Track-driven atmospheric density estimation method

Abstract

The invention discloses an orbit-driven atmospheric density estimation method. The method comprises the following steps: acquiring historical orbit element data and screening satellite numbers in a required time range; based on the obtained historical orbit element data and the screened satellite number, calculating a propagation orbit of the satellite in a future time period through an orbit propagation model; defining historical orbit element data as a real orbit, and calculating a propagation error at the same time point according to the real orbit and the propagation orbit; dividing the propagation errors into a regular class and an irregular class, and selecting satellite orbit parameters of the regular class; constructing an atmospheric density prediction model, and inputting the selected rule-class satellite orbit parameters into the atmospheric density prediction model for training and evaluation to obtain an optimized atmospheric density prediction model; and inputting the orbit parameter, the state vector, the space environment index and the error of the front and back orbit propagation in the previous time period into the optimized atmospheric density prediction model to obtain the atmospheric density at the future moment.

Description

Technical Field

[0001] The present invention belongs to the technical field of atmospheric density estimation, and particularly relates to an orbit-driven atmospheric density estimation method. Background Art

[0002] With the sharp increase in the number of satellites in low Earth orbit (LEO), especially the deployment of large-scale satellite constellations (such as Starlink, OneWeb, etc.), the real-time monitoring of the atmospheric density at the location has become increasingly important. LEO generally refers to an orbit with a height of about 200 kilometers to 2000 kilometers. This orbital region has become a popular choice for satellite deployment due to its relatively low altitude and short orbital period. The LEO orbit is not only a core component of modern communication, navigation, and Earth observation missions, but also provides valuable data support in areas such as meteorological monitoring, environmental monitoring, disaster warning, and scientific research.

[0003] However, with the increase in the number of satellites and the complexity of missions, the accurate prediction of satellite orbits and the real-time monitoring of atmospheric density face unprecedented challenges. Traditional orbit prediction methods usually rely on ground observations and simplified mathematical models. These methods are difficult to meet the real-time and accurate requirements when facing large-scale satellite constellations. At the same time, the atmospheric density in the low Earth orbit region is affected by many factors, such as solar activity and changes in the Earth's magnetic field. These factors make the orbital propagation error and atmospheric density change extremely complex.

[0004] With the increase in the number of satellite constellations, especially the increasing demand for low-latency, high-precision communication and precise navigation, the requirements for atmospheric density monitoring are also getting higher and higher. Large-scale satellite constellations not only require accurate prediction of the orbits of individual satellites, but also need to real-time evaluate the relative positions and dynamic changes of the entire constellation. This high-frequency and high-precision orbit prediction not only needs to overcome the limitations of traditional technologies, but also requires stronger real-time computing and adaptive capabilities.

[0005] In addition, the monitoring of atmospheric density plays a crucial role in this process. Atmospheric density directly affects the orbital decay of satellites, the orbital change speed, and the lifespan of satellites. Therefore, accurately estimating the density data at the position of the satellite at each moment is crucial for optimizing orbit design and improving the stability of satellite missions. However, due to the influence of multiple factors such as the meteorological environment and atmospheric density changes in the low Earth orbit region, traditional static density models cannot meet the dynamic and real-time monitoring requirements. Therefore, how to achieve high-precision real-time estimation of atmospheric density in the context of large-scale satellite constellations has become a key issue in satellite orbit management, space environment monitoring, and communication and navigation mission planning. To address this challenge, the present invention proposes an orbit-driven atmospheric density estimation method. Summary of the Invention

[0006] To solve the above technical problems, the present invention proposes an orbit-driven atmospheric density estimation method to improve the prediction accuracy of atmospheric density and provide important support for the stable operation and precise navigation of large-scale satellite constellations.

[0007] To achieve the above object, the present invention provides an orbit-driven atmospheric density estimation method, including:

[0008] Obtain historical orbit element data and screen out satellite numbers within the required time range;

[0009] Based on the obtained historical orbit element data and the screened satellite numbers, calculate the propagation orbit of the satellite in the future time period through the orbit propagation model;

[0010] Define the historical orbit element data as the true orbit, and calculate the propagation error according to the true orbit and the propagation orbit at the same time point;

[0011] Divide the propagation error into regular and irregular categories, and select the satellite orbit parameters of the regular category;

[0012] Construct an atmospheric density prediction model, input the selected satellite orbit parameters of the regular category into the atmospheric density prediction model for training and evaluation, and obtain an optimized atmospheric density prediction model;

[0013] Input the orbit parameters, state vectors, space environment indices, and errors of the forward and backward orbit propagations in the previous time period into the optimized atmospheric density prediction model to obtain the atmospheric density at a future moment.

