameter to be corrected to obtain the second corrected parameter.

[0020] Replace the second parameter to be corrected with the second corrected parameter to obtain the second corrected data.

[0021] In one possible implementation, the space target orbit prediction model is the Simplified General Perturbation Model, Version 4 (SGP4).

[0022] The prediction of the orbit information of the space target at a preset future time based on the space target orbit prediction model and the corrected data includes:

[0023] The corrected data is input into the fourth version of the simplified general perturbation model to predict the orbital information of the space target at a preset future time. The orbital information includes displacement information and velocity information.

[0024] In one possible implementation, the pre-trained ensemble model includes: the pre-trained parameter correction model and the space target orbit prediction model.

[0025] In one possible implementation, the training steps of the ensemble model include:

[0026] Acquire training data, which includes the historical two-line orbital elements, perturbation coefficients, space environment data, and measured orbital information of space targets that match the historical two-line orbital elements;

[0027] The initial parameter correction model in the initial ensemble model is trained using the training data until the preset convergence condition is met, and the trained ensemble model is obtained. The initial ensemble model includes the initial parameter correction model and the fourth version of the initial simplified general perturbation model.

[0028] In one possible implementation, the space target orbit prediction model is the Simplified General Perturbation Model, Version 4 (SGP4).

[0029] Before acquiring the training data, the method further includes:

[0030] Obtain the fourth version of the initial simplified general perturbation model and the initial parameter correction model;

[0031] The parameters of the initial simplified general perturbation model (version 4) are represented by tensor quantization to form the transformed simplified general perturbation model (version 4).

[0032] The initial parameter correction model is fused with the fourth version of the transformed simplified general perturbation model to obtain the initial integrated model.

[0033] Secondly, embodiments of this application provide a space target orbit prediction device, including: an acquisition module, used to acquire the two rows of orbital elements, space environment data and perturbation coefficients of the space target at the current time, wherein the perturbation coefficients are coefficients in the space target orbit prediction model;

[0034] The correction module is used to correct the number of the two rows of orbital elements and / or the perturbation coefficients based on a pre-trained parameter correction model, the number of the two rows of orbital elements, the perturbation coefficients, and the spatial environment data, to obtain corrected data, wherein the corrected data includes the corrected number of the two rows of orbital elements and / or the corrected preset perturbation coefficients.

[0035] The prediction module is used to predict the orbit information of the space target at a preset future time based on the space target orbit prediction model and the corrected data.

[0036] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0037] The memory stores computer-executed instructions;

[0038] The processor executes computer execution instructions stored in the memory, causing the processor to perform various possible implementations of the first aspect described above.

[0039] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement various possible implementations of the first aspect above.

[0040] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements various possible implementations of the first aspect described above.

[0041] The embodiments of this application provide a method, device, electronic device, and medium for predicting the orbit of a space target. The method includes: acquiring two rows of orbital elements, space environment data, and perturbation coefficients of the space target at the current moment, wherein the perturbation coefficients are coefficients in the space target orbit prediction model, and correcting the two rows of orbital elements and / or perturbation coefficients based on a pre-trained parameter correction model, the two rows of orbital elements, the perturbation coefficients, and the space environment data to obtain corrected data, and then predicting the orbit information of the space target at a preset future moment based on the space target orbit prediction model and the corrected data. Attached Figure Description

[0042] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0043] Figure 1 A schematic diagram illustrating the format of the two-line track elements provided in this application;

[0044] Figure 2 A flowchart illustrating a method for predicting the orbit of a space target, provided in an embodiment of this application;

[0045] Figure 3A A flowchart illustrating a method for obtaining corrected data provided in an embodiment of this application;

[0046] Figure 3BAn architecture diagram of a parameter correction model provided in an embodiment of this application;

[0047] Figure 3C An architectural diagram of yet another parameter correction model provided in an embodiment of this application;

[0048] Figure 4A A flowchart illustrating another method for obtaining corrected data provided in an embodiment of this application;

[0049] Figure 4B An architectural diagram of another parameter correction model provided in an embodiment of this application;

[0050] Figure 4C An architecture diagram of another parameter correction model provided in the embodiments of this application;

[0051] Figure 5A A flowchart illustrating a training method for an ensemble model provided in an embodiment of this application;

[0052] Figure 5B A schematic diagram of a training process provided in an embodiment of this application;

[0053] Figure 6A This application provides a bar chart of position prediction error obtained solely through the SGP4 model in the prior art;

[0054] Figure 6B This application provides a histogram of velocity prediction error obtained solely through the SGP4 model in the prior art.

[0055] Figure 6C A location prediction error comparison histogram provided for embodiments of this application;

[0056] Figure 6D A speed prediction error comparison histogram provided for embodiments of this application;

[0057] Figure 7 A schematic diagram of the structure of a space target trajectory prediction device provided in an embodiment of this application;

[0058] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0059] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0060] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0061] In the description of the embodiments of this application, the terms "inner" and "outer", etc., which indicate the direction or positional relationship, are based on the direction or positional relationship shown in the drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or component must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this application.

[0062] In the description of the embodiments of this application, unless otherwise expressly specified and limited, the terms "connected" and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the embodiments of this application according to the specific circumstances.

[0063] Below, we will first explain the terms used in this application:

[0064] Space targets refer to objects such as satellites, space stations, booster rockets entering space orbit, and protective shields.

[0065] Perturbations: These refer to deviations in the orbit of a space object caused by the gravitational pull of other celestial bodies or other factors. Perturbations can lead to changes in the object's coordinates, velocity, or orbital parameters.

[0066] Two-line elements (TLE) refer to a set of elements generated based on general perturbation theory to predict the position and velocity of Earth-orbiting spacecraft. They are used to determine the orbital elements of a space target orbiting the Earth at a given epoch. They have essentially become an international standard and are widely used to determine the position of space targets.

