Spacecraft on-orbit telemetry data variation detection and correction method and device

Abstract

The invention relates to a spacecraft on-orbit telemetry data variation detection and correction method and device. The method comprises the following steps: processing a historical normal telemetry data set to obtain a time sequence data sample; according to the time sequence data samples and the corresponding label data, quantile regression of the high quantile level, the middle quantile level and the low quantile level is conducted on the time sequence prediction model, and a trained upper limit prediction model, a trained mean value prediction model and a trained lower limit prediction model are obtained; obtaining to-be-detected telemetry data at the current moment and time sequence data corresponding to normal telemetry data in the previous time period, and inputting the time sequence data into the trained upper limit, mean value and lower limit prediction models to obtain prediction upper limits, mean values and lower limits under each time step in a prediction time interval from the current moment; and judging whether the to-be-detected data is abnormal according to the prediction upper limit and the prediction lower limit at the current moment, and correcting the abnormal telemetry data by using the prediction mean value. By adopting the method, the abnormal change detection and correction of the on-orbit telemetry data of the spacecraft can be effectively realized, and the data accuracy and reliability are improved.

Description

Technical Field

[0001] The present application relates to the field of spacecraft engineering technology, and in particular to a method and device for detecting and correcting abnormal changes in on-orbit telemetry data of a spacecraft. Background Art

[0002] In practical engineering problems, downlinked spacecraft telemetry data often deviates from the true value due to factors such as sensor accuracy limitations and inaccuracies during data transmission. Analyzing data with significant deviations can lead to inaccurate judgments about the spacecraft's status and may even cause misjudgments, impacting normal operation. Therefore, detecting anomalies in spacecraft telemetry data and implementing appropriate methods to correct these deviations are pressing challenges.

[0003] Given the vast amount of historical telemetry data accumulated on the ground, data analysis based on deep learning is currently the most commonly used approach. Telemetry data is generally time series data, and time series prediction models can better reflect data trends over time. Common time series prediction models include recurrent neural networks (RNNs) and long short-term memory networks (LSTMs). RNNs consist of recurrent units, each receiving input at each time point in the sequence and using the output from the previous time point as input for the current time point. However, for long time series, they are prone to vanishing or exploding gradients. LSTMs are a variant of RNNs that incorporate gates that control the flow of information, allowing the network to learn when to remember or forget information. By capturing the dependencies between information, they enable long-term prediction. This model has been widely used in aircraft engine remaining life prediction and health monitoring, demonstrating excellent performance.

[0004] Due to factors such as data noise, spacecraft telemetry data on-orbit exhibits data uncertainty. Effectively quantifying this uncertainty is crucial for detecting data anomalies. Traditional deep learning-based time series prediction methods only achieve point estimates and fail to consider the impact of data uncertainty. While existing data uncertainty quantification methods can be effectively combined with deep learning models, the practical engineering challenge of analyzing spacecraft telemetry data on-orbit remains to accurately detect and correct anomalies based on quantified data uncertainty. Summary of the Invention

[0005] Based on this, it is necessary to provide a method and device for detecting and correcting anomalies in spacecraft on-orbit telemetry data in response to the above technical problems.

[0006] A method for detecting and correcting anomalous changes in on-orbit telemetry data of a spacecraft, the method comprising:

[0007] Performing sliding window processing on a historical normal telemetry data set of a spacecraft to obtain a time series data sample; the label data of the time series data sample includes the historical normal telemetry data within the corresponding prediction time interval;

[0008] Performing quantile regression at high, medium, and low quantile levels on the pre-built time series prediction model according to the time series data samples and the corresponding label data to obtain a trained upper limit prediction model, a mean prediction model, and a lower limit prediction model;

[0009] Obtaining time series data corresponding to the telemetry data to be tested of the spacecraft at the current moment and the normal telemetry data of the previous time period, inputting the time series data into the trained upper limit prediction model, mean prediction model, and lower limit prediction model, respectively, to obtain the prediction upper limit, prediction mean, and prediction lower limit at each time step within the prediction time interval starting from the current moment;

[0010] If the telemetry data to be detected does not fall within the prediction data interval formed by the prediction upper limit and prediction lower limit at the current moment, the data to be detected is abnormal telemetry data, and the abnormal telemetry data is corrected to the prediction mean at the current moment.