[0014] Optionally, calculating the propagation error according to the true orbit and the propagation orbit at the same time point includes:

[0015] e i (t0,t1) = O propagation (t0,t1) - O true (t0,t1);

[0016] Wherein, O propagation is the propagation orbit; O true is the true orbit; e i is the error of a certain orbit period.

[0017] Optionally, the regular category is that the change rate of the propagation error is greater than 0.

[0018] Optionally, the irregular category is that the change rate of the propagation error is not greater than 0.

[0019] Optionally, the orbit parameters include semi-major axis, eccentricity, inclination, right ascension of the ascending node, argument of perigee, and true anomaly.

[0020] Optionally, the state vector includes the position vector and velocity vector of the satellite.

[0021] Optionally, the space environment index includes the solar activity index and the geomagnetic activity index.

[0022] Optionally, evaluating the atmospheric density prediction model includes:

[0023]

[0024] where y i is the true value; is the predicted value; N is the total number of samples.

[0025] Technical effects of the present invention: The present invention discloses an orbit-driven atmospheric density estimation method, which estimates the atmospheric density by combining traditional orbit prediction technology with advanced deep learning algorithms, and realizes high-precision atmospheric density estimation based on the corrected orbit data. By accurately modeling the orbit dynamics of low Earth orbit (LEO) satellites, the present invention can improve the prediction accuracy of atmospheric density, thereby providing a more reliable basis for satellite orbit prediction and space environment monitoring. Description of the Drawings

[0026] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0027] Figure 1 is a schematic flowchart of the orbit-driven atmospheric density estimation method according to an embodiment of the present invention;

[0028] Figure 2 is a schematic diagram of the training process of the Transformer according to an embodiment of the present invention. Detailed Embodiments

[0029] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to detail this application.

[0030] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0031] As Figure 1 - Figure 2 shown, this embodiment provides an orbit-driven atmospheric density estimation method, including: obtaining historical orbit element data and screening out the satellite numbers within the required time range;

[0032] Based on the obtained historical orbital element data and the selected satellite numbers, calculate the propagation orbit of the satellite in the future time period through the orbit propagation model;

[0033] Define the historical orbital element data as the true orbit, and calculate the propagation error according to the true orbit and the propagation orbit at the same time point;

[0034] Divide the propagation error into regular and irregular categories, and select the satellite orbit parameters of the regular category;

[0035] Construct an atmospheric density prediction model, input the selected satellite orbit parameters of the regular category into the atmospheric density prediction model for training and evaluation, and obtain an optimized atmospheric density prediction model;

[0036] Input the orbit parameters, state vectors, space environment indices, and the errors of the previous and subsequent orbit propagations in the previous time period into the optimized atmospheric density prediction model to obtain the atmospheric density at the future moment.

[0037] The specific implementation process includes the following steps:

[0038] Step S1: Obtain the TLE dataset (such as the Space-Track platform, etc.) from public data sources. TLE data is the standard format for describing satellite orbits and is widely used in satellite orbit prediction and analysis. Public platforms such as Space-Track provide the latest and historical satellite orbit data. Users can register and apply for an API key to obtain the required TLE data. Each TLE dataset usually contains two lines of key orbit parameters: the first line contains basic information about the satellite, such as satellite identification, launch date, etc.; the second line includes six important orbit parameters: semi-major axis, eccentricity, inclination, right ascension of the ascending node, argument of perigee, and mean anomaly. Through the SGP4 (Simplified General Perturbations Model 4) orbit propagation model, the state vector of the satellite at the corresponding moment of the TLE data can be calculated. SGP4 is a widely used satellite orbit propagation model that can accurately predict the evolution of satellite orbits based on the TLE data. This model calculates the propagation orbit of the satellite in the future time period based on the orbit elements in the TLE data and gives the current position vector and velocity vector of the satellite in the geocentric inertial coordinate system. Since the SGP4 model is widely used, there are implementation interfaces for it in various programming languages and software. For example, libraries such as pyephem and sgp4 in Python, as well as tools such as MATLAB, provide implementations of the SGP4 model. The space environment indices AP and DST can be obtained from the NOAA Space Weather Prediction Center. The satellite density can be calculated through the satellite's spatial position in the JB2008 model.

[0039] Step S2: Perform data screening. First, select the TLE of the corresponding satellite within the desired time range. On this basis, divide the TLE data into time periods to ensure that the selected data can cover multiple orbital periods. Considering that the orbital period of LEO satellites is usually between 90 minutes and 2 hours, 8 hours can be selected as a division unit, so that each time period contains multiple orbital periods, which helps to analyze the variation law between satellite orbits and densities. The key to data screening is to ensure that the selected data covers sufficient orbital information for accurate orbit prediction and error correction. During the screening process, the continuity and consistency of the time series need to be ensured.

[0040] Step S3: Since the LEO orbit is propagated, the SGP4 orbit model is selected for propagation calculation.

