An atmospheric density prediction method based on an empirical atmospheric model
By constructing an atmospheric density deviation ratio correction model and optimizing the empirical atmospheric model using historical spacecraft data, the error problem of the empirical atmospheric model when the spatiotemporal dimensions change rapidly was solved, and higher accuracy density prediction was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHONGKE INSIGHT TECHNOLOGY (XIAN) CO LTD
- Filing Date
- 2026-03-03
- Publication Date
- 2026-07-21
AI Technical Summary
Existing empirical atmospheric models struggle to accurately capture rapid changes in atmospheric density across time and space, especially during periods of intense solar activity or geomagnetic storms, where errors increase significantly.
By collecting historical data from multiple spacecraft, inverting observed atmospheric density values and generating predicted density values using empirical atmospheric models, a training dataset is constructed to train an atmospheric density deviation ratio correction model. This model learns the nonlinear mapping relationship between input features and density deviation ratios, thereby correcting the systematic biases of the empirical atmospheric model.
It improves the accuracy of atmospheric density prediction and reduces prediction errors, especially during periods of intense solar activity or geomagnetic storms.
Smart Images
Figure CN122433458A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of atmospheric density prediction technology, and specifically to an atmospheric density prediction method based on an empirical atmospheric model. Background Technology
[0002] The main non-conservative force acting on low Earth orbit (LEO) spacecraft is atmospheric drag, and the orbital decay caused by atmospheric drag is a key factor affecting the accuracy of spacecraft orbit determination and long-term forecasts.
[0003] Empirical atmospheric models are widely used in engineering and scientific research, with typical examples including JB2008, the MSIS series, and the DTM series. These models are built on a large amount of historical data from sounding rockets and satellite tow, and their core function is to describe the statistical relationship between atmospheric density and various parameters (such as the solar radiation index F10.7, the geomagnetic indices Ap and Kp, as well as geographical location, time, altitude, etc.).
[0004] However, empirical atmospheric models are essentially statistical descriptions of global, long-term average states, and have obvious shortcomings. That is, existing empirical atmospheric models are unable to accurately capture the rapid changes in atmospheric density in the spatiotemporal dimensions, especially during periods of intense solar activity or geomagnetic storms, when model errors can increase significantly, reaching as high as 30% or even 100%. Summary of the Invention
[0005] The purpose of this invention is to solve the technical problem that existing empirical atmospheric models are unable to accurately capture the rapid changes in atmospheric density in the spatiotemporal dimensions and have large errors, and to provide an atmospheric density prediction method based on empirical atmospheric models.
[0006] To solve the above-mentioned technical problems, the technical solution provided by the present invention is as follows: An atmospheric density prediction method based on an empirical atmospheric model includes the following steps: 1) Data preparation and processing Collect precise orbital data from multiple spacecraft, as well as the corresponding spacecraft status parameters and space environment data, as historical data; Remove outliers from historical data and interpolate all historical data to the same time resolution; 2) Constructing the training dataset Atmospheric drag acceleration at various locations along the orbital trajectory can be separated from precise orbital data using the POD dynamics method or the energy conservation method. The observed atmospheric density values ρ_obs at various locations along the orbital trajectory are retrieved based on atmospheric drag acceleration. The empirical atmospheric model is invoked by inputting the observed atmospheric density value ρ_obs, along with the corresponding time, location, spacecraft state parameters, and space environment data, and then the empirical atmospheric model generates the predicted density value ρ_model. The density deviation ratio Ratio is calculated based on the observed atmospheric density value ρ_obs and the predicted density value ρ_model. Ratio = ρ_obs / ρ_model; The precise orbital data at each position along the orbital trajectory, along with the corresponding spacecraft state parameters and space environment data, are used as input features (X). The density deviation ratio (Ratio) at each position along the orbital trajectory is used as the target output (Y). The input features (X) and the target output (Y) are used as the training dataset. 3) Training the atmospheric density deviation ratio correction model An atmospheric density deviation ratio correction model (M) is constructed and trained with a training dataset. The atmospheric density deviation ratio correction model (M) learns the nonlinear mapping relationship between the input feature (X) and the density deviation ratio (Y), i.e., Y=M(X). 4) Predict atmospheric density Input the current space environment data and spacecraft state parameters into the empirical atmospheric model, and the empirical atmospheric model will predict the initial atmospheric density prediction value ρ_model_base. Input the current space environment data, spacecraft state parameters and the initial atmospheric density prediction value ρ_model_base into the atmospheric density deviation ratio correction model (M) to obtain the predicted density deviation ratio Ratio_pred; The optimized atmospheric density value ρ_optimized is obtained by multiplying the predicted density deviation ratio Ratio_pred with the initial atmospheric density prediction value ρ_model_base. ρ_optimized=ρ_model_base*Ratio_pred.
