CN112257901A

CN112257901A - Abnormity early warning method and device for spacecraft

Info

Publication number: CN112257901A
Application number: CN202011019866.9A
Authority: CN
Inventors: 杨浩; 孙健; 房红征; 罗凯; 樊焕贞; 李蕊; 王信峰; 刘勇; 胡伟钢; 王德志
Original assignee: Beijing Aerospace Measurement and Control Technology Co Ltd
Current assignee: Beijing Aerospace Measurement and Control Technology Co Ltd
Priority date: 2020-09-24
Filing date: 2020-09-24
Publication date: 2021-01-22

Abstract

The application provides an anomaly early warning method and device for a spacecraft, wherein the method comprises the following steps: acquiring first monitoring data of the spacecraft, wherein the first monitoring data are monitoring data of the spacecraft in a first time period; inputting the first monitoring data into a target LSTM prediction model to obtain a first prediction result output by the target LSTM prediction model, wherein the target LSTM prediction model is obtained by training an initial LSTM prediction model by using second monitoring data, the second monitoring data are monitoring data of a spacecraft in a normal operation state in a second time period, and the first prediction result is used for indicating a first prediction parameter value of a target operation parameter of the spacecraft in a third time period after the end time of the first time period; and performing abnormal alarm on the spacecraft according to the first prediction parameter value and the first actual parameter value, wherein the first actual parameter value is an actual value of a target operation parameter of the spacecraft in a third time period.

Description

Abnormity early warning method and device for spacecraft

Technical Field

The application relates to the field of space equipment, in particular to an abnormity early warning method and device for a spacecraft.

Background

At present, health management becomes one of core technologies for prolonging the service life of complex equipment, and the service life and stability of a spacecraft serving as typical complex equipment can be remarkably prolonged, so that the use benefit of the spacecraft can be remarkably improved.

The spacecraft runs in outer space for a long time and is continuously interfered by outer space radiation, some complex and unknown faults are easier to occur than those of complex equipment on the ground, and perfect mathematical and mechanism models are difficult to describe all fault modes; moreover, the spacecraft can seriously influence the execution of various tasks when the spacecraft fails, and even can seriously influence the use of the civil, for example, the navigation satellite is abnormal, the performance and the function of the navigation are obviously reduced, the possibility of the failure of the spacecraft is prevented or predicted in advance, and the loss can be reduced to a certain extent. Therefore, performance anomaly prediction for a spacecraft is a very critical spacecraft health management requirement.

However, various factors cause that the abnormal prediction performance and accuracy of the spacecraft cannot meet the requirements, and the abnormal prediction performance and accuracy are mainly reflected in the following points:

1) the spacecraft has a complex structure, the coupling among parameters is high, the characteristics which can really represent the abnormal degree of a certain part of the spacecraft are difficult to extract, the change of the parameters can be caused by noise or working condition switching, the relation between the parameters and the change of the parameters is difficult to distinguish when a fault occurs, and the coupling and the complexity bring great influence on the abnormal early warning performance;

2) the spacecraft is used as high-safety equipment, stability and safety are used as cores during design, so that multiple redundancy measures are generally adopted in the operation process to ensure the normal operation of the equipment, the probability of the spacecraft being in fault is low due to the setting, and particularly the probability of the spacecraft being in fault for multiple times is low. Therefore, data acquired by the spacecraft are basically normal data, the data lack fault characteristics, and the traditional fault prediction mode cannot meet the requirements;

3) some equipment of the spacecraft is a long-term accumulated process from normal to abnormal to final failure, and short-time observation and analysis can not obtain failure analysis results and needs to be analyzed from a long time scale. The prediction accuracy of a new energy anomaly prediction mode in the related technology is insufficient in a scene with a long time scale.

Therefore, the problem that the abnormity cannot be accurately predicted due to the lack of data with fault characteristics exists in the abnormity early warning mode of the spacecraft in the related technology.

Disclosure of Invention

The embodiment of the application provides an anomaly early warning method and device for a spacecraft, and aims to at least solve the problem that an anomaly cannot be accurately predicted due to lack of data with fault characteristics in an anomaly early warning mode of the spacecraft in the related technology.

According to an aspect of an embodiment of the present application, there is provided an anomaly early warning method for a spacecraft, including: acquiring first monitoring data of a spacecraft, wherein the first monitoring data are monitoring data of the spacecraft in a first time period; inputting the first monitoring data into a target LSTM prediction model to obtain a first prediction result output by the target LSTM prediction model, wherein the target LSTM prediction model is obtained by training an initial LSTM prediction model by using second monitoring data, the second monitoring data are monitoring data of the spacecraft in a normal operation state in a second time period, and the first prediction result is used for indicating a first prediction parameter value of a target operation parameter of the spacecraft in a third time period after the end time of the first time period; and performing abnormal alarm on the spacecraft according to the first prediction parameter value and a first actual parameter value, wherein the first actual parameter value is an actual value of the target operation parameter of the spacecraft in the third time period.

Optionally, the performing an abnormal alarm on the spacecraft by using the first predicted parameter value and the first actual parameter value includes: determining that the spacecraft is operating normally when the difference between the first predicted parameter value and the first actual parameter value is less than or equal to a target error; determining that the spacecraft is abnormal in operation under the condition that the difference value between the first prediction parameter value and the first actual parameter value is larger than a prime number target error; and sending an alarm message to a mobile terminal of a target object, wherein the alarm message is used for indicating that the spacecraft is abnormal.

Optionally, before performing an abnormal warning on the spacecraft according to the first predicted parameter value and the first actual parameter value, the method further includes: inputting third monitoring data into the target LSTM prediction model to obtain a second prediction result output by the target LSTM prediction model, wherein the third monitoring data are monitoring data of the spacecraft in a normal operation state in a third time period, and the second prediction result is used for indicating a second prediction parameter value of the target operation parameter of the spacecraft in a fifth time period after the end time of the third time period; determining an average mean square error between the second predicted parameter value and a second actual parameter value as the target error, wherein the second actual parameter value is an actual value of the target operating parameter of the spacecraft over the fifth time period.

Optionally, before inputting the first monitoring data to the target LSTM prediction model, the method further comprises: acquiring monitoring data of the spacecraft in a normal operation state in the second time period to obtain second monitoring data; constructing training samples of the initial LSTM prediction model using the second monitoring data, wherein the training samples comprise: target sub-monitoring data within a target sub-time period of the second time period, and a third actual parameter value of the target operating parameter of the spacecraft within a sixth time period after an end time of the target sub-time period; and training the initial LSTM prediction model by using the training sample to obtain the target LSTM prediction model.

Optionally, training the initial LSTM prediction model using the training samples to obtain the target LSTM prediction model includes: inputting the target sub-monitoring data into the initial LSTM prediction model to obtain a third prediction result output by the initial LSTM prediction model, wherein the third prediction result is used for indicating a third prediction parameter value of the target operation parameter of the spacecraft in the sixth time period; and adjusting the model parameters of the initial LSTM prediction model according to the difference value between the third prediction parameter value and the third actual parameter value to obtain the target LSTM prediction model, wherein the difference value between the fourth prediction parameter value of the target operation parameter and the third actual parameter value output by the target LSTM prediction model is smaller than or equal to a target difference value.