[0011] In one embodiment, the sliding window processing of the historical normal telemetry data set of the spacecraft to obtain time series data samples includes: obtaining the historical normal telemetry data set of the spacecraft; performing sliding window processing on the historical normal telemetry data set according to a preset sliding window size and sliding step size to obtain a number of time series data samples; for each time series data, using the historical normal telemetry data within a predicted time interval from the next moment as the corresponding label data; the interval length of the predicted time interval is the sliding step length.

[0012] In one embodiment, the time series prediction model includes a recurrent neural network or a long short-term memory network.

[0013] In one embodiment, the method of performing quantile regression at high, middle and low quantile levels on a pre-constructed time series prediction model according to the time series data samples and the corresponding label data to obtain a trained upper limit prediction model, a mean prediction model and a lower limit prediction model includes: constructing a first, a second and a third quantile loss function according to the preset high quantile level, the middle quantile level and the low quantile level respectively; and training the pre-constructed time series prediction model according to the time series data samples, the corresponding label data and the first, the second and the third quantile loss functions respectively to obtain a trained upper limit prediction model, a mean prediction model and a lower limit prediction model.

[0014] In one embodiment, the quantile loss function is:

[0015]

[0016] Among them, L(θ) is the quantile loss function, n s is the number of historical normal telemetry data in the training dataset, is the predicted value, y k is the kth historical normal telemetry data, τ is the quantile level, θ is the model parameter, is the indicator function.

[0017] In one embodiment, the method further includes: if the telemetry data to be detected belongs to a prediction data interval consisting of a prediction upper limit and a prediction lower limit at the current moment, the data to be detected is normal telemetry data.

[0018] In one embodiment, the method further includes: the abnormal telemetry data is converted into normal telemetry data after correction.

[0019] A device for detecting and correcting abnormal changes in on-orbit telemetry data of a spacecraft, the device comprising:

[0020] A sample acquisition module is used to perform sliding window processing on the historical normal telemetry data set of the spacecraft to obtain time series data samples; the label data of the time series data samples includes the historical normal telemetry data within the corresponding prediction time interval;

[0021] A model training module is used to perform quantile regression at high, medium and low quantile levels on the pre-built time series prediction model according to the time series data samples and the corresponding label data, so as to obtain a trained upper limit prediction model, a mean prediction model and a lower limit prediction model;

[0022] a data prediction module, configured to obtain time series data corresponding to the telemetry data to be tested of the spacecraft at the current moment and the normal telemetry data of the previous time period, input the time series data into the trained upper limit prediction model, mean prediction model, and lower limit prediction model, respectively, and obtain the prediction upper limit, prediction mean, and prediction lower limit at each time step within the prediction time interval starting from the current moment;

[0023] The detection and correction module is used to determine that if the telemetry data to be detected does not fall within the prediction data interval formed by the prediction upper limit and prediction lower limit at the current moment, the data to be detected is abnormal telemetry data, and the abnormal telemetry data is corrected to the prediction mean at the current moment.

[0024] A computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0025] Performing sliding window processing on a historical normal telemetry data set of a spacecraft to obtain a time series data sample; the label data of the time series data sample includes the historical normal telemetry data within the corresponding prediction time interval;

[0026] Performing quantile regression at high, medium, and low quantile levels on the pre-built time series prediction model according to the time series data samples and the corresponding label data to obtain a trained upper limit prediction model, a mean prediction model, and a lower limit prediction model;

[0027] Obtaining time series data corresponding to the telemetry data to be tested of the spacecraft at the current moment and the normal telemetry data of the previous time period, inputting the time series data into the trained upper limit prediction model, mean prediction model, and lower limit prediction model, respectively, to obtain the prediction upper limit, prediction mean, and prediction lower limit at each time step within the prediction time interval starting from the current moment;

[0028] If the telemetry data to be detected does not fall within the prediction data interval formed by the prediction upper limit and prediction lower limit at the current moment, the data to be detected is abnormal telemetry data, and the abnormal telemetry data is corrected to the prediction mean at the current moment.

