Space debris orbit prediction model training method and space debris orbit prediction method

Abstract

The application provides a space debris orbit prediction model training method and a space debris orbit prediction method, and relates to the technical field of space debris orbit prediction, to solve the technical problem that the prior art cannot effectively extract target information from sparse data, thereby failing to accurately perceive the change of space situation. The space debris orbit prediction model training method comprises the following steps: obtaining historical orbit data of space debris; according to the historical orbit data, extracting the position residual between the prediction state vector and the reference state vector of the space debris at different times; generating a multi-dimensional feature data set according to the position residual; performing data dimension reduction processing on the multi-dimensional feature data set to obtain a reduced sample data set; and using the reduced sample data set, training an initial random forest model through a machine learning algorithm, so that the initial random forest model establishes a mapping relationship between the orbit parameters of the space debris and the future state, and a space debris orbit prediction model is obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of space debris orbit prediction, and can be applied to the technical field of machine learning and artificial intelligence, and more particularly relates to a space debris orbit prediction model training method and a space debris orbit prediction method. BACKGROUND

[0002] Since the first launch of artificial satellites, human exploration of space has been more than sixty years. With frequent space activities, space objects have been spread throughout the Earth's orbit, and most of them are space debris except for a small number of on-orbit spacecraft. According to published data, there are more than 36,500 space debris larger than 10 cm, and more than 1 million space debris larger than 1 cm. These space debris run at high speed in the Earth's orbit and will cause great damage if they collide with spacecraft. Therefore, effective monitoring of space debris is of great significance to the safety of on-orbit spacecraft.

[0003] Currently, human monitoring of space debris mainly relies on ground-based optical and radar equipment. However, due to the constraints of construction cost and construction conditions, the layout of observation equipment is sparse. Therefore, when batch observation of multiple targets is performed simultaneously, if there is no reasonable optimization planning for space debris detection tasks, it will result in insufficient observation time, small amount of observation data and discontinuity for some targets. The existing technology is difficult to effectively extract target information from sparse data, which further affects the timely and accurate perception of space situation changes. SUMMARY

[0004] Therefore, the present application provides a space debris orbit prediction model training method and a space debris orbit prediction method to solve the technical problem that the existing technology is difficult to effectively extract target information from sparse data, which further leads to inaccurate perception of space situation changes.

[0005] One aspect of the present application provides a space debris orbit prediction model training method, comprising: obtaining historical orbit data of space debris; extracting position residuals between prediction state vectors and reference state vectors of space debris at different times according to the historical orbit data, wherein the position residuals represent the position changes of space debris in a space coordinate system; generating a multi-dimensional feature data set according to the position residuals, wherein the multi-dimensional feature data set includes multi-dimensional feature vectors and labels; performing data dimension reduction processing on the multi-dimensional feature data set to transform the multi-dimensional feature vectors therein to a low-dimensional space to obtain a reduced sample data set; and training an initial random forest model by a machine learning algorithm using the reduced sample data set, so that the initial random forest model establishes a mapping relationship between orbit parameters of space debris and future states to obtain a space debris orbit prediction model.

[0006] According to an embodiment of the present application, the position residual between the predicted state vector and the reference state vector of the space debris at different time instants is extracted from historical orbit data, including: setting a main reference time instant, wherein the state vector of the space debris at the main reference time instant is the reference state vector, and the position residual is the difference between the predicted state vector at any time instant and the reference state vector; in response to setting any time instant in N time instants as the main reference time instant, the position residual between the predicted state vector at each time instant before the main reference time instant and the corresponding reference state vector is extracted, wherein N is a positive integer greater than 1.

[0007] According to an embodiment of the present application, the multi-dimensional feature data set is subjected to data dimension reduction processing, and the multi-dimensional feature vectors in the multi-dimensional feature data set are transformed to a low-dimensional space to obtain a reduced sample data set, including: performing standardization processing on the multi-dimensional feature data set; calculating a covariance matrix based on the multi-dimensional feature data set after standardization processing, wherein the covariance matrix is used to capture the correlation between the multi-dimensional feature vectors; performing eigenvalue decomposition on the covariance matrix to obtain eigenvectors and corresponding eigenvalues; determining the part higher than the eigenvalue threshold as the principal component according to the eigenvectors; and mapping the principal component to the low-dimensional space by using a transformation matrix to obtain the reduced sample data set.

[0008] According to an embodiment of the present application, the initial random forest model is trained by a machine learning algorithm using the reduced sample data set, including: determining the to-be-optimized parameters of the initial random forest model, wherein the to-be-optimized parameters include the maximum number of iterations, the maximum depth of the decision tree, the minimum number of separated samples, the minimum number of leaf node samples, and the sampling form; determining the optimal parameter combination of the initial random forest model by using a grid search algorithm, and configuring the optimal parameter combination in the initial random forest model; and inputting the reduced sample data set into the initial random forest model configured with the optimal parameter combination for training, so that the model captures the orbit change rule of the space debris.

[0009] According to an embodiment of the present application, the initial random forest model is configured with a plurality of decision trees, and the training of the initial random forest model configured with the optimal parameter combination by inputting the reduced sample data set includes: performing a shuffling operation on the multi-dimensional feature vectors and labels in the reduced sample data set, wherein the correspondence between the multi-dimensional feature vectors and the labels after the shuffling operation remains unchanged; randomly dividing the sample data set after the shuffling operation into a training set and a test set according to a specified proportion, wherein the training set is used for training the model, and the test set is used for evaluating the performance of the model; randomly dividing the training set into a plurality of sub-sample sets in a manner with replacement, wherein each decision tree randomly selects a sub-sample set for independent training to obtain an independent prediction result; and generating a space debris orbit prediction model according to the plurality of trained decision trees.

[0010] According to an embodiment of the present invention, randomly selecting a subset of samples for independent training of each decision tree includes: using the random subspace method, randomly selecting multiple features from the subset of samples for node splitting for each decision tree; using the Gini index algorithm to determine the optimal splitting features and splitting points; in response to a node in the decision tree that cannot be split, determining the node as a leaf node, at which point the training of the representation decision tree is completed.

[0011] According to an embodiment of the present invention, the method further includes: performing an inverse normalization operation on the prediction results of each decision tree; and integrating the prediction results of multiple decision trees to obtain the orbit prediction results of the space debris.

[0012] According to an embodiment of the present invention, integrating the prediction results of multiple decision trees to obtain the orbit prediction result of space debris includes: acquiring the prediction results of multiple decision trees; and calculating the average of the prediction results of multiple decision trees to obtain the orbit prediction result of space debris.

