CN116680647A - Method for detecting satellite data by constructing LSTM network by using preprocessed data - Google Patents
Method for detecting satellite data by constructing LSTM network by using preprocessed data Download PDFInfo
- Publication number
- CN116680647A CN116680647A CN202310974501.9A CN202310974501A CN116680647A CN 116680647 A CN116680647 A CN 116680647A CN 202310974501 A CN202310974501 A CN 202310974501A CN 116680647 A CN116680647 A CN 116680647A
- Authority
- CN
- China
- Prior art keywords
- lstm network
- data
- satellite
- network
- telemetry data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims description 23
- 238000012549 training Methods 0.000 claims abstract description 41
- 238000001514 detection method Methods 0.000 claims abstract description 35
- 238000012545 processing Methods 0.000 claims abstract description 23
- 238000007781 pre-processing Methods 0.000 claims abstract description 16
- 210000002569 neuron Anatomy 0.000 claims description 23
- 230000009467 reduction Effects 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 10
- 230000009466 transformation Effects 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 7
- 210000004027 cell Anatomy 0.000 claims description 6
- 238000007689 inspection Methods 0.000 claims description 6
- 238000005259 measurement Methods 0.000 claims description 6
- 238000002372 labelling Methods 0.000 claims description 5
- 210000005036 nerve Anatomy 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000003213 activating effect Effects 0.000 claims description 3
- 230000005856 abnormality Effects 0.000 abstract description 3
- 230000036541 health Effects 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 15
- 230000002159 abnormal effect Effects 0.000 description 12
- 230000008569 process Effects 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 6
- 239000004973 liquid crystal related substance Substances 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000015654 memory Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000008713 feedback mechanism Effects 0.000 description 1
- 230000003862 health status Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Radio Relay Systems (AREA)
Abstract
The invention discloses a satellite data detection method by constructing an LSTM network by using preprocessed data, which comprises preprocessing satellite telemetry data; establishing an LSTM network, and training the LSTM network by utilizing the preprocessed data; generalizing the LSTM network with the pre-processed data; inspecting the LSTM network using the preprocessed data; performing satellite telemetry data anomaly detection by using the detected LSTM network; according to the invention, after the satellite telemetry data is preprocessed, the LSTM network is established, the LSTM network is trained, the LSTM network is generalized, and the LSTM network is checked by utilizing the preprocessed data, so that the obtained LSTM network has higher matching property and accuracy, is used for detecting the satellite telemetry data abnormality, can effectively monitor the on-orbit running state of the satellite, solves the problems of high concurrent processing speed requirement, high fault detection false alarm rate, high fault leakage rate and the like in the existing satellite running management, and provides technical support for scientifically predicting the satellite health state.
Description
Technical Field
The invention relates to the technical field of satellite telemetry data anomaly detection, in particular to a satellite data detection method by constructing an LSTM network by using preprocessed data.
Background
The satellite telemetry data received by the measurement and control station can intuitively reflect the working states of the satellite platform and the effective load and the space environment condition. The satellite telemetry data mainly comprises two types of digital quantity and analog quantity, the two types of digital quantity telemetry parameters and serial digital quantity telemetry parameters can be subdivided into two types of digital quantity telemetry parameters, the other types of digital quantity telemetry parameters are mainly physical quantities which continuously change along with time, such as speed, acceleration, angle, angular speed, voltage, current, temperature and the like, and the satellite sensor is used for receiving the analog quantity after acquiring the analog quantity, through analog-digital conversion and modulation demodulation, and then is received by the measurement and control station.
The number of telemetry parameters of one satellite is tens of thousands, the parameter state changes are related to each other, and all telemetry parameters are directly detected and analyzed by a common threshold method, so that the problems of high concurrent processing speed requirement, high fault detection false alarm rate, high fault missing alarm rate and the like exist. In order to accurately evaluate the in-orbit health status of a satellite, predict the operational trend of the satellite, it is necessary to perform anomaly detection for the telemetry data of the satellite.
The recurrent neural network (Recurrent Neural Network, RNN) is a typical neural network for processing time series data in deep learning, and is more suitable for processing data with time series characteristics by adopting a negative feedback mechanism compared with other neural network models due to the increased consideration of time lines. The analog telemetry data generally has the characteristics of large data volume, time change and strong data association, and the RNN is very suitable for the analog telemetry data. However, with continuous iterative computation of the RNN, problems of gradient disappearance and gradient explosion are easily formed, and the process of RNN network training is affected. To improve RNN networks, improvements are made by using a gating cell Memory mechanism, namely Long Short-Term Memory neural networks (LSTM).
The satellite telemetry data anomaly detection analysis result based on the LSTM network can support a user to scientifically predict the health state of the satellite, the user timely discovers the anomaly of a satellite platform and a payload by monitoring satellite telemetry data generated by the satellite in-orbit operation, positions the fault cause, adopts effective means to improve the stability and safety of the satellite, further reduces the satellite damage rate and reduces the occurrence of disastrous accidents.
The patent document with the document number of CN109934337B discloses a spacecraft telemetry data anomaly detection method based on integrated LSTM, which comprises the steps of preprocessing training data, classifying the training data into a training set A and a training set B, and training the training set A and the training set B based on an LSTM model respectively, so that the influence of long-term dependence on detection results can be effectively reduced, and the accuracy of overall anomaly detection is improved; the prediction result of the LSTM model is integrated according to a certain weight to obtain a final predicted value of the telemetry data, the error of the prediction result and the actual value is utilized to make difference, namely the error is smoothed, and an abnormal section of the telemetry data is detected according to a dynamic threshold value. The invention can be applied to the technical field of detection of telemetry data anomalies, but simultaneously has the following advantages: 1. the satellite telemetry data is not effectively preprocessed, so that the establishment precision of the LSTM network is affected; 2. without the process of generalizing and checking the LSTM network, the precision of the LSTM network is not easy to guarantee.