[0029] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0030] Performing sliding window processing on a historical normal telemetry data set of a spacecraft to obtain a time series data sample; the label data of the time series data sample includes the historical normal telemetry data within the corresponding prediction time interval;

[0031] Performing quantile regression at high, medium, and low quantile levels on the pre-built time series prediction model according to the time series data samples and the corresponding label data to obtain a trained upper limit prediction model, a mean prediction model, and a lower limit prediction model;

[0032] Obtaining time series data corresponding to the telemetry data to be tested of the spacecraft at the current moment and the normal telemetry data of the previous time period, inputting the time series data into the trained upper limit prediction model, mean prediction model, and lower limit prediction model, respectively, to obtain the prediction upper limit, prediction mean, and prediction lower limit at each time step within the prediction time interval starting from the current moment;

[0033] If the telemetry data to be detected does not fall within the prediction data interval formed by the prediction upper limit and prediction lower limit at the current moment, the data to be detected is abnormal telemetry data, and the abnormal telemetry data is corrected to the prediction mean at the current moment.

[0034] The above-mentioned method and device for detecting and correcting anomalies in on-orbit telemetry data of spacecraft, by performing sliding window processing on the historical normal telemetry data set of the spacecraft and determining the label data, can provide effective samples for the training of the time series prediction model, facilitate the model to learn data features, and combine the quantile regression training upper limit, mean and lower limit prediction models at high, medium and low quantile levels to effectively quantify data uncertainty, overcome the defect of traditional time series prediction models that ignore data uncertainty, input the current time series data into the trained model to obtain the prediction upper limit, mean and lower limit, construct the prediction interval, and accurately determine whether the telemetry data to be detected is abnormal. Abnormal data can be corrected to the predicted mean, which can make the data return to a reasonable range. The embodiment of the present invention can effectively realize the detection and correction of anomalies in on-orbit telemetry data of spacecraft, improve data accuracy and reliability, and provide strong support for spacecraft reliability analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 1 is a flow chart of a method for detecting and correcting anomalous changes in on-orbit telemetry data of a spacecraft in one embodiment;

[0036] Figure 2 A schematic diagram of the framework of the offline part and the online part in one embodiment;

[0037] Figure 3 A schematic diagram of constructing sequences and labels in multi-step time series prediction in one embodiment;

[0038] Figure 4 1 is a flow chart of a time series prediction model training method according to an embodiment;

[0039] Figure 5 Schematic diagram of a cycle flow of prediction, detection and correction in one embodiment;

[0040] Figure 6 This is a structural block diagram of a device for detecting and correcting anomalies in on-orbit telemetry data of a spacecraft in one embodiment;

[0041] Figure 7 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0043] In one embodiment, Figure 1 As shown, a method for detecting and correcting anomalous changes in on-orbit telemetry data of a spacecraft is provided, comprising the following steps:

[0044] Step 102 : Perform sliding window processing on the historical normal telemetry data set of the spacecraft to obtain time series data samples.

[0045] The historical normal telemetry dataset is a collection of various telemetry data collected and recorded by the spacecraft during its normal operation. By setting a fixed-size sliding window and moving it point by point along the time series, the data within the window is used as a time series data sample. This allows the continuous time series data to be converted into a format suitable for model processing. The label data for the time series data sample includes the historical normal telemetry data corresponding to the prediction time interval. The prediction time interval length is the sliding step size. By labeling the time series data samples with this label data, model training can be supervised, allowing the model to learn how to predict future telemetry data based on historical time series data samples.

[0046] Step 104 , performing quantile regression at high, medium and low quantile levels on the pre-built time series prediction model according to the time series data samples and the corresponding label data, to obtain a trained upper limit prediction model, a mean prediction model and a lower limit prediction model.

[0047] Time series forecasting models are used to predict time series data. Based on the time series information in historical data, quantile regression can model and predict different quantiles of data without relying on specific assumptions about the data's distribution, predicting future data values. Quantile regression is performed on time series forecasting models at high, mid, and low quantile levels to generate forecast models at different quantiles, thereby quantifying data uncertainty. The three models derived from quantile regression are: the upper bound forecast model, which predicts the upper limit of the data and reflects the range of possible values under high probability; the mean forecast model, which predicts the mean of the data and estimates the central trend of the data; and the lower bound forecast model, which predicts the lower limit of the data and reflects the range of possible values under low probability.