[0007] Further, in step 1), the precise orbital data includes TLE data, geographical location (longitude, latitude, altitude), local time, annual day, drag coefficient CD, surface-to-mass ratio A / m, and relative atmospheric velocity v; Space environment data includes the solar index F10.7, the geomagnetic index Ap, and the Kp index.
[0008] Furthermore, in step 2), the specific method for retrieving the observed atmospheric density values ρ_obs at various locations along the orbital trajectory based on atmospheric drag acceleration is as follows: Construct the equation for calculating atmospheric drag acceleration: a_d=-0.5*ρ*(CD*A / m)*v 2 In the formula, a_d is the atmospheric drag acceleration, ρ is the atmospheric density, CD is the drag coefficient, A / m is the surface mass ratio, and v is the relative atmospheric velocity. Constructing the ballistic coefficient equations: BC = CD * A / m; where BC is the ballistic coefficient; The atmospheric density value ρ is obtained based on the equations for calculating atmospheric drag acceleration and ballistic coefficients. ρ=2a_d / (BC*v 2 ); The atmospheric density value ρ is used as the observed atmospheric density value ρ_obs.
[0009] Further, in step 2), the empirical atmospheric model is a JB2008 series model, an MSIS series model, or a DTM series model.
[0010] Furthermore, in step 3), the atmospheric density deviation ratio correction model is a gradient boosting tree, random forest, or deep neural network.
[0011] Furthermore, it also includes step 5). Continuously collect precise orbital data from multiple spacecraft, as well as corresponding spacecraft status parameters and space environment data; Perform steps 1) through 3) periodically or on a triggered basis to train and update the atmospheric density deviation ratio correction model (M).
[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: The atmospheric density prediction method based on an empirical atmospheric model provided by this invention collects historical data from multiple spacecraft, retrieves the observed atmospheric density value ρ_obs, and simultaneously generates a predicted density value ρ_model using an empirical atmospheric model. The two are then divided to obtain the density deviation ratio Ratio. Next, a training dataset is constructed with precise orbital data, corresponding spacecraft state parameters, and space environment data as input features, and the density deviation ratio Ratio as the target output. Based on this training dataset, an atmospheric density deviation ratio correction model is trained, enabling the atmospheric density deviation ratio model to learn the nonlinear mapping relationship between the density deviation ratio and precise orbital data, corresponding spacecraft state parameters, and space environment data. This model is used to capture and correct the systematic biases of the empirical atmospheric model, thereby optimizing the predicted values of the empirical atmospheric model. Compared to existing empirical atmospheric models, this invention uses massive amounts of historical data as a training dataset to train the atmospheric density deviation ratio correction model, enabling the atmospheric density deviation ratio correction model to accurately capture the rapid changes in atmospheric density in the spatiotemporal dimensions. Then, the atmospheric density deviation ratio correction model is used to correct and optimize the prediction results of existing empirical atmospheric models, thereby improving prediction accuracy. Attached Figure Description
[0013] Figure 1 This is a flowchart illustrating an embodiment of the present invention; Figure 2 This is a schematic diagram of the process for retrieving atmospheric density values in an embodiment of the present invention; Figure 3 This is a comparison diagram of atmospheric density values before and after correction in an embodiment of the present invention. Detailed Implementation
[0014] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0015] like Figure 1 As shown, an atmospheric density prediction method based on an empirical atmospheric model includes the following steps: 1) Data preparation and processing Collect precise orbital data from multiple spacecraft, as well as the corresponding spacecraft status parameters and space environment data, as historical data; Precise orbital data includes TLE data, geographic location (longitude, latitude, altitude), local time, annual day, drag coefficient CD, surface mass ratio A / m, and relative atmospheric velocity v; space environment data includes solar index F10.7, geomagnetic index Ap, and Kp index. Remove outliers from historical data and interpolate all historical data to the same time resolution; 2) Constructing the training dataset like Figure 2 As shown, atmospheric drag acceleration at various locations along the orbital trajectory in the precise orbital data can be separated using the POD dynamics method or the energy conservation method. The observed atmospheric density values ρ_obs at various locations along the orbital trajectory are retrieved based on atmospheric drag acceleration. The specific method is as follows: Construct the equation for calculating atmospheric drag acceleration: a_d=-0.5*ρ*(CD*A / m)*v 2In the formula, a_d is the atmospheric drag acceleration, ρ is the atmospheric density, CD is the drag coefficient, A / m is the surface mass ratio, and v is the relative atmospheric velocity. Constructing the ballistic