[0013] According to an embodiment of the present invention, the method further includes: using a test set, independently evaluating the gap between the prediction results of each decision tree and the true values ​​through multiple indicators, and optimizing the model parameters, wherein the multiple indicators include: root mean square error, mean absolute error and Pearson correlation coefficient.

[0014] Another aspect of the present invention provides a method for predicting the orbit of space debris, comprising: acquiring the current orbit data of space debris; inputting the current orbit data into a space debris orbit prediction model to predict the orbit parameters of space debris; wherein the space debris orbit prediction model is trained according to the above-described space debris orbit prediction model training method.

[0015] Compared with existing technologies, the space debris orbit prediction model training method and space debris orbit prediction method provided by this invention have at least the following beneficial effects:

[0016] (1) The space debris orbit prediction model training method provided by the present invention no longer relies solely on real-time sparse observation data, but fully utilizes the historical accumulated information of space debris. That is, by mining the historical data of the orbit and features of space debris, and using the machine learning random forest algorithm to establish a short-term prediction model to analyze the short-term evolution of space debris, it can maximize the detection efficiency of the equipment under the condition of limited detection equipment resources, quickly and accurately grasp the information of space debris, and timely and effectively perceive changes in the space situation.

[0017] (2) The space debris orbit prediction model training method provided by the present invention can provide reliable short-term prediction error for most observation targets by simply inputting their historical observation data after the random forest prediction model is trained. It is simple and easy to implement. In addition, compared with traditional algorithms, the random forest algorithm has a fast training speed, can be parallelized, is suitable for processing large-scale datasets, and has strong adaptability to outliers and noise. Attached Figure Description

[0018] The above and other objects, features and advantages of the present invention will become more apparent from the following description of embodiments of the invention with reference to the accompanying drawings, in which:

[0019] Figure 1 A flowchart illustrating a method for training a space debris orbit prediction model according to an embodiment of the present invention is shown.

[0020] Figure 2 A scree plot of attribute features of a multidimensional feature dataset according to an embodiment of the present invention is illustrated schematically.

[0021] Figure 3 A scatter plot illustrating the relationship between the actual and predicted error values ​​in the U direction according to an embodiment of the present invention is shown.

[0022] Figure 4 A scatter plot illustrating the relationship between the actual and predicted error values ​​in the N direction according to an embodiment of the present invention is shown.

[0023] Figure 5 A scatter plot illustrating the relationship between the actual and predicted error values ​​in the W direction according to an embodiment of the present invention is shown.

[0024] Figure 6 The schematic diagram illustrates the principle of a space debris orbit prediction model training method according to an embodiment of the present invention;

[0025] Figure 7 This schematic diagram illustrates the structural block diagram of a space debris orbit prediction model training device according to an embodiment of the present invention;

[0026] Figure 8 The diagram illustrates the structure of an electronic device suitable for implementing a space debris orbit prediction model training method according to an embodiment of the present invention. Detailed Implementation

[0027] Hereinafter, embodiments of the present application will be described with reference to the accompanying drawings. It is to be understood, however, that the description is merely exemplary of the present application, and is not intended to limit the scope of the present application. In the following detailed description of the embodiments of the present application, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without these specific details. In other instances, well-known structures and functions have not been described in detail in order to avoid obscuring aspects of the present application.

[0028] The terms used herein are merely used to describe specific embodiments, and are not intended to limit the present application. The terms "include" and "have" and the like used herein indicate the presence of the described features, steps, operations, and / or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.

[0029] All terms used herein, including technical and scientific terms, have the same meanings as those generally understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having meanings consistent with the context of the present specification, and should not be interpreted in an idealized or excessively formal manner.

[0030] In the case of using expressions similar to "at least one of A, B, and C, etc.", it should generally be interpreted to include at least one of the items, unless otherwise defined (for example, "a system having at least one of A, B, and C" should include but not be limited to a system having A alone, a system having B alone, a system having C alone, a system having both A and B, a system having both A and C, a system having both B and C, and / or a system having A, B, and C, etc.).

[0031] In the embodiments of the present application, the collection, updating, analysis, processing, use, transmission, provision, disclosure, storage, etc. of the data involved (for example, including but not limited to user personal information) comply with the relevant legal regulations, are used for legal purposes, and do not violate public order and good customs. In particular, necessary measures are taken for user personal information to prevent illegal access to user personal information data and to maintain user personal information security and network security.

[0032] Since the first launch of artificial satellites, human exploration of space has been more than sixty years. With frequent space activities, space objects have been scattered in the Earth's orbit, most of which are space debris in addition to a small number of spacecraft in orbit. According to published data, there are more than 36,500 space debris larger than 10 cm, and more than 1 million space debris larger than 1 cm. These space debris run at high speed in the Earth's orbit and will cause great damage if they collide with spacecraft. Therefore, effective monitoring of space debris is of great significance to the safety of spacecraft in orbit.

[0033] Currently, the monitoring of space debris by humans mainly relies on ground-based optical and radar equipment, but due to the constraints of construction cost and construction conditions, the observation equipment is sparse in layout, and therefore, when a plurality of targets are observed at the same time, if there is no reasonable optimization planning for the space debris detection task, the observation time of some targets will be insufficient, and the amount of observation data will be small and discontinuous. However, the existing technology is difficult to effectively extract target information from sparse data, and thus affects the timely and accurate perception of space situation changes.

[0034] Therefore, how to make full use of the historical data of space debris, construct a short-term prediction model of information change, and then guide the optimization of the space debris detection task is an effective way to maximize the detection efficiency of the equipment under the condition of limited detection equipment resources, quickly and accurately grasp the space debris information, and timely and effectively perceive the space situation change.

[0035] Based on the above idea, the present application provides a space debris orbit prediction model training method and a space debris orbit prediction method to solve the technical problem that the existing technology is difficult to effectively extract target information from sparse data, and thus cannot accurately perceive the space situation change. The space debris orbit prediction model training method comprises:

[0036] acquiring historical orbit data of space debris;

[0037] According to the historical orbit data, the position residual between the prediction state vector and the reference state vector of the space debris at different times is extracted, wherein the position residual represents the position change of the space debris in the space coordinate system;

[0038] According to the position residual, a multi-dimensional feature data set is generated, wherein the multi-dimensional feature data set includes a multi-dimensional feature vector and a label;

[0039] The multi-dimensional feature data set is subjected to data dimension reduction processing, and the multi-dimensional feature vector is transformed to a low-dimensional space to obtain a reduced sample data set;

[0040] An initial random forest model is trained by using the reduced sample data set through a machine learning algorithm, so that the initial random forest model establishes a mapping relationship between the orbit parameters of the space debris and the future state, and a space debris orbit prediction model is obtained.