According to another aspect of the embodiments of the present application, there is also provided an anomaly early warning device for a spacecraft, including: the system comprises a first acquisition unit, a second acquisition unit and a control unit, wherein the first acquisition unit is used for acquiring first monitoring data of a spacecraft, and the first monitoring data is monitoring data of the spacecraft in a first time period; a first input unit, configured to input the first monitoring data into a target LSTM prediction model, so as to obtain a first prediction result output by the target LSTM prediction model, where the target LSTM prediction model is obtained by training an initial LSTM prediction model using second monitoring data, the second monitoring data is monitoring data of the spacecraft in a normal operation state in a second time period, and the first prediction result is used to indicate a first prediction parameter value of a target operation parameter of the spacecraft in a third time period after an end time of the first time period; and the warning unit is used for performing abnormal warning on the spacecraft according to the first prediction parameter value and a first actual parameter value, wherein the first actual parameter value is an actual value of the target operation parameter of the spacecraft in the third time period.

Optionally, the alarm unit includes: a first determining module, configured to determine that the spacecraft operates normally when a difference between the first predicted parameter value and the first actual parameter value is less than or equal to a target error; the second determining module is used for determining that the spacecraft runs abnormally under the condition that the difference value between the first prediction parameter value and the first actual parameter value is larger than a prime number target error; and sending an alarm message to a mobile terminal of a target object, wherein the alarm message is used for indicating that the spacecraft is abnormal.

Optionally, the apparatus further comprises: a second input unit, configured to, before performing an abnormal warning on the spacecraft according to the first predicted parameter value and the first actual parameter value, input third monitoring data into the target LSTM prediction model to obtain a second prediction result output by the target LSTM prediction model, where the third monitoring data is monitoring data of the spacecraft in a normal operation state in a fourth time period, and the second prediction result is used to indicate a second predicted parameter value of the target operation parameter of the spacecraft in a fifth time period after an end time of the third time period; a determining unit, configured to determine an average mean square error between the second predicted parameter value and a second actual parameter value as the target error, where the second actual parameter value is an actual value of the target operating parameter of the spacecraft in the fifth time period.

Optionally, the apparatus further comprises: the second acquisition unit is used for acquiring the monitoring data of the spacecraft in a normal operation state in the second time period before the first monitoring data are input into the target LSTM prediction model, so as to obtain second monitoring data; a construction unit configured to construct training samples of the initial LSTM prediction model using the second monitoring data, wherein the training samples include: target sub-monitoring data within a target sub-time period of the second time period, and a third actual parameter value of the target operating parameter of the spacecraft for a sixth time period after an end time of the target sub-time period; and the training unit is used for training the initial LSTM prediction model by using the training samples to obtain the target LSTM prediction model.

Optionally, the apparatus further comprises: the training unit includes: an input module, configured to input the target sub-monitoring data into the initial LSTM prediction model to obtain a third prediction result output by the initial LSTM prediction model, where the third prediction result is used to indicate a third prediction parameter value of the target operating parameter of the spacecraft in the sixth time period; and the adjusting module is used for adjusting the model parameters of the initial LSTM prediction model according to the difference value between the third prediction parameter value and the third actual parameter value to obtain the target LSTM prediction model, wherein the difference value between a fourth prediction parameter value of the target operation parameter and the third actual parameter value output by the target LSTM prediction model is smaller than or equal to a target difference value.

According to a further embodiment of the present invention, a computer-readable storage medium is also provided, in which a computer program is stored, wherein the computer program is configured to carry out the steps of any of the above-described method embodiments when executed.

According to yet another embodiment of the present invention, there is also provided an electronic device, including a memory in which a computer program is stored and a processor configured to execute the computer program to perform the steps in any of the above method embodiments.

According to the method and the device, an abnormal early warning mode based on an LSTM prediction model is adopted, and first monitoring data of the spacecraft are obtained, wherein the first monitoring data are monitoring data of the spacecraft in a first time period; inputting the first monitoring data into a target LSTM prediction model to obtain a first prediction result output by the target LSTM prediction model, wherein the target LSTM prediction model is obtained by training an initial LSTM prediction model by using second monitoring data, the second monitoring data are monitoring data of a spacecraft in a normal operation state in a second time period, and the first prediction result is used for indicating a first prediction parameter value of a target operation parameter of the spacecraft in a third time period after the end time of the first time period; and carrying out abnormal alarm on the spacecraft according to the first prediction parameter value and the first actual parameter value, wherein, the first actual parameter value is an actual value of a target operating parameter of the spacecraft over a third time period, since the LSTM prediction model is trained based on normal data, the LSTM prediction model predicts a predicted value of a target operating parameter (e.g., flywheel current) over a period of time under normal conditions, compares the predicted value with an actual value of the target operating parameter over the period of time, thereby reversely deducing whether the current system is abnormal or not, realizing the purpose of predicting the abnormality based on normal data, and daily operation monitoring data of the spacecraft are used, so that the technical effects of improving the convenience of data acquisition and improving the accuracy of spacecraft abnormal early warning can be achieved, and the problem that the abnormity of the spacecraft in the related technology cannot be accurately predicted due to the lack of data with fault characteristics in an abnormity early warning mode is solved.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention.

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive exercise.

Fig. 1 is a schematic diagram of a hardware environment of an alternative spacecraft anomaly early warning method according to an embodiment of the invention;

fig. 2 is a schematic flow chart of an alternative spacecraft anomaly early warning method according to an embodiment of the present application;

FIG. 3 is a schematic diagram of an alternative long short term memory network according to an embodiment of the present application;

FIG. 4 is a schematic flow chart diagram illustrating an alternative spacecraft anomaly early warning method according to an embodiment of the present application;

FIG. 5 is a flow chart of an alternative model building method according to an embodiment of the present application;

fig. 6 is a block diagram of an alternative anomaly early warning device for a spacecraft according to an embodiment of the present application;

fig. 7 is a block diagram of an alternative electronic device according to an embodiment of the present application.

Detailed Description

The invention will be described in detail hereinafter with reference to the accompanying drawings in conjunction with embodiments. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.

It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.

According to one aspect of the embodiment of the application, an abnormity early warning method of a spacecraft is provided. Optionally, in this embodiment, the anomaly early warning method for the spacecraft may be applied to a hardware environment formed by the terminal 101 (e.g., anomaly early warning device, model training device) and the server 103 as shown in fig. 1. As shown in fig. 1, a server 103 is connected to a terminal 101 through a network, which may be used to provide services (such as game services, application services, etc.) for the terminal or a client installed on the terminal, and a database may be provided on the server or separately from the server for providing data storage services for the server 103, and the network includes but is not limited to: the terminal 101 is not limited to a PC, a mobile phone, a tablet computer, and the like. The anomaly early warning method for the spacecraft can be executed by the server 103, can also be executed by the terminal 101, and can also be executed by the server 103 and the terminal 101 together. The terminal 101 may execute the anomaly early warning method for the spacecraft according to the embodiment of the present application by a client installed thereon.