[0067] The data format consists of two or three lines of text information. The first line is optional and usually contains the name of the space target. The following two lines contain orbital elements. These mainly include the space target's orbital eccentricity, orbital inclination, argument of perigee, right ascension of the ascending node, mean perigee angle, first-order time derivative of the mean motion, and atmospheric drag ballistic coefficients. An example of a two-line orbital element format is shown below. Figure 1 As shown.

[0068] Figure 1 The interpretation of the data is shown in Table 1.

[0069] Table 1

[0070]

[0071]

[0072]

[0073] The SGP4 model, also known as the dynamic model, is a sophisticated algorithmic model developed in 1970 specifically for calculating the trajectories of low-Earth orbit satellites. It is a simplification of the extensive analytical theory of Lane and Cranford (1969), specifically designed to simulate and predict Earth satellite orbits based on simplified perturbation theory, incorporating the effects of atmospheric drag, solar and lunar gravitational influences. Its applications are extremely wide-ranging, assisting in space target orbit prediction, astronomical observation support, and collision warning missions.

[0074] With the rapid development of aerospace technology, the number of space targets in Earth orbit has increased significantly, including satellites, space stations, and space debris. This trend makes orbit prediction particularly important. Accurate orbit prediction can effectively reduce potential collision risks, thereby improving the safe operation of space targets. Applying artificial intelligence technology to the field of space target orbit prediction is of great significance for the research and development of advanced orbit prediction methods.

[0075] In related technologies, the TLE (Time-to-Earning Error) of a space target is typically input into SGP4 for calculation based on the target's current or historical TLE, predicting the target's orbital information. Then, a pre-trained error compensation model is used to compensate for the errors in the predicted orbital information output by SGP4, yielding the final orbital information. The training process for the pre-trained error compensation model involves obtaining training samples, which consist of historical orbital information predicted by SGP4 and the corresponding actual orbital information. These training samples are then input into the error compensation model for error compensation, yielding the error-compensated orbital information. The training process is then evaluated to determine if a preset convergence condition is met; error compensation models that meet the convergence condition are considered successfully trained error compensation models.

[0076] However, the above methods, because TLE corresponds to the instantaneous orbital coordinate system, have orbital prediction algorithms derived from general perturbation theory. These algorithms not only include coordinate system rotation bias but also model errors, resulting in inaccurate orbital elements in the two rows. The correlation constants used in SGP4 suffer from insufficient precision and dynamic variations. Due to the complexity and high dynamism of the space environment, it is difficult to accurately model various perturbation forces, leading to significant errors in the perturbation coefficients in SGP4, which also affect the accuracy of orbital prediction. Although error compensation models can compensate for errors in the predicted orbital information output by SGP4, the significant discrepancy between the predicted results and the actual values ​​makes this compensation insufficient. Ultimately, the orbital accuracy obtained after error compensation based on the SGP4-predicted orbital information remains inaccurate. Therefore, the orbital information of space targets obtained using the above-mentioned existing technologies is of poor accuracy.

[0077] Therefore, addressing the aforementioned technical problems in the prior art, the inventors discovered during their research that the accuracy of the orbital elements in the two rows before inputting into SGP4 is poor due to inaccurate data or inaccurate perturbation coefficients in SGP4. Thus, by pre-correcting the data input into SGP4, the accuracy of the input data can be improved, leading to more accurate orbital information predicted based on the more accurate input data. Furthermore, research revealed that space environment data refers to parameters related to the space environment, which influences the perturbation terms of space targets. These perturbation terms affect the perturbation coefficients in the SGP4 model. Therefore, space environment data needs to be considered during correction. Specifically, a pre-trained parameter correction model can be used to correct the acquired two rows of orbital elements and / or perturbation coefficients to obtain more accurate corrected data. This corrected data is then used to predict the orbit based on the space target orbit prediction model, effectively improving the accuracy of orbit prediction. The space target orbit prediction model can be SGP4. Based on this, this application proposes a method, device, electronic device, and medium for predicting the orbit of a space target.

[0078] It should be noted that this application is mainly used for orbit prediction of space targets operating in low Earth orbit. Low Earth orbit refers to an orbit with an operating period of less than 225 minutes (operating altitude of approximately 5875 km).

[0079] The application scenario of the method in this application can be an orbit prediction device with a space target. This device can receive an activation command sent by a user through a client and begin periodic predictions based on the command. By executing the orbit prediction method of this application, it obtains the predicted orbit information of the space target. After obtaining the predicted orbit information, it can analyze the possibility of collisions between satellites, either independently or through interaction with other devices, and monitor the actual launch orbit of the satellite to analyze its deviation path. It is understood that the above scenarios are for illustrative purposes only and do not impose limitations on this application.

[0080] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0081] Figure 2 A flowchart illustrating a method for predicting the orbit of a space target provided in an embodiment of this application is shown below. Figure 2 As shown, the executing entity of this method can be a space target orbit prediction device, which can be implemented through a computer program; it can also be implemented through a medium storing the relevant computer program, such as a USB flash drive and / or optical disc; or it can be implemented through a physical device integrating or installing the relevant computer program, such as a chip or electronic device. The electronic device can be a server, server cluster, smart terminal, or other electronic device. The method includes:

[0082] S201. Obtain the two orbital elements, space environment data, and perturbation coefficients of the space target at the current moment. The perturbation coefficients are the coefficients in the space target orbit prediction model.

[0083] In this embodiment, space targets include, but are not limited to, artificial satellites, space stations, spacecraft, and space debris.

[0084] Among them, the two rows of orbital elements can include the atmospheric drag ballistic coefficient B of the space target. * The parameters include the perigee argument ω0, the first time derivative of the average motion n0, the average orbital inclination i0, the orbital eccentricity e0, the mean perigee angle M0, and the right ascension of the ascending node Ω0. Other data in Table 1 may also be included; however, the comparison in this embodiment is not limited.