[0041] The space debris orbit prediction model training method provided by the embodiment of the present application can maximize the detection efficiency of the equipment under the condition of limited detection equipment resources, quickly and accurately grasp the space debris information, and timely and effectively perceive the space situation change, because it no longer simply relies on real-time sparse observation data, but fully utilizes the historical accumulation information of space debris, that is, by mining the orbit and feature historical data of space debris, and using a machine learning random forest algorithm to establish a short-term prediction model to analyze the short-term evolution of space debris.

[0042] In order to make the objects, technical solutions and advantages of the present application clearer, further, the present application will be described in detail below with reference to specific embodiments and drawings.

[0043] Figure 1 A flowchart of a space debris orbit prediction model training method according to an embodiment of the present application is schematically shown.

[0044] As shown in Figure 1 , the space debris orbit prediction model training method of this embodiment may, for example, include operations S1-S5.

[0045] In operation S1, historical orbit data of space debris is acquired.

[0046] In operation S2, according to the historical orbit data, a position residual between a prediction state vector and a reference state vector of the space debris at different times is extracted, wherein the position residual represents the position change of the space debris in the space coordinate system.

[0047] In operation S3, a multi-dimensional feature data set is generated according to the position residual, wherein the multi-dimensional feature data set includes a multi-dimensional feature vector and a label.

[0048] In operation S4, the multi-dimensional feature data set is subjected to data dimension reduction processing, and the multi-dimensional feature vector therein is transformed to a low-dimensional space to obtain a reduced sample data set.

[0049] In operation S5, the reduced sample data set is used to train an initial random forest model by a machine learning algorithm, so that the initial random forest model establishes a mapping relationship between the orbit parameters of the space debris and the future state, and a space debris orbit prediction model is obtained.

[0050] In this embodiment, the historical orbit data of the space debris is acquired in operation S1, for example, 10 or more TLE orbits updated from September 1 to 30, 2017 can be selected for analysis, and the corresponding TLE root number is acquired, wherein the TLE root number refers to various orbit parameters recorded in the TLE data, including orbit inclination, ascending node right ascension, eccentricity, perigee parameter, orbit period, etc., which are used to accurately describe the motion state of the orbit object.

[0051] TLE (Two-Line Orbital Element): satellite ephemeris, also known as two-line orbital data format, is used to describe the expression of the position and velocity of space flying body. Satellite ephemeris can accurately calculate, predict, depict, track the running state of satellite, flying body, such as time, position, velocity; can express the accurate parameters of celestial body, satellite, spacecraft, missile, space debris and other flying bodies; can place the flying body in three-dimensional space, and use time to depict the past, present and future of celestial body.

[0052] In this embodiment, in order to avoid the influence of satellite orbit maneuver on model prediction results, the orbits of active satellites are excluded in the target selection process, and a total of 9493 targets meet the screening conditions, and the target orbits cover all to area in low earth orbit.

[0053] After obtaining the required historical orbit data, then according to the historical orbit data, the position residual between the predicted state vector and the reference state vector of the space debris at different times is extracted as the data set used for subsequent model training and verification.

[0054] According to the embodiment of the application, operation S2 extracts the position residual between the predicted state vector and the reference state vector of the space debris at different times according to the historical orbit data, which can further include:

[0055] Set a main reference time, wherein the state vector of the space debris at the main reference time is the reference state vector, and the position residual is the difference between the predicted state vector at any time and the reference state vector from the main reference time;

[0056] In response to setting any one of the N time points as the main reference time, the position residual between the predicted state vector at each time point before the main reference time and the corresponding reference state vector is extracted, wherein N is a positive integer greater than 1.

[0057] In this embodiment, the position residual can be calculated by the Pair-Wise (pair) difference method, specifically:

[0058] The Nth time point (the time point corresponding to the latest data in all TLE data in a given time period) is set as the main reference time, at this time, the published TLE data is not subject to any prediction, therefore, the state vector (position and velocity) of the space debris is the most reliable.

[0059] The predicted state vector of the space debris from the N-1th time point to the Nth time point is calculated, and then the predicted state vector from the N-1th time point to the Nth time point is subtracted from the state vector at the Nth time point (reference state vector) to obtain the position residual.

[0060] The predicted state vector of the space debris from the N-2th moment to the Nth moment is further calculated, and the state vector at the Nth moment is subtracted. This process is repeated until the predicted state vector from the 1st moment to the Nth moment is calculated, and the state vector at the Nth moment is subtracted.

[0061] Then the N-1th moment is reset as the new main reference moment, and the predicted state vector of the space debris from the N-2th moment to the N-1th moment is calculated, and the state vector at the N-1th moment is subtracted. This process is repeated until the predicted state vector from the 1st moment to the N-1th moment is calculated, and the state vector at the N-1th moment is subtracted.

[0062] The above operation is repeated until the N-(N-2)th moment is set as the new main reference moment, and the predicted state vector of the space debris from the 1st moment to the N-(N-2)th moment is calculated, and the state vector at the N-(N-2)th moment is subtracted.

[0063] The residual of the above state vector is calculated based on the conversion between the position in the star coordinate system and the position in the geocentric equatorial coordinate system The conversion is as follows:

[0064]

[0065]

[0066] wherein, represents the position of the space debris in the radial direction U, the normal direction N, and the tangential direction W in the star coordinate system, represents the position of the space debris in the x, y, and z directions in the geocentric equatorial coordinate system, Ω represents the right ascension of the ascending node, i represents the orbital inclination, ω represents the argument of perigee, f represents the true anomaly, and Rx(·), RY(·), and RZ(·) represent the rotation matrixes of rotating a certain angle along the X axis, the Y axis, and the Z axis of the right-hand coordinate, respectively.

[0067] The position of the target in the geocentric equatorial coordinate system can be calculated by its instantaneous Keplerian elements , wherein a represents the semi-major axis of the orbit, and e represents the orbital eccentricity.

[0068] In this embodiment, the Pair-Wise method is used to generate the position residual data of 9493 targets based on the continuous TLE orbits of the space debris, and the space environment data, including the F10.7 and Ap indices, which are contemporaneous with the orbit data, are obtained to generate 9493 data set files, that is, the multi-dimensional feature data set of operation S3.

[0069] Considering the limited computing power, 200 data files with the most data are selected for processing in this embodiment. The experimental data totals 341932, containing seven attribute characteristic information and error information in three directions. Table 1 describes the statistical characteristics of each attribute characteristic in the sample population.

[0070] Table 1 Statistical description of the sample population

[0071]

[0072] After generating the multi-dimensional feature data set, the multi-dimensional feature data set is further processed for data dimension reduction.