[0041] Step S4: Extract the orbits at the same time as the SGP4 calculation in step S3 from the subsequent same time period, denoted as the true orbits. These orbit data usually come from actual observations or more accurate orbit prediction models, so they are regarded as unbiased true orbits. By comparing with the "propagated orbits" calculated by the SGP4 model, the propagation error can be obtained, providing a basis for subsequent orbit correction. For error analysis, first, the time points of the true orbits need to be matched with the orbit data calculated by SGP4 to ensure data consistency and time synchronization. Then, the error calculation is based on the following formula:

[0042] e i (t0,t1) = O propagation (t0,t1) - O true (t0,t1);

[0043] where O propagation is the propagated orbit; O true is the true orbit; e i is the error of a certain orbital period.

[0044] Step S5: Classify the above calculation errors. The error classification criteria are divided into two categories: regular and irregular, according to the changing trend of the propagation error. This classification method is determined based on the maximum error value within each orbital period. In each time period, first calculate the maximum error value of each sample within one orbital period, and then classify according to the mean and standard deviation of these maximum error values. Observe the change rate of the error in each time period. If the error change rate is greater than 0, it is classified as a regular class; otherwise, it is classified as an irregular class. These changing trends reflect the correction direction of the orbit, facilitating subsequent error compensation. According to different classification results, different data sets can be created to provide more accurate data support for subsequent orbit correction and atmospheric density estimation. When the orbit error increases over time, it usually indicates that the satellite is affected by strong external disturbances (such as atmospheric drag). In this case, the orbit prediction error of the satellite will increase due to the change in atmospheric density. Therefore, the change in atmospheric density can be estimated using the decay of the orbit. Hence, regular class errors are selected for classification in this classification.

[0045] Step S6: Divide the training set and the validation set, and define the input and output. In this system, the input is the orbital parameters (semi-major axis, eccentricity, inclination, right ascension of the ascending node, argument of perigee, true anomaly) of the previous time period, the state vectors (position vector, velocity vector), the space environment indices (AP, DST), and the error (e i (t0,t1)) of the forward and backward orbit propagations. The output data is the atmospheric density of the next time period. To ensure the accuracy and effectiveness of the model, the data set needs to be reasonably divided into a training set and a validation set, where the training set is used for model training, and the validation set is used to evaluate the performance of the trained model. Before data division, basic processing of the data is required, including removing gross errors and outliers to ensure the quality of the data. The processed data can be normalized and standardized to improve the efficiency and convergence speed of model training.

[0046] Normalization is to scale the data to the interval [0,1] according to a ratio, and the formula is as follows,

[0047]

[0048] where x is the original data, and min(x) and max(x) are the minimum and maximum values in the data set respectively.

[0049] Standardization is to convert the data into a distribution with zero mean and unit variance, using the following formula,

[0050]

[0051] where μ(x) is the mean of the data, and σ(x) is the standard deviation of the data.

[0052] Step S7: The Transformer model, based on the self-attention mechanism, can effectively process sequential data and capture global dependencies in the sequence. Different from traditional recurrent neural networks (such as LSTM), Transformer does not rely on sequential processing of the sequence. Instead, it calculates the correlation weights between all input positions through the self-attention mechanism, allowing the model to process the entire sequence in parallel. The core components of the Transformer model are the encoder and the decoder. The encoder is responsible for processing the input sequence and extracting features, while the decoder generates outputs related to the target sequence. The self-attention mechanism generates a weighted representation by calculating the similarity between each input element and other elements, ensuring that the model is more efficient in capturing long-term dependencies. Through the stacking of multiple attention layers and feed-forward network layers, the Transformer model can accurately capture complex patterns in the orbital data and predict future atmospheric densities.

[0053] Step S8: For the Transformer model, the number of neurons, the number of iterations, the learning rate, the dropout rate, etc. can all be used as hyperparameters to be optimized. To improve the performance and accuracy of the model, these hyperparameters need to be optimized. For this purpose, the Grid Search algorithm can be used to find the optimal parameter combination under different error patterns. The Grid Search algorithm will traverse the given hyperparameter space, evaluate the performance of each combination on the validation set, and select the optimal parameter settings. During the optimization process, by adjusting the key hyperparameters of Transformer (such as the number of layers, the number of attention heads, the hidden layer dimension, the learning rate, and the dropout rate, etc.), the best model performance can be ensured under different error conditions. To update the parameters in the network, the ADAM (Adaptive Moment Estimation) optimizer is used. During the model training and evaluation process, the mean squared error (MSE) and the root mean squared error (RMSE) are used as evaluation metrics to measure the difference between the predicted value and the true value. The specific calculation methods are as follows:

[0054]

[0055] where, y i is the true value, is the predicted value, and N is the total number of samples.