Taking an example of an anomaly early warning device (the anomaly early warning device may be a terminal device or a similar electronic device), in this embodiment, an anomaly early warning method for a spacecraft running on the anomaly early warning device is provided, and fig. 2 is a schematic flow diagram of an optional anomaly early warning method for a spacecraft according to an embodiment of the present application, and as shown in fig. 2, the flow includes the following steps:

step S202, acquiring first monitoring data of the spacecraft, wherein the first monitoring data are monitoring data of the spacecraft in a first time period;

step S204, inputting first monitoring data into a target LSTM prediction model to obtain a first prediction result output by the target LSTM prediction model, wherein the target LSTM prediction model is obtained by training an initial LSTM prediction model by using second monitoring data, the second monitoring data are monitoring data of the spacecraft in a normal operation state in a second time period, and the first prediction result is used for indicating a first prediction parameter value of a target operation parameter of the spacecraft in a third time period after the end time of the first time period;

and S206, performing abnormal alarm on the spacecraft according to the first prediction parameter value and the first actual parameter value, wherein the first actual parameter value is an actual value of a target operation parameter of the spacecraft in a third time period.

Optionally, the main body of the above steps may be a terminal device, etc., but is not limited thereto, and may also be a server or other electronic devices, which is not limited in this embodiment.

Through the steps, first monitoring data of the spacecraft are obtained, wherein the first monitoring data are monitoring data of the spacecraft in a first time period; inputting the first monitoring data into a target LSTM prediction model to obtain a first prediction result output by the target LSTM prediction model, wherein the target LSTM prediction model is obtained by training an initial LSTM prediction model by using second monitoring data, the second monitoring data are monitoring data of a spacecraft in a normal operation state in a second time period, and the first prediction result is used for indicating a first prediction parameter value of a target operation parameter of the spacecraft in a third time period after the end time of the first time period; and performing abnormity warning on the spacecraft according to the first prediction parameter value and the first actual parameter value, wherein the first actual parameter value is an actual value of a target operation parameter of the spacecraft in a third time period, so that the problem that abnormity cannot be accurately predicted due to lack of data with fault characteristics in an abnormity early warning mode of the spacecraft in the related technology is solved, the convenience of data acquisition is improved, and the accuracy of abnormity early warning of the spacecraft is improved.

In the technical solution provided in step S202, first monitoring data of the spacecraft is obtained, where the first monitoring data is monitoring data of the spacecraft in a first time period.

The anomaly early warning method for the spacecraft in the embodiment of the application can be applied to anomaly early warning of the spacecraft in a space system, and can also be applied to other scenes with equipment anomaly prediction requirements, for example, anomaly early warning of other high-safety equipment, and the method is not limited in the embodiment.

The anomaly early warning device may acquire first monitoring data of the spacecraft, where the first monitoring data may be monitoring data of the spacecraft in a first time period. The monitoring data may be: the daily operation monitoring data of the spacecraft does not need to be processed for the second time, the data is easy to obtain, and the proportion of the requirement on the fault data is not high.

The first monitoring data of the spacecraft may be monitoring data of a certain operating parameter or certain operating parameters of the spacecraft, or may be monitoring data of a certain component or certain operating parameters of the spacecraft. The component may be a component that has a greater impact on the operation of the spacecraft, such as a flywheel or the like. In this embodiment, the on-orbit operation monitoring data of the spacecraft flywheel is taken as an example for explanation, and for other components, the abnormality prediction can be performed in the same or similar manner as in this embodiment.

The operating parameter may be one or more operating parameters of the flywheel, and may include, but is not limited to, at least one of: attitude angle X, attitude angle Y, attitude angle Z, angular rate X, angular rate Y, angular rate Z, attitude orbit control mode of operation, X flywheel shaft temperature, Y flywheel shaft temperature, Z flywheel shaft temperature, S flywheel shaft temperature, Y flywheel current, Z flywheel current, S flywheel current. The first time period may be a certain period of time in the past, and the duration of the first time period may be: 5min, 10min, half an hour, one hour, etc., which are not limited in this embodiment.

In the technical solution provided in step S204, the first monitoring data is input to the target LSTM prediction model, and a first prediction result output by the target LSTM prediction model is obtained.

A target LSTM (Long Short-Term Memory network) prediction model may be run on the anomaly prediction device, where the target LSTM prediction model is obtained by training an initial LSTM prediction model using second monitoring data, and the second monitoring data is monitoring data of the spacecraft in a normal operation state in a second time period.

LSTM is a long short term memory network, a time-recursive neural network, suitable for processing and predicting relatively long-spaced and delayed events in a time series. Unlike RNN (Recurrent Neural Network), LSTM incorporates a "processor" in the algorithm that determines whether information is useful or not, and the structure of the processor action is called a cell. Three gates, namely an input gate, a forgetting gate and an output gate (as shown in fig. 3), are placed in one cell. A message enters the LSTM network and may be determined to be useful based on rules. Only the information which is in accordance with the algorithm authentication is left, and the information which is not in accordance with the algorithm authentication is forgotten through a forgetting door. The LSTM updates the state (cell state) of the cell by completing addition and deletion of information by an output gate, a forgetting gate, and an input gate.

The model inputs for the target LSTM prediction model are one or more monitored parameters associated with the spacecraft flywheel and the model inputs are target operating parameters. According to prior analysis, when the flywheel is abnormal, the current change of the flywheel has high correlation with the abnormal degree, and the current change can be used as the output of the model. Other monitored parameters of the flywheel may be used as model inputs to predict flywheel current changes, and alternatively the target operating parameter may be X flywheel current.

As an example, the model inputs for the target LSTM predictive model may be as shown in Table 1.

TABLE 1

Input features	Description of the invention	Typical value
			1	Angular rate X	-0.000061
2	Angular rate Y	0.000061
			3	Angular rate Z	-0.000061
4	Attitude and orbit control working mode	187
			5	Y flywheel current	0.024
6	Z flywheel current	0.048
			7	S flywheel current	0

The model output of the target LSTM prediction model may be as shown in table 2.

TABLE 2

Output characteristics	Description of the invention	Typical value
			1	Current of X flywheel	0.012

In this embodiment, the long and short term memory network LSTM is used as a main model, which can deeply mine the time sequence characteristics of the monitoring data, and has more efficient prediction performance compared to the conventional machine learning method regardless of the time sequence of the parameters.

The target LSTM prediction model is obtained by training an initial LSTM prediction model by using normal data (second monitoring data, namely the monitoring data of the spacecraft in a normal operation state in a second time period), and the model is trained according to the normal data, so that the obtained model can predict the current and future changes of related data under a normal condition.

As an optional embodiment, before the first monitoring data is input into the target LSTM prediction model, the monitoring data of the spacecraft in the normal operation state in the second time period may be acquired to obtain second monitoring data; constructing a training sample of the initial LSTM prediction model by using the second monitoring data; and training the initial LSTM prediction model by using the training samples to obtain a target LSTM prediction model.