[0073] According to an embodiment of the present application, operation S4 processes the multi-dimensional feature data set for data dimension reduction, transforms the multi-dimensional feature vectors therein to a low-dimensional space, and obtains a reduced sample data set, which can further include:

[0074] standardizing the multi-dimensional feature data set;

[0075] calculating a covariance matrix based on the standardized multi-dimensional feature data set, wherein the covariance matrix is used to capture the correlation between the multi-dimensional feature vectors;

[0076] performing eigenvalue decomposition on the covariance matrix to obtain eigenvectors and corresponding eigenvalues;

[0077] determining, according to the eigenvectors, a part higher than an eigenvalue threshold as principal components; and mapping the principal components to a low-dimensional space using a transformation matrix to obtain the reduced sample data set.

[0078] In this embodiment, the principal component analysis (PCA) method is used to process the attribute characteristic sample for data dimension reduction.

[0079] Suppose the multi-dimensional feature data set has P features and the output dimension after dimension reduction is p, then the transpose matrix X of the multi-dimensional feature data set can be represented as:

[0080] X=(X1,X2,…,X p ) T

[0081] wherein T represents the transpose operation, X1 represents an N×1 column vector, X represents an N×P matrix, and N represents the number of data in the multi-dimensional feature data set.

[0082] The specific data dimension reduction process is as follows:

[0083] calculating the mean vector of the multi-dimensional feature data set :

[0084]

[0085] wherein X ki represents the ki-th data in the transposed matrix X.

[0086] Calculating the standard matrix of zero mean of the multi-dimensional feature data set :

[0087]

[0088] Calculating the covariance matrix D of the multi-dimensional feature data set:

[0089]

[0090] Calculating the eigenvalues λ1, λ2,..., λ p and the corresponding eigenvectors α1, α2,..., α p of the covariance matrix, and sorting the eigenvalues, and determining the part higher than the eigenvalue threshold as the principal components.

[0091] Calculating the contribution rate σ k and the cumulative contribution rate δ of each principal component:

[0092]

[0093]

[0094] Selecting the d-dimensional features whose eigenvalues meet the condition, and using the corresponding eigenvectors α1, α2,..., α d to form the conversion matrix A = [α1, α2,..., α d ], and calculating the reduced dimension data Y of the multi-dimensional feature data set X:

[0095]

[0096] Most of the information in the original data set is concentrated in the top few principal components, and the number of principal components is selected by combining the eigenvalue threshold and the intuitive analysis of the scree plot.

[0097] In this embodiment, the PCA algorithm is used to reduce the dimension of the attribute features of the multi-dimensional feature data set. The sample population contains seven types of attribute features, which are referred to as original components in the PCA algorithm, and the eigenvalue scree plot of each original component is shown in Figure 2 .

[0098] Figure 2 The scree plot of the attribute features of the multi-dimensional feature data set according to the embodiment of the application is schematically shown.

[0099] As shown in Figure 2As shown, in the stone chart of the multi-dimensional feature data set attribute characteristics of the embodiment of the application, the component number is the sequence number of the attribute characteristics, and the stone chart can assist the number of principal component extraction, and it can be seen from the figure that the curve tends to be flat starting from the fourth component.

[0100] In the present embodiment, three principal components with feature values greater than 1 are extracted, which are represented as F1, F2 and F3 respectively, and cooperate with the construction of subsequent model training set and test set, and the extraction result is shown in Table 2.

[0101] Table 2 Initial feature values and extracted principal component information

[0102]

[0103] Among the three principal components extracted by the above method, the first principal component (F1) is the dominant factor, and its variance accounts for 54.32% of the total variance, followed by the second principal component (F2), and its variance accounts for 16.79% of the total variance, and the cumulative information of the principal components can explain 85.95% of the total variance. The formula of each principal component and its component coefficient is as follows:

[0104]

[0105]

[0106]

[0107] Wherein, Vn represents the nth attribute characteristic in the original sample population.

[0108] After processing, the seven attribute characteristics in the original sample population are replaced by three principal components that can express more than 85% of the total information, and the process realizes data dimension reduction, retains most of the information in the original data, and to some extent, links the information redundancy in the subsequent random forest model construction process.

[0109] According to the embodiment of the application, operation S5 utilizes the dimension-reduced sample data set to train the initial random forest model through a machine learning algorithm, which can further include:

[0110] Determining the to-be-optimized parameters of the initial random forest model, wherein the to-be-optimized parameters include the maximum number of iterations, the maximum depth of the decision tree, the minimum number of separated samples, the minimum number of leaf node samples and the sampling form;

[0111] Determining the optimal parameter combination of the initial random forest model by using a grid search algorithm, and configuring the optimal parameter combination in the initial random forest model;

[0112] Inputting the dimension-reduced sample data set into the initial random forest model configured with the optimal parameter combination for training, so that the model captures the orbit variation law of the spatial fragments.

[0113] Random forest model is a kind of classifier that uses multiple trees to train and predict samples. In machine learning, random forest is a classifier containing multiple decision trees, and its output category is the mode of the categories output by individual trees. It can be used to process classification and regression problems, and is also suitable for dimension reduction problems.

[0114] There are 17 commonly used parameters in random forest model. In order to consider the actual operation cost, five indispensable parameters that determine the complexity of decision tree and sampling method are selected for hyperparameter optimization in this embodiment, which are: maximum number of iterations, maximum depth of decision tree, minimum number of separated samples, minimum number of leaf node samples and sampling form.

[0115] The five parameters to be optimized are listed in grid form to support the screening of the optimal parameter combination in the cross-validation manner. K-fold cross-validation is used to select the combination with the highest score as the optimal combination to participate in model construction.

[0116] Suppose the parameter combination to be optimized is , The objective function represents the score of constructing a random forest model, so the optimal solution of the objective function can be solved as follows:

[0117]

[0118] Before grid search traversal, a wide range of values for the parameters to be optimized needs to be set, which is called hyperparameter.

[0119] If the combination contains two optimal parameters and , the value ranges of the two parameters are , and , respectively, which are defined as grid , and the parameter combination can be expressed as:

[0120]

[0121]

[0122] Where, A=1,2,…,M, B=1,2,…,N. In the case of parameter grid point determination, the optimization of the objective function is carried out, and the parameter combination that makes it maximum is output as the hyperparameter optimization result of this embodiment.

[0123] The embodiment is based on a hyperparameter optimization algorithm to optimize five indispensable parameters in a random forest model, which affect the complexity of a decision tree and a sampling method, and preset parameter ranges and an optimal parameter combination obtained by the hyperparameter optimization algorithm are shown in Table 3.