[0048] It can be understood that the use of quantile regression method combined with time series prediction model fully considers the uncertainty of data, overcomes the shortcomings of traditional time series prediction model that ignores data uncertainty, and obtains a prediction model that can accurately describe the changing laws of data at different quantiles, providing a more comprehensive reference basis for subsequent anomaly detection.

[0049] Step 106: Obtain the time series data corresponding to the telemetry data to be tested of the spacecraft at the current moment and the normal telemetry data of the previous time period, input the time series data into the trained upper limit prediction model, mean prediction model and lower limit prediction model respectively, and obtain the prediction upper limit, prediction mean and prediction lower limit at each time step in the prediction time interval starting from the current moment.

[0050] The time series data corresponding to normal telemetry data from the previous time period is provided as input to the trained model, allowing the model to predict current and future telemetry data based on historical information. The sequence length of the time series data matches the sequence length of the time series data samples, and the end time of the time series data is close to or adjacent to the current time. It is worth noting that after correcting the detected abnormal telemetry data, the corrected data can be used as normal telemetry data and can still be included as an element in the time series data.

[0051] Normal data from spacecraft on-orbit telemetry typically fluctuates within a certain range. By inputting the current time series data into these three models, we obtain the upper, mean, and lower bounds of the prediction, forming a predicted data interval. If the telemetry data to be tested falls outside this interval, we can preliminarily identify it as an anomaly, providing a clear basis for anomaly detection.

[0052] Step 108: If the telemetry data to be detected does not fall within the prediction data interval formed by the prediction upper limit and prediction lower limit at the current moment, the data to be detected is abnormal telemetry data, and the abnormal telemetry data is corrected to the prediction mean at the current moment.

[0053] like Figure 2 The framework diagram of the offline and online parts is shown. The offline part mainly trains the time series prediction model, and the online part detects and corrects anomalies in real-time telemetry data based on the trained model.

[0054] Once the abnormal data is identified, the mean prediction model comprehensively considers the overall trends and patterns of the historical data, and its output, the predicted mean, represents, to a certain extent, the reasonable value of the normal data. Correcting the abnormal telemetry data to the predicted mean can restore the data to a relatively reasonable state, thereby improving the accuracy and reliability of spacecraft data and facilitating subsequent work such as spacecraft status assessment and fault diagnosis.

[0055] In the above-mentioned method for detecting and correcting anomalies in on-orbit telemetry data of spacecraft, by performing sliding window processing on the historical normal telemetry data set of the spacecraft and determining the label data, it is possible to provide effective samples for the training of the time series prediction model, which facilitates the model to learn data features. The upper limit, mean, and lower limit prediction models are trained by combining quantile regression at high, medium, and low quantile levels, which can effectively quantify data uncertainty and overcome the defect of traditional time series prediction models that ignore data uncertainty. The current time series data is input into the trained model to obtain the prediction upper limit, mean, and lower limit, and a prediction interval is constructed. It can accurately determine whether the telemetry data to be detected is abnormal. The abnormal data is corrected to the predicted mean, which can make the data return to a reasonable range. The embodiment of the present invention can effectively realize the detection and correction of anomalies in on-orbit telemetry data of spacecraft, improve data accuracy and reliability, and provide strong support for spacecraft reliability analysis.

[0056] In one embodiment, performing sliding window processing on a historical normal telemetry data set of a spacecraft to obtain time series data samples includes: obtaining a historical normal telemetry data set of the spacecraft; performing sliding window processing on the historical normal telemetry data set according to a preset sliding window size and sliding step size to obtain a number of time series data samples; for each time series data, using the historical normal telemetry data within a predicted time interval from the next moment as corresponding label data; the interval length of the predicted time interval is the sliding step length.