Before using the target LSTM prediction model, the LSTM prediction model may be constructed, but is not limited to, as follows: the LSTM layer is added using the API (Application Programming Interface) model library of Keras (an open source artificial neural network library written by Python): add (LSTM (units, (40, number of parameters))), where units (4, 8, 32, 128, etc.) can try different values and adjust by comparing their final test results; add (full join) layer add (full join) using model.

The model structure of the constructed LSTM prediction model may adopt a Deep learning framework combining LSTM + DNN (Deep Neural Networks), and the model structure may be as shown in table 3.

TABLE 3

Layer (type)	Output size	Number of parameters
			lstm_1(LSTM)	(None,20,128)	70144
lstm_2(LSTM)	(None,20,54)	39528
			lstm_3(LSTM)	(None,54)	23544
dense_1(Dense)	(None,10)	550

As can be seen from Table 3, the number of all parameters in the LSTM prediction model is 133,766, the number of parameters to be tuned is 133,766, and there are no parameters that need not be tuned.

For the constructed LSTM prediction model (initial LSTM prediction model), second monitoring data may be obtained, so that model training is performed on the initial LSTM prediction model using the second monitoring data to obtain a target LSTM prediction model. The device for performing model training may be a model training device, and the model training device and the abnormality early warning device may be the same device or different devices, which is not limited in this embodiment.

The second monitoring data may be monitoring data of the spacecraft in a normal operation state in a second time period, for example, the on-orbit operation monitoring data of the flywheel of the spacecraft may be acquired, and the time period may be 1 year or more than 1 year, and the second monitoring data is acquired according to the on-orbit operation monitoring data of the flywheel. For example, the target LSTM prediction model is a flywheel fault prediction model, and can obtain flywheel on-orbit operation monitoring signal data of a spacecraft in a spacecraft control system in a whole year (which may be the same year or across years), where the flywheel may be abnormal in some months, and the rest is normal data, or all may be normal data.

Based on the existing monitoring data (e.g., the flywheel on-orbit operation monitoring data), a data set can be constructed, which includes: a training dataset of the initial LSTM prediction model, which may include: training samples, the number of which can be one or more. Each training sample may include: monitoring data (e.g., target sub-monitoring data) within a certain sub-period (e.g., target sub-period) of the second time period; a third actual parameter value of the target operating parameter of the spacecraft for a certain time period (e.g. a sixth time period) after the end time of the sub-time period.

For example, the training samples may include: monitoring data for a certain time period (e.g., 10min) of the flywheel, and a change in the flying current for a period of time (e.g., 5min) after the end of the time period.

After the training samples are obtained, the model training device may train the initial LSTM prediction model using the training samples to obtain a target LSTM prediction model.

The LSTM prediction model is an unsupervised learning model, the unsupervised learning model is adopted, manual participation in data processing is not needed, artificial error factors can be eliminated by constructing and using the model, and the model training efficiency and the model training accuracy are improved.

According to the embodiment, the monitoring data of the spacecraft in the normal operation state in a certain time period are used for constructing the training sample and training the LSTM prediction model, so that the model training efficiency and the model training accuracy can be improved.

As an alternative embodiment, training the initial LSTM prediction model using the training samples to obtain the target LSTM prediction model may include: inputting the target sub-monitoring data into the initial LSTM prediction model to obtain a third prediction result output by the initial LSTM prediction model, wherein the third prediction result is used for indicating a third prediction parameter value of the target operation parameter of the spacecraft in a sixth time period; and adjusting the model parameters of the initial LSTM prediction model according to the difference value between the third prediction parameter value and the third actual parameter value to obtain a target LSTM prediction model, wherein the difference value between the fourth prediction parameter value and the third actual parameter value of the target operation parameter, which is output by the target LSTM prediction model, is smaller than or equal to the target difference value.

The number of training samples may be multiple, and the way of training the initial LSTM prediction model using the training samples may be: sequentially inputting the sub-monitoring data in each training sample into an LSTM prediction model to obtain a prediction parameter value of a target operation parameter output by the LSTM prediction model; and then, adjusting the model parameters of the LSTM prediction model according to the errors of the prediction parameter values and the actual parameter values, and the like, and finishing the training when the loss function of the model is converged through one or more rounds of training to obtain the target LSTM prediction model.

According to the convergence condition, the difference between the predicted parameter value and the actual parameter value of the target operating parameter, which is output by the target LSTM prediction model, is less than or equal to the target difference, which may be preset, or may be determined according to the target LSTM prediction model after the model converges, which is not specifically limited in this embodiment.

In model training, the batch _ size (i.e., sample size) and the number of iterations may be determined, and the model may be iteratively optimized by a loss function until the value of the loss function is not decreasing, or a preset condition is met. The MSE (Mean Square Error) between the predicted parameter value and the actual parameter value may be used as a loss function for model training, or other loss functions may be used, which is not limited in this embodiment.

According to the embodiment, the parameters of the model are adjusted by comparing the predicted parameter values and the actual parameter values of the model, so that the accuracy of model parameter adjustment can be ensured, and the efficiency of model training is improved.

Inputting the first monitoring data into the target LSTM prediction model, obtaining a first prediction result output by the target LSTM prediction model, wherein the first prediction result can be used for indicating a first prediction parameter value of the target operation parameter of the spacecraft in a third time period after the end time of the first time period.

The end time of the first time period is the first time, and the third time period is a time period after the first time. For example, the first period is a period of 10min in duration, and the third period is a period of 5min in duration after the end time of the first period.

In the technical scheme provided in step S206, an abnormal warning is given to the spacecraft according to the first predicted parameter value and the first actual parameter value.

And the actual value of the target operation parameter of the spacecraft in the third time period is the first actual parameter value. The abnormality early warning device can obtain the first actual value and perform abnormality warning on the spacecraft according to the first prediction parameter value and the first actual parameter value. For example, according to the difference degree of the first prediction parameter value and the first actual parameter value, the spacecraft is subjected to abnormal alarm.

When the equipment (such as the spacecraft) normally operates, the LSTM can be trained by data to fit some rules and mechanisms of the internal operation of the equipment, so that the future change trend of the equipment can be predicted. When an abnormality occurs inside the device, internal rules and mechanisms also change, and the change is reflected in real changes of the future device. The model trained based on normal data still represents the rules and mechanisms of normal equipment. The difference degree between the predicted value and the actual value can reflect the abnormal degree of the equipment, and further realize the early warning of the abnormality.

As an alternative example, the anomaly prediction algorithm uses LSTM as a core model to predict the current change over a future period of time based on the existing current signal. And comparing the predicted value with the actual value of the change of the current, solving the predicted error to form an error curve, and when the error precision of the model prediction deviates to a certain degree, considering that the equipment is abnormal, and further performing abnormal early warning.

As an alternative embodiment, performing an abnormal alarm on the spacecraft according to the first predicted parameter value and the first actual parameter value may include: determining a mean square error between the first predicted parameter value and the first actual parameter value; determining that the spacecraft operates normally under the condition that the difference value between the mean square error and the target error is within the target difference value range; determining that the spacecraft operates abnormally under the condition that the difference between the mean square error and the target error exceeds the target difference range; and sending an alarm message to the mobile terminal of the target object, wherein the alarm message is used for indicating that the spacecraft is abnormal.