[0124] Table 3 Value ranges of parameters to be optimized and optimal solutions

[0125]

[0126] After obtaining the optimal parameter combination, the optimal parameter combination is configured in the initial random forest model.

[0127] According to an embodiment of the present application, the initial random forest model is configured with a plurality of decision trees, wherein the dimension-reduced sample data set is input into the initial random forest model configured with the optimal parameter combination for training, and further comprising:

[0128] Shuffling the multi-dimensional feature vectors and labels in the dimension-reduced sample data set, wherein the correspondence between the shuffled multi-dimensional feature vectors and labels remains unchanged;

[0129] Randomly dividing the shuffled sample data set into a training set and a test set according to a specified proportion, wherein the training set is used for training the model, and the test set is used for evaluating the performance of the model;

[0130] Randomly dividing the training set into a plurality of sub-sample sets with replacement, wherein each decision tree randomly selects a sub-sample set for independent training to obtain an independent prediction result;

[0131] Generating a space debris orbit prediction model according to the plurality of trained decision trees.

[0132] In the embodiment, the multi-dimensional feature vectors and labels in the dimension-reduced sample data set are shuffled in sequence, and the correspondence is ensured not to be changed in the process to ensure the generalization ability of the model, and the training set and the test set are randomly divided, and the division proportion may be 4:1, for example.

[0133] Then, a plurality of sub-sample sets, i.e., root node samples, are randomly sampled from the training set with replacement, and each decision tree randomly selects a sub-sample set for independent training to obtain an independent prediction result.

[0134] According to an embodiment of the present application, randomly selecting a sub-sample set for each decision tree to perform independent training comprises:

[0135] Using a random subspace method to randomly extract a plurality of features from the sub-sample set for each decision tree to perform node splitting;

[0136] Using a Gini index algorithm to determine the best splitting feature and splitting point;

[0137] In response to the node in the decision tree being unable to split, the node is determined as a leaf node, at this time, the training of the decision tree is completed.

[0138] In the embodiment, the random subspace method is used to randomly draw m features from the attribute feature set of the sample with replacement, and the Gini index algorithm is used to determine the split node attribute, when the node in the decision tree is unable to split, the training of each decision tree is completed without pruning, and finally the space debris orbit prediction model is generated.

[0139] According to the embodiment of the present application, the space debris orbit prediction model training method further comprises:

[0140] The prediction results of each decision tree are subjected to inverse normalization operation;

[0141] The prediction results of the plurality of decision trees are integrated to obtain the orbit prediction result of the space debris.

[0142] In the embodiment, the prediction results are converted into three kinds of predicted parameter values through inverse normalization, and then the prediction results of the plurality of decision trees are integrated.

[0143] According to the embodiment of the present application, the prediction results of the plurality of decision trees are integrated to obtain the orbit prediction result of the space debris, further comprising:

[0144] The prediction results of the plurality of decision trees are obtained;

[0145] The prediction results of the plurality of decision trees are subjected to mean value calculation to obtain the orbit prediction result of the space debris.

[0146] In the embodiment, according to the principle of the random forest regression algorithm, the mean value of the prediction results of the plurality of decision trees is taken as the model prediction result:

[0147]

[0148] wherein, represents the prediction result of the model, represents the output of each decision tree, represents the independent variable, and Z represents the number of decision trees.

[0149] According to the embodiment of the present application, the space debris orbit prediction model training method further comprises:

[0150] The prediction results of each decision tree and the true value are independently evaluated by a plurality of indexes through a test set, and the model parameters are optimized, wherein the plurality of indexes include root mean square error, mean absolute error and Pearson correlation coefficient.

[0151] In the embodiment, three quantitative precision evaluation indexes are selected to estimate the gap between the estimated value of the algorithm and the true value of the test set, so as to optimize and adjust the model parameters.

[0152] The three indexes are respectively:

[0153] (1) Root Mean Squared Error (RMSE):

[0154]

[0155] (2) Mean Absolute Error (MAE):

[0156]

[0157] (3) Pearson Correlation Coefficient (PCC):

[0158]

[0159] Wherein, represents the true value of the predicted parameter in the test set, represents the predicted value of the estimated parameter obtained by the model, represents the average value of the data.

[0160] The index can intuitively describe the degree of linear correlation between the predicted value and the true value of the model, and the closer the index value is to 1, the higher the linear correlation between the two.

[0161] And The parameter can reflect the difference between the true value and the predicted value of the parameter, and a lower index value represents higher prediction accuracy.

[0162] In the embodiment, the short-term prediction error du, dn, dw of the space debris orbit in the radial U, normal N and tangential W directions is analyzed based on 273546 training data sets and 68386 experimental verification data sets.

[0163] (1) U direction error

[0164] The prediction result of the model on the U direction error of the verification data set is shown in Figure 3 .

[0165] Figure 3 The scatter plot between the true value and the predicted value of the error number in the U direction according to the embodiment of the application is schematically shown.

[0166] AsFigure 3 As shown in FIG. 6, the scatter plot intuitively shows the scatter distribution between the true value and the predicted value of the U-direction error of the verification data set. It can be found that:

[0167] The predicted value and the true value are basically in the same numerical range, between 0.1 and 0.65, and there is no obvious prediction range deviation; most of the scatter points are distributed near the 1:1 line, indicating that the difference between the predicted value and the true value is small, and there is no significant overestimation or underestimation phenomenon; although the overall prediction deviation is small, most of the scatter points are below the 1:1 line, representing a small underestimation phenomenon.

[0168] The scatter plot can intuitively show the difference and approximate distribution between the predicted data and the true data, but cannot quantitatively describe the prediction accuracy, therefore, in this embodiment, three quantitative accuracy verification indicators are also calculated as shown in Table 4.

[0169] Table 4. Accuracy verification of U-direction error prediction results

[0170]

[0171] From the PCC index calculation results, the U-direction error prediction accuracy is 0.9996, representing that there is a significant linear correlation between the true value and the predicted value of the parameter, and the prediction accuracy is high; and the RMSE and MAE parameter accuracies are 0.0049 and 0.0038 respectively, which are much lower than the statistical mean value of the U-direction error samples, representing that the model has high prediction accuracy.