[0057] like Figure 3 As shown in Figure 1, a schematic diagram of the construction of sequences and labels in multi-step time series prediction is provided. First, the time series data of a specific window size is obtained based on the sliding window method. The samples of the normal spacecraft telemetry data set accumulated historically are represented as {x1, x2, …, x n Define the sliding window size as ws and the sliding step as sl. Each sequence is composed of ws normal spacecraft telemetry data. Based on the time order, the latter sequence is obtained by moving the previous sequence by sl steps. Figure 2 As shown, the first sequence x 1 By {x1,x2,…,x ws}, the second sequence x 2 By {x sl+1 ,x sl+2 ,…,x sl+ws In addition, according to the sliding step size, label data can be further obtained. The first sequence x 1 The corresponding label is y 1 , which is composed of {x ws+1 ,x ws+2 ,…,x sl+ws}, the second sequence x 2 The corresponding label is y 2 , which is composed of {x sl+ws+1,x sl+ws+2 ,…,x sl+ws+sl}. By analogy, we can obtain the training data set {(x i ,y i )|i=1,2,…,n s}.

[0058] In one embodiment, the time series prediction model includes a recurrent neural network or a long short-term memory network. In this embodiment, the model type and model structure used by the time series prediction model can be defined according to the actual problem requirements.

[0059] In one embodiment, quantile regression of high, middle and low quantile levels is performed on a pre-constructed time series prediction model according to time series data samples and corresponding label data to obtain trained upper limit prediction model, mean prediction model and lower limit prediction model, including: constructing first, second and third quantile loss functions according to preset high quantile level, middle quantile level and low quantile level respectively; training the pre-constructed time series prediction model according to time series data samples, corresponding label data and first, second and third quantile loss functions respectively to obtain trained upper limit prediction model, mean prediction model and lower limit prediction model.

[0060] like Figure 4 As shown, a flow chart of a time series prediction model training method is provided. The present invention selects high quantile level, middle quantile level and low quantile level to train the model respectively, specifically τ up =0.975, τ low =0.025, τ mean = 0.5, thus obtaining the predicted mean and the upper and lower limits. After building the model structure, select appropriate parameters such as batch size, learning rate, iteration cycle, and optimization algorithm, and continuously update the model parameters during the training process to finally obtain the optimal parameter values. Corresponding to the three quantile levels τ up =0.975, τ low =0.025, τ mean =0.5, we need to calculate three different quantile loss functions: L(τ up ;θ), L(τ low ;θ) and L(τ mean ; θ), and obtain three corresponding upper limit prediction models, lower limit prediction models and mean prediction models: M up (x,τ up θ up ), M low (x,τ low θ low ) and M mean (x,τ mean θ mean ).

[0061] In one embodiment, the quantile loss function is:

[0062]

[0063] Among them, L(θ) is the quantile loss function, n s is the number of historical normal telemetry data in the training dataset, is the predicted value, y k is the kth historical normal telemetry data, τ is the quantile level, θ is the model parameter, is the indicator function.

[0064] After obtaining the three trained models based on historical data, time series prediction of real-time telemetry data can be achieved. Assume that the normal telemetry data for a period of time is x real ={x'1,x'2,…,x' ws}, the time series length is ws, and x real Input to model M low (x,τ low θ low ) and M up (x,τ up θ up ), the lower limit of prediction is obtained as The upper limit of the forecast is That is, the prediction result of the data with a time span of sl is obtained. The prediction lower limit and the prediction upper limit constitute the final prediction data interval, thereby effectively quantifying the data uncertainty. Similarly, x real Input to model M mean (x,τ mean θ mean ), the predicted mean can be obtained

[0065] In one embodiment, the method further includes: if the telemetry data to be detected belongs to the prediction data interval formed by the prediction upper limit and the prediction lower limit at the current moment, then the data to be detected is normal telemetry data.

[0066] In this embodiment, the telemetry data at time ws+1 is x' ws+1 , judge x' ws+1 Whether it is within the prediction interval, that is, judging Is it true? If it is within the prediction interval, then the data x' ws+1 If no exception occurs, continue to receive new data; otherwise, it is considered that data x' ws+1 Anomalies have occurred and further correction is required. Use the predicted mean to correct the abnormal data. That is, the abnormal telemetry data x' ws+1 Corrected to

[0067] In one embodiment, the method further includes: converting the abnormal telemetry data into normal telemetry data after correction.