[0085] The space environment data includes the geomagnetic index KP, planetary equivalent amplitude (AP), international sunspot number (ISN), solar 10.7 cm radio flux observation (F10.7_OBS), and the average solar 10.7 cm radio flux over the past 81 days (F10.7_LAST8).

[0086] The perturbation coefficient can include the j2 perturbation rate and the j2 perturbation rate, etc.

[0087] Optionally, the current number of two orbital elements can be obtained by accessing a website or platform that can store real-time two-line orbital elements, and the spatial environment data can be obtained from a database that stores spatial environment data. The perturbation coefficient can be obtained by the user through client input.

[0088] S202. Based on the pre-trained parameter correction model, the number of orbital elements in the two rows, the perturbation coefficients, and the space environment data, the number of orbital elements in the two rows and / or the perturbation coefficients are corrected to obtain the corrected data.

[0089] The corrected data includes the corrected number of two orbital elements and / or the corrected perturbation coefficients.

[0090] Optionally, the pre-trained parameter correction model can be a neural network model.

[0091] The acquired two rows of orbital elements, perturbation coefficients, and space environment data are input into the trained parameter correction model. The parameters to be corrected are indicated to the parameter correction model. If the parameters to be corrected include the two rows of orbital elements, the parameter correction model corrects the two rows of orbital elements. If the parameters to be corrected include the perturbation coefficients, the parameter correction model corrects the perturbation coefficients. If the parameters to be corrected include both the two rows of orbital elements and the perturbation coefficients, the parameter correction model corrects both the two rows of orbital elements and the perturbation coefficients to obtain the corrected two rows of orbital elements.

[0092] In this embodiment, the space target orbit prediction model can be the Simplified General Perturbation Model, Version 4 (SGP4).

[0093] S203. Based on the space target orbit prediction model and the corrected data, predict the orbit information of the space target at a preset future time.

[0094] SGP4 can be applied to near-Earth objects with orbital periods of less than 225 minutes. It simulates and predicts the orbits of Earth satellites based on simplified perturbation theory, and combines factors such as atmospheric drag, solar and lunar gravitational influences to predict the orbits of space targets.

[0095] In this embodiment, the corrected data is input into the Simplified General Perturbation Model, Version 4 (SGP4), to predict the orbital information of a space target at a preset future time. The orbital information includes displacement and velocity information.

[0096] One possible approach to using the Simplified Universal Perturbation Model 4 (SGP4) for prediction is as follows: After obtaining the corrected data, SGP4 analyzes the data and initializes itself based on the analyzed data. This initialization process involves calculating constants and parameters used in subsequent orbit integration and prediction. A preset timeframe and time step are determined; typically, the time step can be adjusted based on required accuracy and computational resources. Starting from the initial timeframe, time is progressively advanced. At each time step, SGP4 calculates the orbital position and velocity of the space target based on the current orbital state and time. Simultaneously, at each time step, SGP4 considers the influence of various perturbation factors on the orbit, including but not limited to the Earth's non-spherical gravitational field, atmospheric drag, and gravitational perturbations from the Sun and Moon. The final output is the orbital parameters of the space target at the predicted future time, including the position and velocity vectors.

[0097] In the above embodiments of this application, the orbital elements, perturbation coefficients, and space environment data of a space target at the current moment are obtained. The orbital elements and / or perturbation coefficients are then corrected based on a pre-trained parameter correction model to obtain corrected data. Finally, based on the space target orbit prediction model and the corrected data, the orbital information of the space target at a preset future time is predicted. In this embodiment, the pre-trained parameter correction model corrects the orbital elements, and the perturbation coefficients are corrected by considering the influence of space environment data on the perturbation coefficients. This makes the data input to the space target orbit prediction model more accurate, thereby making the orbital information of the space target predicted based on the corrected data more accurate.

[0098] Furthermore, based on the above embodiments, the process of correcting the two orbital elements and / or perturbation coefficients based on a pre-trained parameter correction model to obtain the corrected data is described.

[0099] Please see Figure 3A , Figure 3A The flowchart illustrates a method for obtaining corrected data according to an embodiment of this application, wherein the pre-trained parameter correction model includes a pre-trained first neural network model, and the method includes the following steps:

[0100] S301. Determine at least one first parameter to be corrected and at least one first auxiliary parameter from among the multiple parameters included in the two-line orbital elements and / or perturbation coefficients.

[0101] The first parameter to be corrected is the parameter that needs to be corrected by the first neural network model in this embodiment. The parameter is one or more of the two orbital elements and the perturbation coefficient.

[0102] Among them, the first auxiliary parameter is the remaining parameter among the multiple parameters excluding the first parameter to be corrected.

[0103] Please see Figure 3B , Figure 3B This application provides an architecture diagram of a parameter correction model, based on two rows of track elements [B]. * From the multiple parameters including [j1,j2,j3,...] and / or perturbation coefficients [j2,j3,...], determine the parameter to be corrected. Auxiliary parameters are Among them, auxiliary parameters are the remaining parameters excluding the parameters to be corrected.

[0104] Optionally, in response to a user's instruction to determine the parameter to be corrected triggered on the client, two rows of orbital elements and perturbation coefficients are displayed; and in response to a user's selection operation on the client of the two rows of orbital elements and / or perturbation coefficients, the selected two rows of orbital elements and / or perturbation coefficients are determined as the parameter to be corrected.

[0105] S302. Input the first parameter to be corrected, the first auxiliary parameter, and the space environment data into the first neural network model to obtain the first corrected parameter.

[0106] The first neural network model is a trained model that can be used to correct the input parameters to obtain the corrected parameters. During its training, the input training data includes two rows of orbital data and multiple parameters from the perturbation coefficients. The labeled data is at least one accurate value of the parameter to be corrected, which can be the accurate value of the parameter in the corresponding two rows of orbital data or the accurate value of the parameter in the corresponding perturbation coefficients.