[0172] (2) N-direction error

[0173] The prediction results of the model on the N-direction error of the verification data set are shown in FIG. 7. Figure 4

[0174] FIG. 8 schematically shows the scatter plot between the true value and the predicted value of the N-direction error according to an embodiment of the present application. Figure 4 As shown in FIG. 8, the scatter plot intuitively shows the difference between the predicted value and the true value and the data distribution, and it can be seen from the figure that:

[0175] Figure 4

[0176] ​​The majority of the predicted N-direction error data are distributed in the range of 0 to 0.5, with a small number of sample parameter values ​​ranging from 0.5 to 1.8, which is highly consistent with the distribution of the true data. Although there are fluctuations, most of the scattered points are still distributed near the 1:1 line, and there is no significant underestimation or overestimation, indicating that the prediction effect is good. Overall, most of the scattered points are distributed below the 1:1 line, indicating that the model prediction results have an underestimation phenomenon, and the underestimation phenomenon is more prominent in the lower value range of 0 to 0.5 and the higher value range of 1 to 1.8.

[0177] Scatter plots can visually display the differences and approximate distribution between predicted and actual data, but they cannot quantitatively describe the prediction accuracy. Therefore, in this embodiment, three quantitative accuracy verification indicators are statistically analyzed, as shown in Table 5.

[0178] Table 5. Accuracy Verification of N-Direction Error Prediction Results

[0179]

[0180] The Pearson correlation coefficient reflects the linear correlation between the model's predicted values ​​and the true values. In this experiment, the PCC index for the prediction error in the N direction reached 0.9961, indicating a high linear correlation between the two and reflecting good prediction accuracy to some extent. The root mean square error and mean absolute error can reflect the difference between the predicted values ​​and the true values. In this embodiment, the two evaluation parameters reached 0.0248 and 0.0175 respectively, which are at a low level, indicating that the difference between the model's prediction error in the N direction and the true values ​​in the training set is low, and the model accuracy is high.

[0181] (3) Error in the W direction

[0182] The model's prediction results for the W-direction error on the validation dataset are as follows: Figure 5 As shown.

[0183] Figure 5 A scatter plot illustrating the relationship between the actual and predicted error values ​​in the W direction according to an embodiment of the present invention is shown.

[0184] like Figure 5 As shown in the figure, the predicted error value in the W direction is within the range of 0 to 1.1, and the data distribution is relatively uniform, which is consistent with the data distribution of the actual values. The scatter points of the predicted and actual values ​​are concentrated near the 1:1 line, indicating that the consistency between the two is high and the model provides reasonable estimates for most of the validation set. The model does not have significant prediction errors in the W direction error data distribution range, and the scatter point fluctuations are small, indicating that the model prediction results are highly stable.

[0185] The scatter plot can intuitively show the difference and approximate distribution between the predicted data and the real data, but cannot quantitatively describe the prediction accuracy, therefore, in the embodiment, three kinds of quantitative accuracy verification indexes are simultaneously counted, as shown in Table 6.

[0186] Table 6 W-direction error prediction result accuracy verification

[0187]

[0188] From the quantitative accuracy verification index, the random forest prediction model constructed in the experiment has high prediction accuracy, specifically, the Pearson correlation coefficient of the model predicting the W-direction error is as high as 0.9999, which fully proves the good linear correlation between the predicted value and the real value of the verification set; and from the two kinds of error verification indexes, the model accuracy is 0.0092 and 0.0069, which are much lower than the real value mean of the training set parameters, and are at a high level.

[0189] The space debris orbit prediction model training method provided by the embodiment of the application can give reliable short-term prediction errors for most observation targets by only inputting the historical observation data of the observation targets after the random forest prediction model training is completed, which is simple and easy to implement; in addition, the random forest algorithm has fast training speed, is parallel, is suitable for processing large-scale data sets, and has strong adaptability to abnormal values and noises.

[0190] The embodiment of the application further provides a space debris orbit prediction method, comprising:

[0191] Obtaining current orbit data of the space debris;

[0192] Inputting the current orbit data into the space debris orbit prediction model to predict the orbit parameters of the space debris; wherein the space debris orbit prediction model is trained according to the space debris orbit prediction model training method, and therefore, will not be described here.

[0193] The space debris orbit prediction method provided by the embodiment of the application can maximize the detection efficiency of the detection equipment under the condition of limited detection equipment resources, quickly and accurately grasp the space debris information, and timely and effectively perceive the space situation change, because the space debris orbit prediction method does not simply rely on real-time sparse observation data, but fully utilizes the historical accumulation information of the space debris, that is, by mining the orbit and feature historical data of the space debris, and using the machine learning random forest algorithm to establish a short-term prediction model to analyze the short-term evolution of the space debris.

[0194] Figure 6 The principle diagram of the space debris orbit prediction model training method according to the embodiment of the application is schematically shown.

[0195] As Figure 6As shown, the principle implementation of the space debris orbit prediction model training method of the embodiment of the application is as follows:

[0196] I. Multi-dimensional feature data set generation

[0197] Calculate the position residual between the TLE root number zero time prediction data and the TLE root number prediction data to a new time, and generate a multi-dimensional feature data set.

[0198] II. Data preprocessing

[0199] Clean and standardize the data to ensure uniform input data format, handle missing values and outliers.

[0200] III. Standardized data set

[0201] Standardize the data set (mean 0, variance 1) to ensure that different features are on the same scale, so that high-variance features do not dominate principal component selection.

[0202] IV. Data dimensionality reduction

[0203] Use principal component analysis to perform data dimensionality reduction on attribute feature samples.

[0204] V. Data division

[0205] Divide the sample data after dimensionality reduction into a training set and a test set according to a 4:1 division ratio, where the training set is used to train the model and the test set is used to evaluate the model performance.

[0206] VI. Parameter optimization

[0207] The short-term prediction errors du, dn, dw of the space debris orbit in the radial U, normal N, and tangential W directions are analyzed, and the objective function is optimized to achieve global optimal solution.

[0208] VII. Prediction model training

[0209] Use the optimized parameters and training data set to train a random forest model (RF model) to obtain the prediction results of the model.

[0210] VIII. Effect evaluation

[0211] Perform precision verification on the prediction results of the model based on an independent validation data set.

[0212] Figure 7 The structure block diagram of the space debris orbit prediction model training device according to the embodiment of the application is schematically shown.

[0213] As Figure 7As shown, the space debris orbit prediction model training apparatus 700 of the embodiment of the present application comprises an acquisition module 710, an extraction module 720, a generation module 730, a data dimension reduction module 740 and a training module 750.

[0214] The acquisition module 710 is configured to acquire historical orbit data of space debris.

[0215] The extraction module 720 is configured to extract, according to the historical orbit data, a position residual between a prediction state vector and a reference state vector of the space debris at different moments, wherein the position residual represents a position change of the space debris in a space coordinate system.

[0216] The generation module 730 is configured to generate a multi-dimensional feature data set according to the position residual, wherein the multi-dimensional feature data set comprises a multi-dimensional feature vector and a label.