[0068] like Figure 5 The diagram of the prediction, detection, and correction cycle shown in the figure shows that the detection and correction of spacecraft telemetry data changes are based on real-time data. The three processes of prediction, detection, and correction are continuously repeated and terminated when the data transmission stops. The entire process can effectively realize the detection and judgment of data, avoid the influence of abnormal data on data interpretation, and thus improve the accuracy of data. After the abnormal data is corrected, the data prediction for the next time period can be further realized based on the sliding window method. For example, the corrected data is defined as x revise ={x' sl+1 ,x' sl+2 ,…,x' sl+ws}, input to three models M up (x,τ up θ up ), M low (x,τ low θ low ) and M mean (x,τ mean θ mean ), obtain the predicted upper and lower limits and the predicted mean, Therefore, the new real-time data x' can be continuously ws+sl+1 Perform mutation detection and correction.

[0069] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0070] In one embodiment, Figure 6 As shown, a device for detecting and correcting abnormalities in on-orbit telemetry data of a spacecraft is provided, comprising:

[0071] The sample acquisition module 602 is configured to perform sliding window processing on the historical normal telemetry data set of the spacecraft to obtain a time series data sample; the label data of the time series data sample includes the historical normal telemetry data within the corresponding prediction time interval;

[0072] The model training module 604 is used to perform quantile regression at high, medium and low quantile levels on the pre-built time series prediction model based on the time series data samples and the corresponding label data to obtain a trained upper limit prediction model, a mean prediction model and a lower limit prediction model;

[0073] The data prediction module 606 is used to obtain the time series data corresponding to the telemetry data to be tested by the spacecraft at the current moment and the normal telemetry data in the previous time period, input the time series data into the trained upper limit prediction model, mean prediction model, and lower limit prediction model, and obtain the prediction upper limit, prediction mean, and prediction lower limit at each time step in the prediction time interval starting from the current moment;

[0074] The detection and correction module 608 is used to determine that if the telemetry data to be detected does not fall within the prediction data interval formed by the prediction upper limit and prediction lower limit at the current moment, the data to be detected is abnormal telemetry data, and correct the abnormal telemetry data to the prediction mean at the current moment.

[0075] In one embodiment, performing sliding window processing on a historical normal telemetry data set of a spacecraft to obtain time series data samples includes: obtaining a historical normal telemetry data set of the spacecraft; performing sliding window processing on the historical normal telemetry data set according to a preset sliding window size and sliding step size to obtain a number of time series data samples; for each time series data, using the historical normal telemetry data within a predicted time interval from the next moment as corresponding label data; the interval length of the predicted time interval is the sliding step length.

[0076] In one embodiment, the time series prediction model includes a recurrent neural network or a long short-term memory network.

[0077] In one embodiment, quantile regression at high, middle and low quantile levels is performed on a pre-constructed time series prediction model according to time series data samples and corresponding label data to obtain trained upper limit prediction model, mean prediction model and lower limit prediction model, including: constructing first, second and third quantile loss functions according to preset high quantile level, middle quantile level and low quantile level respectively; training the pre-constructed time series prediction model according to time series data samples, corresponding label data and first, second and third quantile loss functions respectively to obtain trained upper limit prediction model, mean prediction model and lower limit prediction model.

[0078] In one embodiment, the quantile loss function is:

[0079]

[0080] Among them, L(θ) is the quantile loss function, n s is the number of historical normal telemetry data in the training dataset, is the predicted value, y k is the kth historical normal telemetry data, τ is the quantile level, θ is the model parameter, is the indicator function.

[0081] In one embodiment, if the telemetry data to be detected falls within a prediction data interval formed by a prediction upper limit and a prediction lower limit at the current moment, the data to be detected is normal telemetry data.

[0082] In one embodiment, abnormal telemetry data is converted into normal telemetry data after correction.

[0083] Regarding the specific limitations of the spacecraft on-orbit telemetry data anomaly detection and correction device, please refer to the limitations of the spacecraft on-orbit telemetry data anomaly detection and correction method above, and will not be repeated here. The various modules in the above-mentioned spacecraft on-orbit telemetry data anomaly detection and correction device can be implemented in whole or in part through software, hardware, and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0084] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 7 As shown. The computer device includes a processor, memory, network interface, display screen and input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for detecting and correcting anomalies in telemetry data of a spacecraft in orbit is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0085] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0086] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps of the method in the above embodiment when executing the computer program.