coefficient equations: BC = CD * A / m; where BC is the ballistic coefficient; The atmospheric density value ρ is obtained based on the equations for calculating atmospheric drag acceleration and ballistic coefficients. ρ=2a_d / (BC*v 2 ); The atmospheric density value ρ is taken as the observed atmospheric density value ρ_obs; Call the empirical atmospheric model, which can be the JB2008 series model, the MSIS series model, or the DTM series model; The observed atmospheric density value ρ_obs, along with the corresponding time, location, spacecraft state parameters, and space environment data, are input into the empirical atmospheric model, which then generates the predicted density value ρ_model. The density deviation ratio Ratio is calculated based on the observed atmospheric density value ρ_obs and the predicted density value ρ_model. Ratio = ρ_obs / ρ_model; The precise orbital data at each position along the orbital trajectory, along with the corresponding spacecraft state parameters and space environment data, are used as input features (X). The density deviation ratio (Ratio) at each position along the orbital trajectory is used as the target output (Y). The input features (X) and the target output (Y) are used as the training dataset. 3) Training the atmospheric density deviation ratio correction model Construct an atmospheric density deviation ratio correction model (M), which can be a gradient boosting tree, random forest, or deep neural network. The atmospheric density deviation ratio correction model (M) is trained using the training dataset, so that the atmospheric density deviation ratio correction model (M) learns the nonlinear mapping relationship between the input feature (X) and the density deviation ratio (Y), i.e., Y=M(X); 4) Predict atmospheric density Input the current space environment data and spacecraft state parameters into the empirical atmospheric model, and the empirical atmospheric model will predict the initial atmospheric density prediction value ρ_model_base. Input the current space environment data, spacecraft state parameters and the initial atmospheric density prediction value ρ_model_base into the atmospheric density deviation ratio correction model (M) to obtain the predicted density deviation ratio Ratio_pred; The optimized atmospheric density value ρ_optimized is obtained by multiplying the predicted density deviation ratio Ratio_pred with the initial atmospheric density prediction value ρ_model_base. ρ_optimized=ρ_model_base*Ratio_pred; In this embodiment, the JB2008 series empirical atmospheric model is used as an example to illustrate the technical solution of this embodiment: Download TLE data, geographic location (longitude, latitude, altitude), local time, day of year, drag coefficient CD, surface mass ratio A / m, relative atmospheric velocity v, and spacecraft status parameters for approximately 1,000 LEO spacecraft over the past 10 years from the Space-Track.org platform. Simultaneously download the corresponding solar index F10.7, geomagnetic index Ap, and Kp index from the NASAOMNIWeb database.
[0016] The atmospheric drag acceleration at each position along the orbital trajectory was separated from the precise orbital data using the POD dynamics method; and the observed atmospheric density value ρ_obs at each position along the orbital trajectory was retrieved based on the atmospheric drag acceleration. For each location, the JB2008 series model is used to calculate the predicted density value ρ_jb2008; Calculate the density deviation ratio: Ratio = ρ_obs / ρ_jb2008; The precise orbital data at each position along the orbital trajectory, along with the corresponding spacecraft state parameters and space environment data, are used as input features (X). The density deviation ratio (Ratio) at each position along the orbital trajectory is used as the target output (Y). The input features (X) and the target output (Y) are used as the training dataset. Each training dataset contains TLE data, geographic location (longitude, latitude, altitude), local time, day of year, drag coefficient CD, surface-to-mass ratio A / m, relative atmospheric velocity v, spacecraft state parameters, solar index F10.7, geomagnetic index Ap, Kp index, ρ_jb2008, and calculated density deviation ratio Ratio. Gradient boosting trees are used as the atmospheric density deviation ratio correction model; The training dataset is divided into a training set and a test set in an 8:2 ratio. The model hyperparameters (such as the maximum tree depth and learning rate) are tuned using grid search and cross-validation to achieve the minimum root mean square error (RMSE) on the test set. A SWARM satellite that was not used in training was selected for prediction verification under a simulated geomagnetic storm event. The observed atmospheric density value ρ_obs retrieved from the SWARM satellite using atmospheric drag acceleration over 24 hours, the predicted density value ρ_jb2008 predicted using the JB2008 series model, and the atmospheric density value ρ_optimized predicted using the atmospheric density prediction method based on the empirical atmospheric model in this embodiment were compared. Specifically, as follows... Figure 3 .
[0017] in accordance with Figure 3 It can be seen that using the atmospheric density prediction method based on the empirical atmospheric model for orbit prediction reduces the position error of the 24-hour forecast by 40% compared with the standard JB2008 model.