[0217] The data dimension reduction module 740 is configured to perform data dimension reduction processing on the multi-dimensional feature data set, to transform the multi-dimensional feature vector in the multi-dimensional feature data set to a low-dimensional space, and to obtain a sample data set after dimension reduction.

[0218] The training module 750 is configured to train an initial random forest model by using the sample data set after dimension reduction through a machine learning algorithm, to make the initial random forest model establish a mapping relationship between an orbit parameter of the space debris and a future state, and to obtain a space debris orbit prediction model.

[0219] Any one or more of the modules, sub-modules, units, sub-units according to the embodiments of the present application, or at least part of the functions of any one or more of them, can be implemented in one module. Any one or more of the modules, sub-modules, units, sub-units according to the embodiments of the present application can be split into multiple modules for implementation. Any one or more of the modules, sub-modules, units, sub-units according to the embodiments of the present application can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on board, a system on package, an application specific integrated circuit (ASIC), or any other reasonable hardware or firmware that can be integrated or packaged with a circuit, or in any one of software, hardware and firmware, or in an appropriate combination of any one or more of them. Alternatively, one or more of the modules, sub-modules, units, sub-units according to the embodiments of the present application can be at least partially implemented as computer program modules that can perform corresponding functions when executed.

[0220] For example, any multiple of the acquisition module 710, the extraction module 720, the generation module 730, the data dimension reduction module 740 and the training module 750 can be combined in one module / unit / subunit for implementation, or any one of the modules / units / subunits can be split into multiple modules / units / subunits. Alternatively, at least part of the functions of one or more of the modules / units / subunits can be combined with at least part of the functions of other modules / units / subunits, and implemented in one module / unit / subunit. According to an embodiment of the present application, at least one of the acquisition module 710, the extraction module 720, the generation module 730, the data dimension reduction module 740 and the training module 750 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on board, a system on package, an application specific integrated circuit (ASIC), or any other reasonable manner of integrating or packaging a circuit, etc. hardware or firmware, or any one of the three implementation ways of software, hardware and firmware or a proper combination of any of them. Alternatively, at least one of the acquisition module 710, the extraction module 720, the generation module 730, the data dimension reduction module 740 and the training module 750 can be at least partially implemented as a computer program module which can perform corresponding functions when executed.

[0221] It should be noted that the space debris orbit prediction model training device part in the embodiments of the present application corresponds to the space debris orbit prediction model training method part in the embodiments of the present application, and the description of the space debris orbit prediction model training device part is specifically referred to the space debris orbit prediction model training method part, which will not be repeated here.

[0222] Figure 8 The structure block diagram of an electronic device suitable for implementing the space debris orbit prediction model training method according to an embodiment of the present application is schematically shown. Figure 8 The electronic device shown is only an example and should not bring any limitation to the function and use range of the embodiments of the present application.

[0223] As Figure 8As shown, the electronic device 800 according to an embodiment of the present application includes a processor 801 which can perform various appropriate actions and processes in accordance with a program stored in a read only memory (ROM) 802 or a program loaded into a random access memory (RAM) 803 from a storage section 808. The processor 801 can include, for example, a general purpose microprocessor (e.g., a CPU), an instruction set processor, and / or a related chip set, and / or a dedicated microprocessor (e.g., an application specific integrated circuit (ASIC)), and so on. The processor 801 can also include an on-board memory for cache use. The processor 801 can include a single processing unit or multiple processing units for executing different actions of the method processes according to embodiments of the present application.

[0224] In the storage section 808, various programs and data required for the operation of the electronic device 800 are stored. The processor 801, the ROM 802, and the storage section 808 are connected to each other through a bus 804. The processor 801 performs various operations of the method processes according to embodiments of the present application by executing the programs in the ROM 802 and / or the storage section 808. Note that the programs can also be stored in one or more memories other than the ROM 802 and the storage section 808. The processor 801 can also perform various operations of the method processes according to embodiments of the present application by executing the programs stored in the one or more memories.

[0225] According to an embodiment of the present application, the electronic device 800 can further include an input / output (I / O) interface 805 which is also connected to the bus 804. The electronic device 800 can further include one or more of the following components connected to the input / output (I / O) interface 805: an input section 806 including a keyboard, a mouse, etc.; an output section 807 including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, a modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output (I / O) interface 805 as necessary. A removable recording medium 811 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is attached to the drive 810 as necessary, so that a computer program read out therefrom is installed in the storage section 808 as necessary.

[0226] According to an embodiment of the present application, the method flow according to the embodiments of the present application can be implemented as a computer software program. For example, the embodiments of the present application include a computer program product comprising a computer program carrying computer program code for executing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network through the communication section 809, and / or installed from the detachable medium 811. When the computer program is executed by the processor 801, the above-described functions defined in the system / apparatus according to the embodiments of the present application are performed. According to an embodiment of the present application, the system, apparatus, module, unit, etc. described above can be implemented by computer program modules.

[0227] The present application also provides a computer readable storage medium, which can be included in the apparatus / system / device described in the above embodiments, or exist separately without being assembled into the apparatus / system / device. The above computer readable storage medium carries one or more programs, which when executed, implement the method according to the embodiments of the present application.

[0228] According to an embodiment of the present application, the computer readable storage medium can be a non-volatile computer readable storage medium. For example, it can include but not limited to portable computer diskette, hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any appropriate combination thereof. In the present application, the computer readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus or device.

[0229] For example, according to an embodiment of the present application, the computer readable storage medium can include the ROM 802 and / or the storage section 808 described above, and / or one or more memories other than the ROM 802 and the storage section 808.

[0230] The embodiments of the present application also include a computer program product comprising a computer program containing program code for executing the method provided by the embodiments of the present application, which program code is used to make the electronic device implement the method provided by the embodiments of the present application when the computer program product is run on the electronic device.

[0231] When the computer program is executed by the processor 801, the above-described functions defined in the system / apparatus according to the embodiments of the present application are performed. According to an embodiment of the present application, the system, apparatus, module, unit, etc. described above can be implemented by computer program modules.

[0232] In one embodiment, the computer program can be tangibly embodied in a non-transitory computer readable medium, such as the optical storage device, the magnetic storage device, etc. In another embodiment, the computer program can be tangibly embodied in a signal, such as a download singal, and be distributed over the networks, such as the Internet, and be downloaded into the computer system 800 via the communication interface 809, and / or the removable storage device 811. The computer program embodied in the computer program can be transmitted in any suitable form including, but not limited to, radio frequency, electrical, magnetic, optical, electromagnetic, infrared, or other suitable transmission mediums. The computer program embodied in the computer program may, for example, be transmitted over a communication network, such as the Internet, using any of the transmission mediums.