[0087] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method in the above embodiment are implemented.

[0088] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0089] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0090] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and such modifications and improvements are intended to fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for detecting and correcting abnormal changes in on-orbit telemetry data of a spacecraft, characterized in that: The method comprises: Performing sliding window processing on a historical normal telemetry data set of a spacecraft to obtain a time series data sample; the label data of the time series data sample includes the historical normal telemetry data within the corresponding prediction time interval; Performing quantile regression at high, medium, and low quantile levels on the pre-built time series prediction model according to the time series data samples and the corresponding label data to obtain a trained upper limit prediction model, a mean prediction model, and a lower limit prediction model; Obtaining time series data corresponding to the telemetry data to be tested of the spacecraft at the current moment and the normal telemetry data of the previous time period, inputting the time series data into the trained upper limit prediction model, mean prediction model, and lower limit prediction model, respectively, to obtain the prediction upper limit, prediction mean, and prediction lower limit at each time step within the prediction time interval starting from the current moment; If the telemetry data to be detected does not fall within the prediction data interval formed by the prediction upper limit and prediction lower limit at the current moment, the data to be detected is abnormal telemetry data, and the abnormal telemetry data is corrected to the prediction mean at the current moment.

2. The method according to claim 1, characterized in that The time series data samples obtained by performing sliding window processing on the historical normal telemetry data set of the spacecraft include: Acquire historical normal telemetry data sets from spacecraft; Perform sliding window processing on the historical normal telemetry data set according to the preset sliding window size and sliding step size to obtain several time series data samples; For each time series data, the historical normal telemetry data within the predicted time interval from the next moment is used as the corresponding label data; the interval length of the predicted time interval is the sliding step length.

3. The method according to claim 1, characterized in that The time series prediction model includes a recurrent neural network or a long short-term memory network.

4. The method according to claim 1, wherein The method of performing quantile regression of high, medium and low quantile levels on the pre-built time series prediction model according to the time series data samples and the corresponding label data to obtain the trained upper limit prediction model, mean prediction model and lower limit prediction model includes: Construct the first, second and third quantile loss functions according to the pre-set high quantile level, middle quantile level and low quantile level respectively; The pre-built time series prediction model is trained according to the time series data samples, the corresponding label data, and the first, second, and third quantile loss functions to obtain a trained upper limit prediction model, a mean prediction model, and a lower limit prediction model.

5. The method according to claim 4, characterized in that The quantile loss function is: Among them, L(θ) is the quantile loss function, n s is the number of historical normal telemetry data in the training dataset, is the predicted value, y k is the kth historical normal telemetry data, τ is the quantile level, θ is the model parameter, is the indicator function.

6. The method according to claim 1, characterized in that The method further comprises: If the telemetry data to be detected falls within the prediction data interval formed by the prediction upper limit and the prediction lower limit at the current moment, the data to be detected is normal telemetry data.

7. The method according to claim 1, characterized in that The method further comprises: The abnormal telemetry data is converted into normal telemetry data after correction.

8. A device for detecting and correcting abnormalities in on-orbit telemetry data of a spacecraft, characterized in that: The device comprises: A sample acquisition module is used to perform sliding window processing on the historical normal telemetry data set of the spacecraft to obtain time series data samples; the label data of the time series data samples includes the historical normal telemetry data within the corresponding prediction time interval; A model training module is used to perform quantile regression at high, medium and low quantile levels on the pre-built time series prediction model according to the time series data samples and the corresponding label data, so as to obtain a trained upper limit prediction model, a mean prediction model and a lower limit prediction model; a data prediction module, configured to obtain time series data corresponding to the telemetry data to be tested of the spacecraft at the current moment and the normal telemetry data of the previous time period, input the time series data into the trained upper limit prediction model, mean prediction model, and lower limit prediction model, respectively, and obtain the prediction upper limit, prediction mean, and prediction lower limit at each time step within the prediction time interval starting from the current moment; The detection and correction module is used to determine that if the telemetry data to be detected does not fall within the prediction data interval formed by the prediction upper limit and prediction lower limit at the current moment, the data to be detected is abnormal telemetry data, and the abnormal telemetry data is corrected to the prediction mean at the current moment.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images