After obtaining the first predicted parameter value, the abnormality warning apparatus may determine a difference between the first predicted parameter value and the first actual parameter value. The predicted parameter value and the actual parameter value may be indicative of different profiles of the flywheel current over the same time period, the difference of which may be represented by a mean square error.

The predicted parameter value can be the change of the flywheel current in a future period (for example, a third period) predicted based on the LSTM prediction model, the change of the current prediction precision MSE can be calculated in real time, the MSE is compared with a normal error (target error), and the deviation value of the MSE and the normal error can be used as an evaluation index of the abnormal early warning. For example, if the MSE is within a normal error range (the difference from the target error is within a target difference range), it may be determined that the spacecraft is operating properly and an exception handling procedure is not triggered.

If the MSE exceeds the normal error range (the difference value between the MSE and the target error exceeds the target difference range), the spacecraft operation can be determined to be abnormal, and an abnormal processing flow is triggered. The exception handling process may be sending an alarm message to a mobile terminal of a target object, where the target object may be an object capable of handling an exception, may also be an object coordinating exception handling, and may also be another object, which is not limited in this embodiment. The warning message may be used to indicate that the spacecraft is abnormal, and may also be at least one of the following: the time period of the occurrence of the anomaly, the identification of the anomalous spacecraft, the current state of the anomalous spacecraft, and the like, which are not limited in this embodiment.

Through the embodiment, different processing is performed on the spacecraft according to different states of the spacecraft, the efficiency of exception handling can be improved, and reasonable distribution of human resources and the like is guaranteed.

As an optional embodiment, before performing an abnormal alarm on the spacecraft according to the first predicted parameter value and the first actual parameter value, third monitoring data may be input into the target LSTM prediction model to obtain a second prediction result output by the target LSTM prediction model, where the third monitoring data is monitoring data of the spacecraft in a normal operation state in a fourth time period, and the second prediction result is used to indicate a second predicted parameter value of the target operation parameter of the spacecraft in a fifth time period after an end time of the third time period; and determining the average mean square error between the second prediction parameter value and the second actual parameter value as a target error, wherein the second actual parameter value is the actual value of the target operation parameter of the spacecraft in the fifth time period.

In order to determine the range of the normal error, the third monitoring data of the spacecraft in the normal operation state in the fourth time period can be input into the target LSTM prediction model, and a second prediction result output by the target LSTM prediction model is obtained. If the number of the third monitoring data is multiple, a second prediction result corresponding to each third monitoring data can be obtained respectively.

After obtaining the second prediction result, an average mean square error between the second prediction parameter value and the second actual parameter value may be obtained and determined as the target error. The target error may be determined based on the mean square error corresponding to each third monitored data and may be an average of the mean square errors corresponding to each third monitored data, or a target error range of the mean square error may be determined, and the target error range may be [0, MSE_max]And performing an abnormal alarm when the mean square error between the predicted parameter value and the actual parameter value exceeds the target error range, which is not limited in this embodiment.

The third monitoring data and the second monitoring data may be obtained based on the same on-orbit operation monitoring data of the spacecraft flywheel. Optionally, in this embodiment, in order to be able to detect and evaluate the accuracy of the LSTM algorithm, when constructing the data set, the constructed data set further includes: a test data set of the LSTM prediction model, the test data in the test data set being usable to determine the target error and to validate the model.

The data set may be constructed by segmenting the data set and transforming the data structure. The data set may be partitioned into the training data set and the test data set at a target scale (partition scale), which may be a preconfigured scale, e.g., 8:2, (2/3): 1/3. The test samples in the test data set are similar to the training samples, and are not described herein. Alternatively, in order to improve the operation efficiency, data preprocessing, normalization, or the like may be performed before dividing the data.

For data normalization, the data set may be normalized, scaling the data so that all feature data is in the range of 0 to 1, facilitating comparison and weighting of indices in different units or magnitudes.

For data reconstruction and segmentation, normalized data can be reconstructed into a new data set in a fixed window step size, the window size can be set to be 60 pieces of data (40+20), and the data are taken out according to a set sequence and combined into the new data set. For example, assume that a dataset D is D1 to dn (D1 to dn may represent D1, D2,. and. dn), and t1 is D1 to D60; t2 ═ d2 to d 61; ...; d (n-59) -dn, and then T1-T (n-59) of the new data set; the new data set T is divided into training data T _ train and test data T _ test in a 2:1 ratio.

The training data T _ train and the test data T _ test are divided into training input data X _ train, training output data Y _ train, test input data X _ test, and test output data Y _ test, respectively. The division method is as follows:

each set of data T-da to d (a +59) in T _ train and T _ test may be divided into tx (da to d (a +39)) and ty (d (a +40) to d (a +59)), and combined as follows:

x _ train is tx 1-txi, Y _ train is ty 1-tyi, and i is the total number of training samples;

x _ test-tx (i +1) -txj, Y _ test-ty (i +1) -tyj, j being the total number of samples in the data set.

As an example, the data set D is D1 to D119 (here, this is an example, the actual data amount may be large), then t1 is D1 to D60; t2 ═ d2 to d 61; ...; t (60) ═ d (60) to d119, and T ═ T1 to T (60). The data set T is divided into training data T _ train (T1-T (40)) and test data T _ test (T41-T (60)) at a ratio of 2: 1. Each set of data (training samples) in T _ train can be divided into: training input data X _ train and training output data Y _ train, for a total of 40 pairs: (d 1-d 40), (d 41-d 60); (d 2-d 41), (d 42-d 61); ...; (d 40-d 79) and (d 80-d 99). Each set of data (test samples) in T _ test can be divided into: the test input data X _ test and the test output data Y _ test, for a total of 20 pairs: (d 41-d 80), (d 81-d 100); (d 42-d 81), (d 82-d 101); ...; (d 60-d 99), (d 100-d 119).

Although the window size of the sliding window is 60, the window step size is 1, and the division ratio is 2:1 in the present embodiment, the present invention is not limited to this, but the present invention is also applicable to sliding windows of other window sizes and/or window steps, and is also applicable to other division ratios (for example, 8: 2).

For example, the training data (i.e., training samples) for the model has a total of 996416, which are: n _ data _ x (996416,8), n _ data _ y (996416, 1). The test data for the model had a total of 172493 pieces: f _ data _ x (172493,8), f _ data _ y (172493,1), where 8 is the number of model input parameters and 1 is the number of model output parameters.

After the model training is finished, the trained model can be verified by using a test set, and whether the MSE of the model loss curve is converged or not is evaluated so as to verify the effectiveness and generalization capability of the model. When the model verification is carried out, the prediction can be carried out on normal data and abnormal data, and the MSE of the normal data prediction is smaller than that of the abnormal data prediction.