[0107] The parameters to be corrected and auxiliary parameters The space environment parameters [KP, AP, ISN, F10.7_OBS, F10.7_LAST81] are input into the first neural network model to obtain the corrected parameters [x′1,…x′]. N ] and other parameters that do not need to be corrected.

[0108] S303. Replace the first parameter to be corrected with the first corrected parameter to obtain the first corrected data.

[0109] The corrected parameters are replaced with the corrected parameters to obtain the corrected data. The corrected data includes the corrected parameters and other data that does not need to be corrected, including auxiliary parameters and space environment data.

[0110] like Figure 3B As shown, the modified data is input into SGP4, and the final output is the displacement vector of the spatial target in three preset directions (x, y, and z) at a preset future time Δt. x, Position y, Position z, The velocity vectors of the target in space at a predetermined future time Δt in three predetermined directions: x, y, and z. [Velocity] x, Velocity y, Velocity z, ].

[0111] Figure 3B The first neural network model shown can be a multi-parameter correction neural network model, meaning that a multi-parameter correction neural network model can simultaneously correct multiple parameters to be corrected. For a multi-parameter correction neural network model, during training, the input training data includes two rows of orbital data and multiple parameters from the perturbation coefficients. The labeled data are the accurate values ​​of the multiple parameters to be corrected, which can be the accurate values ​​of the parameters in the corresponding two rows of orbital data or the accurate values ​​of the parameters in the corresponding perturbation coefficients.

[0112] Optionally, in this embodiment, the first neural network model can also be a module composed of multiple single-parameter correction neural network models, wherein each single-parameter correction neural network model can correct one parameter to be corrected, and the module composed of multiple single-parameter correction neural network models replaces... Figure 3B The multi-parameter correction neural network model in the middle, thus obtaining Figure 3C Please refer to the architecture diagram shown. Figure 3C , Figure 3C This is an architectural diagram of another parameter correction model provided in an embodiment of this application. For a single-parameter correction neural network model, during training, the input training data includes two rows of orbital data and multiple parameters from the perturbation coefficients. The labeled data is an accurate value of a parameter to be corrected, which can be the accurate value of the parameter in the corresponding two rows of orbital data or the accurate value of the parameter in the corresponding perturbation coefficients. Different single-parameter correction neural network models use different types of labeled parameters during training.

[0113] In the above embodiments of this application, at least one first parameter to be corrected and at least one first auxiliary parameter are determined from multiple parameters including the two rows of orbital elements and / or perturbation coefficients. The first parameter to be corrected, the first auxiliary parameter, and space environment data are input into a first neural network model to obtain a first corrected parameter. The first corrected parameter is then used to replace the first parameter to be corrected to obtain the first corrected data. The method of this embodiment corrects the determined parameter to be corrected by a trained first neural network model. Since the trained first neural network model can accurately learn the relationship between the first parameter to be corrected, the first auxiliary parameter, the space environment data, and the first corrected parameter, the corrected parameter is closer to the truth than the uncorrected parameter, reducing the error and thus obtaining more accurate corrected data. Consequently, the orbital information of the space target predicted based on the corrected data is more accurate.

[0114] The following describes another process for correcting the two orbital elements and / or perturbation coefficients based on a pre-trained parameter correction model to obtain the corrected data.

[0115] Please see Figure 4A , Figure 4A A flowchart illustrating another method for obtaining corrected data provided in this application embodiment, wherein the pre-trained parameter correction model includes a pre-trained second neural network model, and the method includes the following steps:

[0116] S401. Determine at least one second parameter to be corrected and at least one second auxiliary parameter from among the multiple parameters included in the two-line orbital elements and / or perturbation coefficients.

[0117] The second parameter to be corrected is the parameter that needs to be corrected by the second neural network model in this embodiment. The parameter is one or more of the two orbital elements and the perturbation coefficient.

[0118] The second auxiliary parameter is the remaining parameter among the multiple parameters, excluding the second parameter to be corrected.

[0119] Please see Figure 4B , Figure 4B This application provides an embodiment of another parameter correction model, which determines the parameter to be corrected from multiple parameters including the two rows of orbital elements and / or perturbation coefficients. Auxiliary parameters are Among them, auxiliary parameters are the remaining parameters excluding the parameters to be corrected.

[0120] Optionally, in response to a user's instruction to determine the parameter to be corrected triggered on the client, two rows of orbital elements and perturbation coefficients are displayed; and in response to a user's selection operation on the client of the two rows of orbital elements and / or perturbation coefficients, the selected two rows of orbital elements and / or perturbation coefficients are determined as the parameter to be corrected.

[0121] S402. Input the second parameter to be corrected, the second auxiliary parameter, and the space environment data into the second neural network model to obtain the residual value of the second parameter to be corrected.

[0122] The second neural network model is a trained model that can be used to correct the input parameters to obtain the corrected parameters. During its training, the input training data includes two rows of orbital data and multiple parameters from the perturbation coefficients. The labeled data is the accurate residual value of at least one parameter to be corrected. This accurate residual value can be the accurate residual value of the parameter in the corresponding two rows of orbital data or the accurate residual value of the parameter in the corresponding perturbation coefficients.

[0123] The parameters to be corrected And auxiliary parameters are The space environment parameters [KP, AP, ISN, F10.7_OBS, F10.7_LAST81] are input into the second neural network model to obtain the residual values ​​[x′1,…x′] of the parameters to be corrected. N ], and other data that does not require correction.

[0124] S403. Add the residual value to the original value of the second parameter to be corrected to obtain the second corrected parameter.

[0125] The residual values ​​of the parameters to be corrected [x′1,…x′] N ] and the original values ​​[x1,…,x N Add them together to get the second corrected parameter.

[0126] S404. Replace the second parameter to be corrected with the second corrected parameter to obtain the second corrected data.

[0127] The corrected parameters are replaced with the corrected parameters to obtain the corrected data. The corrected data includes the corrected parameters and other data that does not need to be corrected, including auxiliary parameters and space environment data.