[0233] According to an embodiment of the present application, program code embodied on one or more computer readable media may be used by, for example, a computer or computer system, to perform the computer-readable media. Specifically, the program code may, for example, be executed by a computer system to perform the computer-readable media. The program code may, for example, be written more than one of a plurality of programming languages, such as, for example, Java, C++, python, "C", or the like. The program code may, for example, be executed on a user computer, partially on the user computer, partially on a remote computer, or entirely on the remote computer or server. In the latter scenario, the remote computer may, for example, be connected to the user computer through any type of network, including, for example, a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0234] The computer program instructions may, for example, also be loaded onto a computer or other programmable information processing system to cause a series of operations to be performed on the computer or other programmable information processing system to produce a computer implemented process such that the instructions which execute on the computer or other programmable information processing system implement the functions / acts specified in the computer program instructions.

[0235] The above described embodiments of the application have been described. However, these embodiments are merely meant to be illustrative and not limiting of the scope of the application. Although each of the embodiments has been described separately, this does not mean that measures from the individual embodiments cannot be used advantageously in combination. Numerous alternatives and modifications will be apparent to those skilled in the art without departing from the scope of the present application, which is defined in the appended claims.

Claims

1. A method for training a space debris orbit prediction model, characterized in that, The method includes: Acquire historical orbital data of space debris, as well as space environment data contemporaneous with the historical orbital data; Based on the historical orbit data, the position residuals between the predicted state vector and the reference state vector of the space debris at different times are extracted using the Pair-Wise difference method. These position residuals characterize the positional changes of the space debris in the spatial coordinate system. The extraction of these position residuals using the Pair-Wise difference method based on the historical orbit data includes: A master reference time is set, wherein the state vector of the space debris at the master reference time is the reference state vector, and the position residual is the difference between the predicted state vector from any time to the master reference time and the reference state vector. In response to setting any one of the N time points as the main reference time, the position residual between the predicted state vector and the corresponding reference state vector at each time point before the main reference time is extracted, where N is a positive integer greater than 1. The position residual is calculated based on the geocentric coordinate system, from the position in the celestial coordinate system. Position in the geocentric equatorial coordinate system The conversion between them is as follows: in, Let U represent the position of the space debris in the stellar coordinate system along the radial direction U, the normal direction N, and the tangential direction W. Let Ω represent the position of the space debris in the geocentric equatorial coordinate system along the x, y, and z directions, where Ω represents the right ascension of the ascending node, i represents the orbital inclination, ω represents the argument of perigee, f represents the true perigee angle, and Rx(·), RY(·), and RZ(·) represent rotation matrices that rotate the space debris by a certain angle along the X, Y, and Z axes of the right-hand coordinate system, respectively. Based on the location residual and the spatial environment data, a multidimensional feature dataset is generated, wherein the multidimensional feature dataset includes multidimensional feature vectors and labels, and the multidimensional feature vectors include seven dimensions: apogee, perigee, tilt angle, eccentricity, F10.7, Ap index, and forecast period. The multidimensional feature dataset is subjected to dimensionality reduction processing, which transforms the multidimensional feature vectors in it to a low-dimensional space to obtain the dimensionality-reduced sample dataset. Using the dimensionality-reduced sample dataset, an initial random forest model is trained through a machine learning algorithm. This initial random forest model establishes a mapping relationship between the orbital parameters of the space debris and its future state, resulting in a space debris orbit prediction model, including: The parameters to be optimized for the initial random forest model are determined, including the maximum number of iterations, the maximum depth of the decision tree, the minimum number of separated samples, the minimum number of leaf node samples, and the sampling method. The optimal parameter combination of the initial random forest model is determined by a grid search algorithm, and the optimal parameter combination is configured in the initial random forest model. The dimensionality-reduced sample dataset is input into the initial random forest model configured with the optimal parameter combination for training, so that the model can capture the orbital change patterns of the space debris; the initial random forest model is configured with multiple decision trees; The step of inputting the dimensionality-reduced sample dataset into the initial random forest model configured with the optimal parameter combination for training includes: The multidimensional feature vectors and labels in the dimensionality-reduced sample dataset are shuffled, while the correspondence between the multidimensional feature vectors and labels remains unchanged after the shuffling operation. The shuffled sample dataset is randomly divided into a training set and a test set according to a specified ratio. The training set is used to train the model, and the test set is used to evaluate the model performance. The training set is randomly divided into multiple subsets with replacement. Each decision tree randomly selects one subset for independent training, resulting in an independent prediction result, including: Using the random subspace method, multiple features are randomly selected from the subsample set for node splitting in each decision tree; The optimal splitting features and splitting points are determined using the Gini index algorithm; If a node in the decision tree cannot be split, the node is determined to be a leaf node. At this point, it indicates that the decision tree training is complete. A space debris trajectory prediction model is generated based on multiple trained decision trees.

2. The method according to claim 1, characterized in that, The step of performing dimensionality reduction processing on the multidimensional feature dataset, transforming the multidimensional feature vectors therein to a lower-dimensional space, to obtain the dimensionality-reduced sample dataset includes: The multidimensional feature dataset is standardized. Based on the standardized multidimensional feature dataset, a covariance matrix is ​​calculated, wherein the covariance matrix is ​​used to capture the correlation between the multidimensional feature vectors; The covariance matrix is ​​decomposed into eigenvalues ​​to obtain eigenvectors and corresponding eigenvalues. Based on the feature vector, the portion above the feature value threshold is determined as the principal component; The principal components are mapped to a lower-dimensional space using a transformation matrix to obtain the dimensionality-reduced sample dataset.

3. The method according to claim 1, characterized in that, Also includes: Perform inverse normalization on the prediction results of each decision tree; By integrating the prediction results of multiple decision trees, the orbit prediction results of the space debris are obtained.

4. The method according to claim 3, characterized in that, The orbit prediction results of the space debris obtained by integrating the prediction results of multiple decision trees include: Obtain the prediction results from multiple decision trees; The orbit prediction result of the space debris is obtained by averaging the prediction results of multiple decision trees.

5. The method according to claim 4, characterized in that, Also includes: Using the test set, the gap between the prediction results and the true values ​​of each decision tree is independently evaluated through multiple indicators to optimize the model parameters. These indicators include: root mean square error, mean absolute error, and Pearson correlation coefficient.

6. A method for predicting the orbit of space debris, characterized in that, The method includes: Obtain the current orbital data of space debris; The current orbital data is input into the space debris orbital prediction model to predict the orbital parameters of the space debris. The space debris orbit prediction model is trained using the space debris orbit prediction model training method according to any one of claims 1 to 5.

Images