Based on the LSTM prediction model, an early warning model can be constructed. For example, the average prediction error init _ MSE (average mean square error) of the LSTM prediction model may be determined as the normal error of the model prediction. And predicting the current change in a future period of time based on the prediction model, and calculating the change of the current prediction precision MSE in real time. The init _ MSE is compared with the MSE predicted in real time, and the deviation value between the init _ MSE and the MSE predicted in real time is used as an evaluation index of the abnormal early warning. If the deviation is not large (within the range of the target difference value), the spacecraft is determined to normally operate, and if the deviation is large (beyond the range of the target difference value), the spacecraft is determined to abnormally operate, and abnormal early warning is carried out.

Through this embodiment, through the monitoring data that uses the spacecraft normal operation confirm the mean square error to confirm the mean square error as the target error, can improve the convenience that the target error acquireed, promote the accuracy that the target error is confirmed.

The anomaly early warning method for the spacecraft is explained by combining an optional example. The abnormality warning method in this example can also be applied to fault diagnosis and Health Management (PHM) of other equipment.

As shown in fig. 4, the anomaly early warning method for a spacecraft in this example may include the following steps:

in step S402, a training data set (normal data) is acquired.

The LSTM prediction model is used for anomaly prediction in this example. With the LSTM as a core, an anomaly early warning model of the spacecraft can be constructed, and the early warning model can be constructed based on an LSTM prediction model.

Step S404, training the LSTM model.

The LSTM prediction model may be trained using training data in the training dataset to obtain a trained LSTM model.

Step S406, a test data set of the LSTM model is obtained, and the LSTM model is verified by using data in the test data set.

The trained LSTM model may be validated using test data in the test data set, and the normal error of the LSTM model may also be determined.

For the construction process of the LSTM prediction model including the training process and the verification process of the model, as shown in fig. 5, the construction process of the LSTM prediction model may include the following steps:

step S502, data preprocessing.

For example, the on-orbit operation monitoring signal data of the flywheel of the spacecraft for a whole year can be preprocessed, standardized and the like.

Step S504, the data set is divided into a training data set and a test data set (random sampling, division ratio 8: 2).

Step S506, training data in the training data set is acquired.

Step S508, test data in the test data set is obtained.

In step S510, the LSTM prediction model is trained using the training data (input of 10 minutes of data is preset, and output of 5 minutes of data is preset).

Step S512, the prediction model is verified by using the test data.

Step S514, the prediction model is evaluated by using the test data.

When the predictive model is verified, a normal error (MSE, e.g., average predictive error) of the predictive model may be determined, which may be used for subsequent abnormal predictions.

Step S408, obtaining the parameter changes of the current and future periods of time output by the prediction model.

During abnormal prediction, monitoring data of the spacecraft flywheel within a period of time can be input into the LSTM prediction model, parameter changes of the current and future periods of time output by the prediction model are obtained, and a prediction result of the prediction model is obtained.

In step S410, it is determined that the prediction actually matches, and step S412 is performed.

Step S412, determining that the system is operating normally.

In step S414, it is determined that the actual difference from the prediction is large, and step S416 is performed.

In step S416, it is determined that the system is operating abnormally.

With the present example, the LSTM prediction model is trained on normal data, and the resulting LSTM prediction model can predict current and future changes in the relevant data under normal conditions. When the system is abnormal, the LSTM prediction model can deviate, when the deviation is large, the LSTM prediction model can be determined to be abnormal in fitting with the current system, the LSTM prediction model is obtained by training based on normal data, so that the current system can be reversely deduced to be abnormal, and further abnormal early warning can be realized, because the system is based on daily operation monitoring data of a spacecraft, secondary processing of the data is not needed, the data is easy to obtain, and the proportion required by fault data is not high; an unsupervised learning model is adopted, manual participation in data processing is not needed, and the construction and use of the model can get rid of human error factors; by adopting the LSTM as the main model, the time sequence characteristics of the monitoring data can be deeply mined, and the method has more efficient prediction performance compared with the existing machine learning method which does not consider the time sequence of the parameters.

Through the above description of the embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but the former is a better implementation mode in many cases. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (e.g., a mobile phone, a computer, a server, or a network device) to execute the method according to the embodiments of the present invention.

According to another aspect of the embodiments of the present application, there is provided a heterogeneous early warning apparatus for a spacecraft, which is used for implementing the above method for early warning an anomaly of a spacecraft. Optionally, the apparatus is used to implement the above embodiments and preferred embodiments, and details are not repeated for what has been described. As used below, the term "module" may be a combination of software and/or hardware that implements a predetermined function. Although the means described in the embodiments below are preferably implemented in software, an implementation in hardware, or a combination of software and hardware is also possible and contemplated.

Fig. 6 is a block diagram of an alternative anomaly early warning device for a spacecraft according to an embodiment of the present application, where as shown in fig. 6, the device includes:

(1) a first obtaining unit 602, configured to obtain first monitoring data of a spacecraft, where the first monitoring data is monitoring data of the spacecraft in a first time period;

(2) a first input unit 604, connected to the first obtaining unit 602, configured to input first monitoring data into a target LSTM prediction model to obtain a first prediction result output by the target LSTM prediction model, where the target LSTM prediction model is obtained by training an initial LSTM prediction model using second monitoring data, the second monitoring data is monitoring data of a spacecraft in a normal operation state in a second time period, and the first prediction result is used to indicate a first prediction parameter value of a target operation parameter of the spacecraft in a third time period after an end time of the first time period;

(3) and the warning unit 606 is connected to the first input unit 604, and configured to perform an abnormal warning on the spacecraft according to the first predicted parameter value and the first actual parameter value, where the first actual parameter value is an actual value of a target operating parameter of the spacecraft in a third time period.

Alternatively, the first obtaining unit 602 may be used in step S202 in the foregoing embodiment, the first input unit 604 may be used in step S204 in the foregoing embodiment, and the warning unit 606 may be used to execute step S206 in the foregoing embodiment.

The method comprises the steps that first monitoring data of the spacecraft are obtained through the modules, wherein the first monitoring data are monitoring data of the spacecraft in a first time period; inputting the first monitoring data into a target LSTM prediction model to obtain a first prediction result output by the target LSTM prediction model, wherein the target LSTM prediction model is obtained by training an initial LSTM prediction model by using second monitoring data, the second monitoring data are monitoring data of a spacecraft in a normal operation state in a second time period, and the first prediction result is used for indicating a first prediction parameter value of a target operation parameter of the spacecraft in a third time period after the end time of the first time period; and performing abnormity warning on the spacecraft according to the first prediction parameter value and the first actual parameter value, wherein the first actual parameter value is an actual value of a target operation parameter of the spacecraft in a third time period, so that the problem that abnormity cannot be accurately predicted due to lack of data with fault characteristics in an abnormity early warning mode of the spacecraft in the related technology is solved, the convenience of data acquisition is improved, and the accuracy of abnormity early warning of the spacecraft is improved.