[0128] like Figure 4B As shown, the corrected data is input into SGP4, and the final output is the displacement vector of the spatial target in three preset directions (x, y, and z) at a preset future time Δt. x, Positiony, Position z, The velocity vectors of the target in space at a predetermined future time Δt in three predetermined directions: x, y, and z. [Velocity] x, Velocity y, Velocity z, ].

[0129] Figure 4B The second neural network model shown can be a multi-parameter residual correction neural network model, meaning that a multi-parameter residual correction neural network model can simultaneously correct the residuals of multiple parameters to be corrected. For the multi-parameter residual correction neural network model, during training, the input training data includes two rows of orbital data and multiple parameters from the perturbation coefficients. The labeled data are the accurate values ​​of the residuals of the multiple parameters to be corrected. These accurate values ​​can be the accurate values ​​of the parameter residuals in the corresponding two rows of orbital data or the accurate values ​​of the parameter residuals in the corresponding perturbation coefficients.

[0130] Optionally, in this embodiment, the second neural network model can also be a module composed of multiple single-parameter residual correction neural network models, wherein each single-parameter residual correction neural network model can correct the residual of a parameter to be corrected, and the module composed of multiple single-parameter residual correction neural network models replaces... Figure 4B The multi-parameter residual correction neural network model in the middle is used to obtain Figure 4C Please refer to the architecture diagram shown. Figure 4C , Figure 4C This is an architecture diagram of another parameter correction model provided in an embodiment of this application. For a single-parameter residual correction neural network model, during training, the input training data includes two rows of orbital data and multiple parameters from the perturbation coefficients. The labeled data is an accurate value of the parameter residual to be corrected, which can be the accurate value of the parameter residual in the corresponding two rows of orbital data or the accurate value of the parameter residual in the corresponding perturbation coefficients. Different single-parameter correction neural network models use different types of labeled parameters during training.

[0131] In the above embodiments of this application, at least one second parameter to be corrected and at least one second auxiliary parameter are determined from multiple parameters including the two rows of orbital elements and / or perturbation coefficients. The second parameter to be corrected, the second auxiliary parameter, and space environment data are input into a second neural network model to obtain the residual value of the second parameter to be corrected. The residual value is added to the original value of the second parameter to be corrected to obtain the second corrected parameter. The second corrected parameter replaces the second parameter to be corrected to obtain the second corrected data. The method of this embodiment corrects the determined parameter to be corrected by a trained second neural network model. Since the trained second neural network model can accurately learn the relationship between the second parameter to be corrected, the second auxiliary parameter, the space environment data, and the residual of the second corrected parameter, the corrected parameter is closer to the truth than the uncorrected parameter, reducing errors and obtaining more accurate corrected data. This makes the orbital information of the space target predicted based on the corrected data more accurate.

[0132] In this application, the pre-trained ensemble model includes the aforementioned pre-trained parameter correction model and the aforementioned space target orbit prediction model. The space target orbit prediction model can be SGP4.

[0133] Optionally, please see Figure 5A , Figure 5A A flowchart illustrating a training method for an ensemble model provided in this application embodiment is shown. The method includes the following steps:

[0134] S501. Acquire training data, which includes the historical two-line orbital elements, perturbation coefficients, space environment data, and measured orbital information of space targets that match the historical two-line orbital elements.

[0135] Acquiring training data may include the historical two-line orbital elements of a space target [B] * [j0′,ω0′,n0′,i0′,e0′,M0′,Ω0′,…], perturbation coefficients [j2′,j3′,…], and precise ephemeris data corresponding to the two historical orbital elements of the space target, i.e., orbital information in different directions [Position′]. x, Position y Position z, Velocity′ x, Velocity′ y, Velocity′ z′[and space environment data [KP1, AP1, ISN1, F10.7_OBS1, F10.7_LAST811]]. Among them, the precise ephemeris data is a post-hoc precise ephemeris, which includes measured orbit information. After obtaining the precise ephemeris data, it is transformed into the coordinate system to obtain the orbit information in the TME coordinate system. The TME coordinate system is the coordinate system required for subsequent model prediction.

[0136] For any epoch time t0, determine the precise ephemeris data corresponding to that epoch time t0, set the predicted future time point t, and calculate the orbit information at the predicted time Δt′=t-t0.

[0137] The historical two-line orbital elements, precise ephemeris data, perturbation coefficients, space environment data, and Δt of the space target are merged into a total training dataset, from which training dataset, validation dataset, and test dataset are divided, and training begins.

[0138] S502. Use training data to train the initial parameter correction model in the initial ensemble model until the preset convergence condition is met, and obtain the trained ensemble model. The initial ensemble model includes the initial parameter correction model and the fourth version of the initial simplified general perturbation model.

[0139] The historical two-line orbital elements, perturbation coefficients, and space environment data of the space target are input into the initial ensemble model for training, obtaining the predicted orbital information of the space target at a preset time Δt′. If the error between the predicted and measured orbital information does not meet the preset error threshold, the preset parameter weights and bias values ​​of the initial ensemble model are adjusted until the error between the predicted and measured orbital information meets the preset error threshold, resulting in the trained ensemble model, as shown below. Figure 5B The diagram shown is as follows. Figure 5B This is a schematic diagram of a training process provided in an embodiment of this application.

[0140] During training, by setting the number of epochs (e.g., 30), the learning rate (e.g., 1e-3), the batch size (e.g., 2048), and hyperparameters such as the number of model layers (e.g., two hidden layers with a hidden layer input-output size of 64), the model is trained using a training dataset. Since SGP4 is integrated into the ensemble model, the model can directly calculate the partial derivatives of the parameters throughout the computation, enabling automatic parameter iteration. This makes the predicted orbital information closer to the actual orbital information, thus improving the accuracy of orbital prediction and achieving better training results. Simultaneously, the neural network framework can fully utilize hardware computing resources such as GPUs to accelerate instructions like matrix multiplication and can tensorize the entire batch of training data at once, significantly improving the computational efficiency of SGP4. After training, the trained ensemble model is evaluated and tested using validation and test datasets, ultimately yielding the trained ensemble model. By leveraging mature neural network frameworks and fully utilizing computing resources (such as GPUs) for computational acceleration, along with tensor-based parallel computing and distributed computing, not only is model training more efficient, but model inference is also faster. When predicting orbits for large-scale low-Earth orbit targets, this method offers significantly higher computational efficiency compared to existing techniques that rely solely on SGP4, thereby substantially reducing time overhead.