[0018] In other embodiments of the present invention, step 5 is also included. Continuously collect precise orbital data from multiple spacecraft, as well as corresponding spacecraft status parameters and space environment data; Steps 1) through 3) are executed periodically or triggered to train and update the atmospheric density deviation ratio correction model (M) so that it can adapt to long-term changes in the space environment and continuously improve its accuracy.
[0019] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present invention should be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A method for predicting atmospheric density based on an empirical atmospheric model, characterized in that, Includes the following steps: 1) Data preparation and processing Collect precise orbital data from multiple spacecraft, as well as the corresponding spacecraft status parameters and space environment data, as historical data; Remove outliers from historical data and interpolate all historical data to the same time resolution; 2) Constructing the training dataset Atmospheric drag acceleration at various locations along the orbital trajectory can be separated from precise orbital data using the POD dynamics method or the energy conservation method. The observed atmospheric density values ρ_obs at various locations along the orbital trajectory are retrieved based on atmospheric drag acceleration. The empirical atmospheric model is invoked by inputting the observed atmospheric density value ρ_obs, along with the corresponding time, location, spacecraft state parameters, and space environment data, and then the empirical atmospheric model generates the predicted density value ρ_model. The density deviation ratio Ratio is calculated based on the observed atmospheric density value ρ_obs and the predicted density value ρ_model. Ratio = ρ_obs / ρ_model; The precise orbital data at each position along the orbital trajectory, along with the corresponding spacecraft state parameters and space environment data, are used as input features (X). The density deviation ratio (Ratio) at each position along the orbital trajectory is used as the target output (Y). The input features (X) and the target output (Y) are used as the training dataset. 3) Training the atmospheric density deviation ratio correction model An atmospheric density deviation ratio correction model (M) is constructed and trained with a training dataset. The atmospheric density deviation ratio correction model (M) learns the nonlinear mapping relationship between the input feature (X) and the density deviation ratio (Y), i.e., Y=M(X). 4) Predict atmospheric density Input the current space environment data and spacecraft state parameters into the empirical atmospheric model, and the empirical atmospheric model will predict the initial atmospheric density prediction value ρ_model_base. Input the current space environment data, spacecraft state parameters and the initial atmospheric density prediction value ρ_model_base into the atmospheric density deviation ratio correction model (M) to obtain the predicted density deviation ratio Ratio_pred; The optimized atmospheric density value ρ_optimized is obtained by multiplying the predicted density deviation ratio Ratio_pred with the initial atmospheric density prediction value ρ_model_base. ρ_optimized=ρ_model_base*Ratio_pred.
2. The atmospheric density prediction method based on an empirical atmospheric model according to claim 1, characterized in that: In step 1), the precise orbital data includes TLE data, geographical location (longitude, latitude, altitude), local time, day of year, drag coefficient CD, surface mass ratio A / m, and relative atmospheric velocity v; Space environment data includes the solar index F10.7, the geomagnetic index Ap, and the Kp index.
3. The atmospheric density prediction method based on an empirical atmospheric model according to claim 1, characterized in that, In step 2), the specific method for retrieving the observed atmospheric density values ρ_obs at various locations along the orbital trajectory based on atmospheric drag acceleration is as follows: Construct the equation for calculating atmospheric drag acceleration: a_d=-0.5*ρ*(CD*A / m)*v 2 In the formula, a_d is the atmospheric drag acceleration, ρ is the atmospheric density, CD is the drag coefficient, A / m is the surface mass ratio, and v is the relative atmospheric velocity. Constructing the ballistic coefficient equations: BC = CD * A / m; where BC is the ballistic coefficient; The atmospheric density value ρ is obtained based on the equations for calculating atmospheric drag acceleration and ballistic coefficients. ρ=2a_d / (BC*v 2 ); The atmospheric density value ρ is used as the observed atmospheric density value ρ_obs.
4. The atmospheric density prediction method based on an empirical atmospheric model according to claim 1, characterized in that: In step 2), the empirical atmospheric model is a JB2008 series model, an MSIS series model, or a DTM series model.
5. The atmospheric density prediction method based on an empirical atmospheric model according to claim 1, characterized in that: In step 3), the atmospheric density deviation ratio correction model is a gradient boosting tree, random forest, or deep neural network.
6. The atmospheric density prediction method based on an empirical atmospheric model according to claim 1, characterized in that, It also includes step 5). Continuously collect precise orbital data from multiple spacecraft, as well as corresponding spacecraft status parameters and space environment data; Perform steps 1) through 3) periodically or on a triggered basis to train and update the atmospheric density deviation ratio correction model (M).