As an alternative embodiment, the alarm unit 606 includes:

a first determining module for determining a mean square error between the first predicted parameter value and the first actual parameter value;

the second determination module is used for determining that the spacecraft operates normally under the condition that the difference value between the mean square error and the target error is within the target difference value range;

the third determining module is used for determining that the spacecraft operates abnormally under the condition that the difference value between the mean square error and the target error exceeds the target difference value range; and the sending module is used for sending an alarm message to the mobile terminal of the target object, wherein the alarm message is used for indicating that the spacecraft is abnormal.

As an alternative embodiment, the apparatus further comprises:

the second input unit is used for inputting third monitoring data into the target LSTM prediction model before abnormal warning is conducted on the spacecraft according to the first prediction parameter value and the first actual parameter value to obtain a second prediction result output by the target LSTM prediction model, wherein the third monitoring data are monitoring data of the spacecraft in a normal operation state in a fourth time period, and the second prediction result is used for indicating a second prediction parameter value of a target operation parameter of the spacecraft in a fifth time period after the end time of the third time period;

and the determining unit is used for determining the average mean square error between the second prediction parameter value and the second actual parameter value as the target error, wherein the second actual parameter value is the actual value of the target operating parameter of the spacecraft in the fifth time period.

As an alternative embodiment, the apparatus further comprises:

the second acquisition unit is used for acquiring the monitoring data of the spacecraft in a normal operation state in a second time period before the first monitoring data are input into the target LSTM prediction model to obtain second monitoring data;

a construction unit, configured to construct a training sample of the initial LSTM prediction model using the second monitoring data, where the training sample includes: target sub-monitoring data in a target sub-time period of the second time period, and a third actual parameter value of a target operating parameter of the spacecraft in a sixth time period after the end time of the target sub-time period;

and the training unit is used for training the initial LSTM prediction model by using the training samples to obtain a target LSTM prediction model.

As an alternative embodiment, the training unit comprises:

the input module is used for inputting the target sub-monitoring data into the initial LSTM prediction model to obtain a third prediction result output by the initial LSTM prediction model, wherein the third prediction result is used for indicating a third prediction parameter value of the target operation parameter of the spacecraft in a sixth time period;

and the adjusting module is used for adjusting the model parameters of the initial LSTM prediction model according to the difference value between the third prediction parameter value and the third actual parameter value to obtain a target LSTM prediction model, wherein the difference value between the fourth prediction parameter value and the third actual parameter value of the target operation parameter, which is output by the target LSTM prediction model, is smaller than or equal to the target difference value.

It should be noted that, the above modules may be implemented by software or hardware, and for the latter, the following may be implemented, but not limited to: the modules are all positioned in the same processor; alternatively, the modules are respectively located in different processors in any combination.

According to yet another aspect of embodiments herein, there is provided a computer-readable storage medium. Optionally, the storage medium has a computer program stored therein, where the computer program is configured to execute the steps in any one of the methods provided in the embodiments of the present application when the computer program is executed.

Alternatively, in the present embodiment, the storage medium may be configured to store a computer program for executing the steps of:

acquiring first monitoring data of the spacecraft, wherein the first monitoring data are monitoring data of the spacecraft in a first time period;

inputting the first monitoring data into a target LSTM prediction model to obtain a first prediction result output by the target LSTM prediction model, wherein the target LSTM prediction model is obtained by training an initial LSTM prediction model by using second monitoring data, the second monitoring data are monitoring data of a spacecraft in a normal operation state in a second time period, and the first prediction result is used for indicating a first prediction parameter value of a target operation parameter of the spacecraft in a third time period after the end time of the first time period;

and performing abnormal alarm on the spacecraft according to the first prediction parameter value and the first actual parameter value, wherein the first actual parameter value is an actual value of a target operation parameter of the spacecraft in a third time period.

Optionally, in this embodiment, the storage medium may include, but is not limited to: various media capable of storing computer programs, such as a usb disk, a ROM (Read-Only Memory), a RAM (Random Access Memory), a removable hard disk, a magnetic disk, or an optical disk.

According to another aspect of the embodiments of the present application, there is also provided an electronic device for implementing the above method for warning an anomaly of a spacecraft, where the electronic device may be a server, a terminal, or a combination thereof.

Fig. 7 is a block diagram of an alternative electronic device according to an embodiment of the present application, and as shown in fig. 7, the electronic device includes a memory 702 and a processor 704, the memory 702 stores a computer program, and the processor 704 is configured to execute the steps in any one of the method embodiments described above through the computer program.

Optionally, in this embodiment, the electronic apparatus may be located in at least one network device of a plurality of network devices of a computer network.

Optionally, in this embodiment, the processor may be configured to execute the following steps by a computer program:

The memory 702 may be configured to store software programs and modules, such as program instructions/modules corresponding to the method and apparatus for early warning an anomaly of a spacecraft in the embodiment of the present invention, and the processor 704 executes various functional applications and data processing by running the software programs and modules stored in the memory 702, so as to implement the method for early warning an anomaly of a spacecraft. The memory 702 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 702 can further include memory located remotely from the processor 704, which can be connected to the terminal over a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

As an example, as shown in fig. 7, the memory 702 may include, but is not limited to, the first obtaining unit 602, the first input unit 604, and the warning unit 606 in the abnormality warning apparatus of the spacecraft. In addition, the abnormal early warning device of the spacecraft may further include, but is not limited to, other module units in the abnormal early warning device of the spacecraft, which is not described in detail in this example.

Optionally, the transmitting device 706 is used for receiving or sending data via a network. Examples of the network may include a wired network and a wireless network. In one example, the transmission device 706 includes a Network adapter (NIC) that can be connected to a router via a Network cable and other Network devices to communicate with the internet or a local area Network. In one example, the transmission device 706 is a Radio Frequency (RF) module, which is used to communicate with the internet in a wireless manner.

In addition, the electronic device further includes: a connection bus 708 for connecting the respective module components in the electronic apparatus.

Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and this embodiment is not described herein again.

It can be understood by those skilled in the art that the structure shown in fig. 7 is only an illustration, and the device implementing the above-mentioned anomaly early-warning method for the spacecraft may be a terminal device, and the terminal device may be a terminal device such as a smart phone (e.g., an Android phone, an iOS phone, etc.), a tablet computer, a palm computer, a Mobile Internet Device (MID), a PAD, and the like. Fig. 7 is a diagram illustrating a structure of the electronic device. For example, the terminal device may also include more or fewer components (e.g., network interfaces, display devices, etc.) than shown in FIG. 7, or have a different configuration than shown in FIG. 7.

Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by a program instructing hardware associated with the terminal device, where the program may be stored in a computer-readable storage medium, and the storage medium may include: flash disks, Read-Only memories (ROMs), Random Access Memories (RAMs), magnetic or optical disks, and the like.

Optionally, for an optional example in this embodiment, reference may be made to the examples described in the above embodiment and optional implementation, and this embodiment is not described herein again.

The integrated unit in the above embodiments, if implemented in the form of a software functional unit and sold or used as a separate product, may be stored in the above computer-readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or a part of or all or part of the technical solution contributing to the prior art may be embodied in the form of a software product stored in a storage medium, and including instructions for causing one or more computer devices (which may be personal computers, servers, network devices, or the like) to execute all or part of the steps of the method described in the embodiments of the present application.