[0141] In this embodiment, the initial ensemble model can be obtained in the following manner.

[0142] The process involves obtaining the initial simplified general perturbation model (SGP4) version 4 and the initial parameter correction model. Tensor quantization is applied to the parameters of the initial simplified general perturbation model version 4, endowing the previously unlearnable SGP4 algorithm with the characteristics of neural networks, such as automatic differentiation, backpropagation, and parallel computation, resulting in a transformed simplified general perturbation model version 4. The initial parameter correction model is then fused with the transformed simplified general perturbation model version 4 to obtain the initial ensemble model. The initial parameter correction model can be a neural network model.

[0143] Optionally, the common variables and parameters of the initial simplified general perturbation model (4th edition) and the initial parameter correction model are determined. The data of the two models are standardized, and key features that need to be spliced ​​in the two models are identified and selected. If the feature dimensions of the two models are different, feature mapping or transformation is performed to ensure consistent processing during integration. A suitable splicing algorithm, such as a machine learning algorithm, is selected to automatically learn the optimal splicing method and finally obtain the fused initial integrated model.

[0144] In the above embodiments of this application, training data is acquired, including historical two-line orbital elements of the space target, space environment data, and measured orbital information of space targets matching the historical two-line orbital elements. The initial ensemble model is trained using this training data until a preset convergence condition is met, resulting in a trained ensemble model. This embodiment trains the initial ensemble model, making the predicted orbital information more accurate when using the trained ensemble model for space target orbit prediction. This embodiment organically integrates the initial simplified general perturbation model (SGP4) and the initial parameter correction model (i.e., the neural network model), and uses tensor quantization language of neural networks for unified construction. This ensures that the backpropagation process using the neural network not only considers data fitting but also incorporates strictly defined spacecraft dynamics principles using SGP4, making the final trained ensemble model more accurate and effective, thereby improving the accuracy of orbit prediction.

[0145] Furthermore, the technical effects brought about by applying this application will be illustrated below with specific data.

[0146] Taking B* (BSTAR) in TLE as an example, the parameter to be corrected is input into the parameter correction model along with other TLE parameters, perturbation coefficients, and space environment parameters. The output is a single parameter, which is regarded as the corrected B*. This parameter, along with other TLE parameters and the prediction time Δt, is input into SGP4 to predict the position and velocity vectors of the space target in days 1-15.

[0147] Please see Figure 6A , Figure 6A This application provides a bar chart illustrating the location prediction error obtained solely through the SGP4 model in a prior art technique. The horizontal axis represents the number of days, and the vertical axis represents the location error. Figure 6A As can be seen, the predicted position error is approximately 1 km on the first day, 2 km on the second day, 4 km on the third day, 5 km on the fourth day, 8 km on the fifth day, 11 km on the sixth day, 16 km on the seventh day, 18 km on the eighth day, 26 km on the ninth day, 32 km on the tenth day, 42 km on the eleventh day, 53 km on the twelfth day, 64 km on the thirteenth day, 74 km on the fourteenth day, and 85 km on the fifteenth day. The position error increases significantly with each day.

[0148] Figure 6BThis application provides a bar chart of velocity prediction error obtained solely through the SGP4 model in a prior art technique, where the horizontal axis represents the number of days and the vertical axis represents the velocity error. Figure 6B As can be seen, the predicted speed error is approximately 0.001 km / s on the first day, 0.002 km / s on the second day, 0.004 km / s on the third day, 0.005 km / s on the fourth day, 0.008 km / s on the fifth day, 0.011 km / s on the sixth day, 0.017 km / s on the seventh day, 0.022 km / s on the eighth day, 0.031 km / s on the ninth day, 0.037 km / s on the tenth day, 0.048 km / s on the eleventh day, 0.056 km / s on the twelfth day, 0.068 km / s on the thirteenth day, 0.078 km / s on the fourteenth day, and 0.095 km / s on the fifteenth day. The speed error increases with each day.

[0149] Figure 6CThis application provides a location prediction error comparison bar chart, where the horizontal axis represents the number of days and the vertical axis represents the location error. The location error predicted by the method of this application for the first day is approximately 1 km, almost the same as in the prior art. The location error predicted by the method of this application for the second day is approximately 2.5 km, very small compared to the prior art. The location error predicted by the method of this application for the third day is approximately 3.5 km, lower than the 4 km in the prior art. The location error predicted by the method of this application for the fourth day is approximately 4.5 km, lower than the 5 km in the prior art. The location error predicted by the method of this application for the fifth day is approximately 5 km, lower than the 8 km in the prior art. The location error predicted by the method of this application for the sixth day is approximately 10 km, lower than the 11 km in the prior art. The location error predicted by the method of this application for the seventh day is approximately 14 km, lower than the 16 km in the prior art. The location error predicted by the method of this application for the eighth day is approximately 16.5 km, lower than... The positional error predicted by the method of this application for the ninth day is approximately 18.2 kilometers, lower than the 26 kilometers in the prior art; the positional error predicted by the method of this application for the tenth day is approximately 26 kilometers, lower than the 32 kilometers in the prior art; the positional error predicted by the method of this application for the eleventh day is approximately 32 kilometers, lower than the 42 kilometers in the prior art; the positional error predicted by the method of this application for the twelfth day is approximately 33 kilometers, significantly lower than the 53 kilometers in the prior art; the positional error predicted by the method of this application for the thirteenth day is approximately 42 kilometers, significantly lower than the 64 kilometers in the prior art; the positional error predicted by the method of this application for the fourteenth day is approximately 46 kilometers, significantly lower than the 74 kilometers in the prior art; and the positional error predicted by the method of this application for the fifteenth day is approximately 54 kilometers, significantly lower than the 85 kilometers in the prior art. With the increase of the number of days, the positional error of this application is significantly reduced compared to the prior art.