[0056] Step S9: After the model training and optimization are completed, the Transformer model will be used to predict the atmospheric density in the next time period. By inputting the orbital parameters, space environment indices, and propagation errors in the previous time period, the model will estimate and predict the space density in the future time period, thus providing accurate support for orbital prediction and satellite mission planning.

[0057] The specific application examples of the present invention are as follows:

[0058] In the application of improving the accuracy of LEO space debris orbit prediction, the historical orbital element data of Starlink satellites were obtained, and the Starlink satellite numbers within a specific time range were screened. Subsequently, the orbital propagation model was used to calculate the propagation orbits of these satellites in the future based on the acquired data.

[0059] The historical orbital element data is used as the real orbit, and the propagation orbits at the same time point are compared to accurately calculate the propagation errors. Then, these propagation errors are carefully divided into regular and irregular categories, and the satellite orbit parameters of the regular category are selected. Using these parameters, an atmospheric density prediction model is constructed, trained and evaluated, and an optimized atmospheric density prediction model is obtained.

[0060] Afterwards, the orbital parameters, state vector, space environment index of the previous time period and the errors of the previous and subsequent orbital propagation are input into the optimized atmospheric density prediction model to invert the atmospheric density of the area.

[0061] The inverted atmospheric density is applied to the orbit prediction of LEO satellites or space debris. After actual verification, compared with the traditional orbit prediction method, the orbit prediction based on the atmospheric density inverted by the method of the present invention can improve the prediction accuracy by about 5%. This application example fully demonstrates the effectiveness and advancement of the present invention in improving the accuracy of space debris orbit prediction, which is of great significance for ensuring the safety of the low earth orbit environment and the normal operation of satellites.

[0062] In addition, the density information obtained through the present invention can also be applied to: in space launch missions, optimizing the rocket launch trajectory based on the acquired atmospheric density data, reducing energy loss during the launch process, and improving launch efficiency; providing key data support for space environment monitoring, helping researchers to deeply understand the temporal and spatial changes in atmospheric density, and thus better grasp the dynamics of the space environment; in terms of satellite orbit maintenance and maneuvering operations, adjusting the working parameters of the satellite propulsion system in real time according to the atmospheric density, to ensure that the satellite can accurately complete orbit adjustment and mission execution.

[0063] The present invention discloses an orbit-driven atmospheric density estimation method, which estimates atmospheric density by combining traditional orbit prediction technology with advanced deep learning algorithms, and realizes high-precision atmospheric density estimation based on corrected orbit data. By accurately modeling the orbital dynamics of low earth orbit (LEO) satellites, the present invention can improve the prediction accuracy of atmospheric density, thereby providing a more reliable basis for satellite orbit prediction and space environment monitoring.

[0064] The above are only the preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A track-driven atmospheric density estimation method, characterized in that: include: Get historical orbital element data and filter out satellite numbers within the required time range; Based on the historical orbital element data and the selected satellite numbers, the propagation orbit of the satellite in the future time period is calculated through the orbital propagation model; The historical orbit element data is defined as a true orbit, and a propagation error is calculated at the same time point according to the true orbit and the propagation orbit; Classifying the propagation errors into regular and irregular classes, and selecting satellite orbit parameters of the regular class; Constructing an atmospheric density prediction model, inputting the selected regular satellite orbit parameters into the atmospheric density prediction model for training and evaluation, and obtaining an optimized atmospheric density prediction model; The orbital parameters, state vector, space environment index and the errors of the previous orbital propagation are input into the optimized atmospheric density prediction model to obtain the atmospheric density at the future time.

2. The orbit-driven atmospheric density estimation method according to claim 1, characterized in that: Calculating the propagation error according to the true track and the propagation track at the same time point comprises: it is i (t0,t1)=O propagation (t0,t1)-O true (t0,t1); Among them, O propagation is the propagation track; true is the real track; e i is the error of a certain orbital period.

3. The orbit-driven atmospheric density estimation method according to claim 1, characterized in that: The rule class is that the rate of change of the propagation error is greater than zero.

4. The orbit-driven atmospheric density estimation method according to claim 1, characterized in that: The irregular category is that the rate of change of the propagation error is not greater than 0.

5. The orbit-driven atmospheric density estimation method according to claim 1, characterized in that: The orbital parameters include semi-major axis, eccentricity, inclination, right ascension of ascending node, angular distance of perigee, and true anomaly.

6. The orbit-driven atmospheric density estimation method according to claim 1, characterized in that: The state vector includes a position vector and a velocity vector of the satellite.

7. The orbit-driven atmospheric density estimation method according to claim 1, characterized in that: The space environment index includes a solar activity index and a geomagnetic activity index.

8. The orbit-driven atmospheric density estimation method according to claim 1, characterized in that: Evaluating the atmospheric density prediction model includes: Among them, y i is the true value; is the predicted value; N is the total number of samples.

Images