In the above embodiments of the present application, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In the several embodiments provided in the present application, it should be understood that the disclosed client may be implemented in other manners. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one type of division of logical functions, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution provided in the embodiment.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The foregoing is only a preferred embodiment of the present application and it should be noted that those skilled in the art can make several improvements and modifications without departing from the principle of the present application, and these improvements and modifications should also be considered as the protection scope of the present application.

Claims

1. An anomaly early warning method for a spacecraft is characterized by comprising the following steps:

acquiring first monitoring data of a spacecraft, wherein the first monitoring data are monitoring data of the spacecraft in a first time period;

inputting the first monitoring data into a target long-term memory network (LSTM) prediction model to obtain a first prediction result output by the target LSTM prediction model, wherein the target LSTM prediction model is obtained by training an initial LSTM prediction model by using second monitoring data, the second monitoring data are monitoring data of the spacecraft in a normal operation state in a second time period, and the first prediction result is used for indicating a first prediction parameter value of a target operation parameter of the spacecraft in a third time period after the end time of the first time period;

and performing abnormal alarm on the spacecraft according to the first prediction parameter value and a first actual parameter value, wherein the first actual parameter value is an actual value of the target operation parameter of the spacecraft in the third time period.

2. The method of claim 1, wherein alerting the spacecraft of the anomaly based on the first predicted parameter value and the first actual parameter value comprises:

determining a mean square error between the first predicted parameter value and the first actual parameter value;

determining that the spacecraft operates normally under the condition that the difference value between the mean square error and the target error is within a target difference value range;

determining that the spacecraft is abnormal in operation under the condition that the difference value between the mean square error and the target error exceeds a target difference value range; and sending an alarm message to a mobile terminal of a target object, wherein the alarm message is used for indicating that the spacecraft is abnormal.

3. The method of claim 2, wherein prior to the anomaly warning of the spacecraft based on the first predicted parameter value and the first actual parameter value, the method further comprises:

inputting third monitoring data into the target LSTM prediction model to obtain a second prediction result output by the target LSTM prediction model, wherein the third monitoring data are monitoring data of the spacecraft in a normal operation state in a fourth time period, and the second prediction result is used for indicating a second prediction parameter value of the target operation parameter of the spacecraft in a fifth time period after the end time of the third time period;

determining an average mean square error between the second predicted parameter value and a second actual parameter value as the target error, wherein the second actual parameter value is an actual value of the target operating parameter of the spacecraft over the fifth time period.

4. The method of any of claims 1-3, wherein prior to inputting the first monitored data into the target LSTM predictive model, the method further comprises:

acquiring monitoring data of the spacecraft in a normal operation state in the second time period to obtain second monitoring data;

constructing training samples of the initial LSTM prediction model using the second monitoring data, wherein the training samples include: target sub-monitoring data within a target sub-time period of the second time period, and a third actual parameter value of the target operating parameter of the spacecraft for a sixth time period after an end time of the target sub-time period;

and training the initial LSTM prediction model by using the training sample to obtain the target LSTM prediction model.

5. The method of claim 4, wherein training the initial LSTM prediction model using the training samples to obtain the target LSTM prediction model comprises:

inputting the target sub-monitoring data into the initial LSTM prediction model to obtain a third prediction result output by the initial LSTM prediction model, wherein the third prediction result is used for indicating a third prediction parameter value of the target operation parameter of the spacecraft in the sixth time period;

and adjusting the model parameters of the initial LSTM prediction model according to the difference value between the third prediction parameter value and the third actual parameter value to obtain the target LSTM prediction model, wherein the difference value between the fourth prediction parameter value of the target operation parameter and the third actual parameter value output by the target LSTM prediction model is smaller than or equal to a target difference value.

6. An anomaly early warning device for a spacecraft, comprising:

the system comprises a first acquisition unit, a second acquisition unit and a control unit, wherein the first acquisition unit is used for acquiring first monitoring data of a spacecraft, and the first monitoring data is monitoring data of the spacecraft in a first time period;

a first input unit, configured to input the first monitoring data into a target long-term memory network LSTM prediction model, so as to obtain a first prediction result output by the target LSTM prediction model, where the target LSTM prediction model is obtained by training an initial LSTM prediction model using second monitoring data, the second monitoring data is monitoring data of the spacecraft in a normal operation state in a second time period, and the first prediction result is used to indicate a first prediction parameter value of a target operation parameter of the spacecraft in a third time period after an end time of the first time period;

and the warning unit is used for performing abnormal warning on the spacecraft according to the first prediction parameter value and a first actual parameter value, wherein the first actual parameter value is an actual value of the target operation parameter of the spacecraft in the third time period.

7. The apparatus of claim 6, wherein the alarm unit comprises:

the second determination module is used for determining that the spacecraft operates normally under the condition that the difference value between the mean square error and the target error is within a target difference value range;

the third determining module is used for determining that the spacecraft is abnormal in operation under the condition that the difference value between the mean square error and the target error exceeds a target difference value range; the sending module is used for sending an alarm message to a mobile terminal of a target object, wherein the alarm message is used for indicating that the spacecraft is abnormal.

8. The apparatus of claim 7, further comprising:

a second input unit, configured to, before performing an abnormal warning on the spacecraft according to the first predicted parameter value and the first actual parameter value, input third monitoring data into the target LSTM prediction model to obtain a second prediction result output by the target LSTM prediction model, where the third monitoring data is monitoring data of the spacecraft in a normal operation state in a fourth time period, and the second prediction result is used to indicate a second predicted parameter value of the target operation parameter of the spacecraft in a fifth time period after an end time of the third time period;

a determining unit, configured to determine an average mean square error between the second predicted parameter value and a second actual parameter value as the target error, where the second actual parameter value is an actual value of the target operating parameter of the spacecraft in the fifth time period.

9. The apparatus of any one of claims 6 to 8, further comprising:

the second acquisition unit is used for acquiring the monitoring data of the spacecraft in a normal operation state in the second time period before the first monitoring data are input into the target LSTM prediction model, so as to obtain second monitoring data;

a construction unit configured to construct training samples of the initial LSTM prediction model using the second monitoring data, wherein the training samples include: target sub-monitoring data within a target sub-time period of the second time period, and a third actual parameter value of the target operating parameter of the spacecraft for a sixth time period after an end time of the target sub-time period;

and the training unit is used for training the initial LSTM prediction model by using the training samples to obtain the target LSTM prediction model.

10. The apparatus of claim 9, wherein the training unit comprises:

an input module, configured to input the target sub-monitoring data into the initial LSTM prediction model to obtain a third prediction result output by the initial LSTM prediction model, where the third prediction result is used to indicate a third prediction parameter value of the target operating parameter of the spacecraft in the sixth time period;

and the adjusting module is used for adjusting the model parameters of the initial LSTM prediction model according to the difference value between the third prediction parameter value and the third actual parameter value to obtain the target LSTM prediction model, wherein the difference value between a fourth prediction parameter value of the target operation parameter and the third actual parameter value output by the target LSTM prediction model is smaller than or equal to a target difference value.