[0150] Figure 6D This application provides a speed prediction error comparison histogram, where the horizontal axis represents the number of days and the vertical axis represents the speed error. From... Figure 6DAs can be seen, the speed error predicted by the method of this application on the first day is approximately 0.001 km / s, almost the same as the prior art. The speed error predicted by the method of this application on the second day is approximately 0.0025 km / s, which is very small compared to the prior art. The speed error predicted by the method of this application on the third day is approximately 0.003 km / s, lower than the 0.004 km / s in the prior art. The speed error predicted by the method of this application on the fourth day is approximately 0.0035 km / s, lower than the 0.005 km / s in the prior art. The speed error predicted by the method of this application on the fifth day is approximately 0.0055 km / s, lower than the 0.008 km / s in the prior art. The speed error predicted by the method of this application on the sixth day is approximately 0.01 km / s, lower than the 0.011 km / s in the prior art. The speed error predicted by the method of this application on the seventh day is approximately 0.012 km / s, lower than the 0.017 km / s in the prior art. The speed error predicted by the method of this application on the eighth day is approximately 0.018 km / s, lower than the prior art. The speed error predicted by the method of this application for the ninth day is approximately 0.021 km / s, which is lower than the 0.031 km / s in the prior art. The speed error predicted by the method of this application for the tenth day is approximately 0.02 km / s, which is lower than the 0.037 km / s in the prior art. The speed error predicted by the method of this application for the eleventh day is approximately 0.03 km / s, which is lower than the 0.048 km / s in the prior art. The speed error predicted by the method of this application for the twelfth day is approximately 0.036 km / s, which is significantly lower than the 0.056 km / s in the prior art. The speed error predicted by the method of this application for the thirteenth day is approximately 0.042 km / s, which is significantly lower than the 0.068 km / s in the prior art. The speed error predicted by the method of this application for the fourteenth day is approximately 0.05 km / s, which is significantly lower than the 0.078 km / s in the prior art. The speed error predicted by the method of this application for the fifteenth day is approximately 0.056 km / s, which is significantly lower than the 0.095 km / s in the prior art. As the number of days increases, the speed error of this application is significantly reduced compared to the prior art.

[0151] In summary, the position prediction error obtained through the ensemble model is significantly lower than that obtained through SGP4 alone in existing technologies. Similarly, the velocity prediction error obtained through the ensemble model is also significantly lower than that obtained through SGP4 alone in existing technologies. Statistical analysis shows that the 7-15 day orbit prediction accuracy of the method in this application is improved by 10%-40% compared to existing technologies, indicating a significant improvement in the accuracy of space target orbit prediction.

[0152] This application also provides a space target orbit prediction device; please refer to [link to relevant documentation]. Figure 7 , Figure 7 A schematic diagram of the structure of a space target trajectory prediction device provided in an embodiment of this application is shown below. Figure 7 As shown, it includes:

[0153] The acquisition module 701 is used to acquire the two orbital elements, space environment data, and perturbation coefficients of the space target at the current moment. The perturbation coefficients are the coefficients in the space target orbit prediction model.

[0154] The correction module 702 is used to correct the number of orbital elements and / or perturbation coefficients based on the pre-trained parameter correction model, perturbation coefficients, number of orbital elements in two rows, and space environment data, to obtain corrected data, which includes the corrected number of orbital elements in two rows and / or the corrected perturbation coefficients.

[0155] The prediction module 703 is used to predict the orbit information of a space target at a preset future time based on the space target orbit prediction model and the corrected data.

[0156] In one possible implementation, the pre-trained parameter correction model includes a pre-trained first neural network model, and correction module 702 is specifically used for:

[0157] From the multiple parameters included in the two-line orbital elements and / or perturbation coefficients, at least one first parameter to be corrected and at least one first auxiliary parameter are determined, wherein the first auxiliary parameter is the remaining parameter among the multiple parameters excluding the first parameter to be corrected.

[0158] The first parameter to be corrected, the first auxiliary parameter, and the space environment data are input into the first neural network model to obtain the first corrected parameter.

[0159] Replace the first parameter to be corrected with the first corrected parameter to obtain the first corrected data.

[0160] In one possible implementation, the pre-trained parameter correction model includes a pre-trained second neural network model, and the correction module 702 is specifically used for:

[0161] From the multiple parameters included in the two-line orbital elements and / or perturbation coefficients, at least one second parameter to be corrected and at least one second auxiliary parameter are determined, wherein the second auxiliary parameter is the remaining parameter among the multiple parameters other than the second parameter to be corrected.

[0162] The second parameter to be corrected, the second auxiliary parameter, and the space environment data are input into the second neural network model to obtain the residual value of the second parameter to be corrected.

[0163] The residual value is added to the original value of the second parameter to be corrected to obtain the second corrected parameter.

[0164] Replace the second parameter to be corrected with the second corrected parameter to obtain the second corrected data.

[0165] In one possible implementation, the space target orbit prediction model is the Simplified General Perturbation Model, Version 4 (SGP4), with prediction module 703 specifically used for:

[0166] The corrected data is input into the Simplified General Perturbation Model 4 (SGP4) to predict the orbital information of the space target at a preset future time. The orbital information includes displacement and velocity information.

[0167] In one possible implementation, the pre-trained ensemble model includes a pre-trained parameter correction model and a space target orbit prediction model.

[0168] In one possible implementation, a training module is also included, for:

[0169] Acquire training data, which includes the historical two-line orbital elements, perturbation coefficients, space environment data, and measured orbital information of space targets that match the historical two-line orbital elements.

[0170] The initial parameter correction model in the initial ensemble model is trained using training data until a preset convergence condition is met, resulting in a trained ensemble model. The initial ensemble model includes the initial parameter correction model and the initial simplified general perturbation model version 4 (SGP4). In one possible implementation, the space target orbit prediction model is the simplified general perturbation model version 4 (SGP4). Before acquiring the training data, the training module is further used for:

[0171] Obtain the fourth edition of the initial simplified general perturbation model and the initial parameter correction model.

[0172] The parameters of the initial simplified general perturbation model (version 4) are represented by tensor quantization to form the transformed simplified general perturbation model (version 4).

[0173] The initial parameter correction model is fused with the fourth version of the simplified general perturbation model after transformation to obtain the initial ensemble model.

[0174] The orbit prediction device for space targets provided in this embodiment can execute the method provided in the above-described method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0175] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 8As shown, the electronic device 80 provided in this embodiment includes at least one processor 801 and a memory 802. Optionally, the device 80 further includes a communication component 803. The processor 801, memory 802, and communication component 803 are connected via a bus 804.

[0176] In a specific implementation, at least one processor 801 executes computer execution instructions stored in memory 802, causing at least one processor 801 to perform the above-described method.

[0177] The specific implementation process of processor 801 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0178] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0179] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0180] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0181] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0182] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0183] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0184] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0185] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0186] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0187] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0188] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0189] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0190] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for predicting the orbit of a space target, characterized in that, include: The system obtains the two orbital elements, space environment data, and perturbation coefficients of the space target at the current moment, wherein the perturbation coefficients are coefficients in the space target orbit prediction model. Based on the pre-trained parameter correction model, the two rows of orbital elements, the perturbation coefficients, and the space environment data, the two rows of orbital elements and / or the perturbation coefficients are corrected to obtain corrected data, which includes the corrected two rows of orbital elements and / or the corrected perturbation coefficients. Based on the space target orbit prediction model and the corrected data, the orbit information of the space target at a preset future time is predicted.

2. The method according to claim 1, characterized in that, The pre-trained parameter correction model includes a pre-trained first neural network model; The method involves correcting the two rows of orbital elements and / or the perturbation coefficients based on a pre-trained parameter correction model, the two rows of orbital elements, the perturbation coefficients, and the spatial environment data to obtain corrected data, including: From the two rows of orbital elements and / or the perturbation coefficient, at least one first parameter to be corrected and at least one first auxiliary parameter are determined, wherein the first auxiliary parameter is the remaining parameter among the multiple parameters excluding the first parameter to be corrected. The first parameter to be corrected, the first auxiliary parameter, and the space environment data are input into the first neural network model to obtain the first corrected parameter; Replace the first parameter to be corrected with the first corrected parameter to obtain the first corrected data.

3. The method according to claim 1, characterized in that, The pre-trained parameter correction model includes a pre-trained second neural network model; The method involves correcting the two rows of orbital elements and / or the perturbation coefficients based on a pre-trained parameter correction model, the two rows of orbital elements, the perturbation coefficients, and the spatial environment data to obtain corrected data, including: From the two rows of orbital elements and / or the perturbation coefficient, at least one second parameter to be corrected and at least one second auxiliary parameter are determined, wherein the second auxiliary parameter is the remaining parameter among the multiple parameters excluding the second parameter to be corrected; The second parameter to be corrected, the second auxiliary parameter, and the space environment data are input into the second neural network model to obtain the residual value of the second parameter to be corrected. The residual value is added to the original value of the second parameter to be corrected to obtain the second corrected parameter; Replace the second parameter to be corrected with the second corrected parameter to obtain the second corrected data.

4. The method according to claim 2 or 3, characterized in that, The space target orbit prediction model is the Simplified General Perturbation Model, Version 4 (SGP4). The prediction of the orbit information of the space target at a preset future time based on the space target orbit prediction model and the corrected data includes: The corrected data is input into the Simplified General Perturbation Model 4 (SGP4) to predict the orbital information of the space target at a preset future time. The orbital information includes displacement information and velocity information.

5. The method according to claim 1, characterized in that, The pre-trained ensemble model includes: the pre-trained parameter correction model and the space target orbit prediction model.

6. The method according to claim 5, characterized in that, The training steps of the ensemble model include: Acquire training data, which includes the historical two-line orbital elements, perturbation coefficients, space environment data, and measured orbital information of space targets that match the historical two-line orbital elements; The initial parameter correction model in the initial ensemble model is trained using the training data until the preset convergence condition is met, and the trained ensemble model is obtained. The initial ensemble model includes the initial parameter correction model and the fourth version of the initial simplified general perturbation model.

7. The method according to claim 6, characterized in that, The space target orbit prediction model is the Simplified General Perturbation Model, Version 4 (SGP4). Before acquiring the training data, the method further includes: Obtain the fourth version of the initial simplified general perturbation model and the initial parameter correction model; The parameters of the initial simplified general perturbation model (version 4) are represented by tensor quantization to form the transformed simplified general perturbation model (version 4). The initial parameter correction model is fused with the fourth version of the transformed simplified general perturbation model to obtain the initial integrated model.

8. A device for predicting the orbit of a space target, characterized in that, include: The acquisition module is used to acquire the two rows of orbital elements, space environment data, and perturbation coefficients of the space target at the current moment. The perturbation coefficients are the coefficients in the space target orbit prediction model. The correction module is used to correct the two rows of orbital elements and / or the perturbation coefficients based on the pre-trained parameter correction model, the two rows of orbital elements, the perturbation coefficients, and the space environment data, to obtain corrected data, wherein the corrected data includes the corrected two rows of orbital elements and / or the corrected perturbation coefficients. The prediction module is used to predict the orbit information of the space target at a preset future time based on the space target orbit prediction model and the corrected data.

9. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

Images