Disclosure of Invention
The invention aims to provide a satellite data detection method by establishing an LSTM network by using preprocessed data, which solves the problems that the lack of preprocessing on satellite telemetry data is low in establishing precision of the LSTM network and the abnormal detection accuracy of the satellite telemetry data is affected.
The aim of the invention can be achieved by the following technical scheme: a satellite data detection method by constructing an LSTM network by using preprocessed data comprises the following steps:
s1, carrying out data preprocessing on satellite telemetry data;
s2, establishing an LSTM network, and training the LSTM network by utilizing the preprocessed data;
s3, generalizing the LSTM network by using the preprocessed data for the trained LSTM network;
s4, generalizing the LSTM network; inspecting the LSTM network using the preprocessed data; if the LSTM network is not up to the inspection requirement after inspection, repeating the steps S2 to S3;
s5, satellite telemetry data anomaly detection is carried out by using the LSTM network after detection.
Further: the step S1 of preprocessing satellite telemetry data comprises the following steps:
s11, performing outlier rejection processing on satellite telemetry data;
s12, performing dimension reduction on satellite telemetry data subjected to outlier rejection processing;
s13, performing scale transformation on the satellite telemetry data subjected to the dimension reduction treatment;
s14, performing feature labeling on the satellite telemetry data after the scale transformation.
Further: the method for performing scale transformation processing on the satellite telemetry data after the dimension reduction processing comprises the following steps:
normalizing the telemetry data to the same dimension, and processing by adopting a maximum and minimum normalization method data normalization method, wherein the formula is as follows:
wherein x is the data value to be normalized,for the telemetry data to be at its maximum value,for the telemetry data to be at a minimum value,is the normalized data value.
Further: the step of establishing the LSTM network by the S2 comprises the following steps:
s21, initializing an LSTM network, and determining the network layer numbers and the neuron numbers in an input layer, a hidden layer and an output layer of the LSTM network;
the number of neurons in the input layer is 2, the number of neurons in the output layer is p (p is more than or equal to 2), the hidden layer is 3 layers, the number q (q is more than or equal to 5) of neurons in each layer is variable, and p and q are positive integers.
S22, setting an objective function of an LSTM network:
wherein, the liquid crystal display device comprises a liquid crystal display device,in order to be able to input the input,in the state of the cell,for the current cell state,to conceal the output of the layer state or instant t,、、、respectively a forgetting gate, an input gate, a current unit state input part calculating module and a weight matrix of an output gate,、、、for the corresponding offset vector to be used,activating functions for sigmoid, i.e.Tanh is a hyperbolic tangent function, i.e；
The objective function adopts a Root Mean Square Error (RMSE) as a standard of measurement, the RMSE is the square root of the ratio of the sum of squares of the LSTM network predicted value and the actual value error to the number of times of prediction m, the RMSE is used for measuring the deviation between the network predicted value and the actual value, and the calculation formula of the RMSE is as follows:
wherein the method comprises the steps ofAndrespectively the actual true value of the satellite telemetry parameter at the time t,And the network predicted value, n, is the number of data.
S23, setting training parameters of an LSTM network;
further: the step of training the LSTM network by the S2 through the preprocessing data comprises the following steps:
s31, forward calculating output values of all nerve units in an input layer, a hidden layer and an output layer in the LSTM network by utilizing the preprocessed data;
s32, reversely calculating the error value of each neuron in the LSTM network according to the output value of each neural unit;
s33, calculating the descent gradient of the LSTM network objective function according to the gradient descent principle, and updating each weight matrix and bias vector of each neuron in the LSTM network according to the descent gradient of the objective function;
s34, judging whether the LSTM network training reaches a preset stopping condition, and if the LSTM network training does not reach the training stopping condition, continuing to step S31 to step S33.
The invention has the beneficial effects that:
1. according to the invention, after the satellite telemetry data is preprocessed, the LSTM network is established, the LSTM network is trained, the LSTM network is generalized, and the LSTM network is checked by utilizing the preprocessed data, so that the obtained LSTM network has higher matching property and accuracy, is used for detecting the satellite telemetry data abnormality, can effectively monitor the on-orbit running state of the satellite, solves the problems of high concurrent processing speed requirement, high fault detection false alarm rate, high fault leakage rate and the like in the existing satellite running management, and provides technical support for predicting the satellite health state.
2. The invention solves the problems of frame loss, random abnormal value occurrence, large parallel processing data quantity and the like of the existing satellite telemetry data through the whole preprocessing process of removing the satellite telemetry data outlier, dimension reduction processing, dimension conversion and characteristic labeling, and the processed satellite telemetry data, has higher precision when the LSTM network is utilized to carry out the abnormal detection of the satellite telemetry data, and can effectively monitor the on-orbit running state of the satellite.
3. According to the invention, the network layer number and the neuron number of the LSTM network are reasonably set according to the characteristic of the abnormal value of the satellite telemetry data, the complexity of the LSTM network is matched with the accuracy of the abnormal detection of the satellite telemetry data, and the running accuracy of the LSTM network is ensured by reasonably setting the objective function of the LSTM network.
Drawings
FIG. 1 is a flow chart of a method for satellite data detection by using pre-processed data to build an LSTM network according to the present invention;
FIG. 2 is a schematic diagram of an LSTM network according to the present invention;
fig. 3 is a schematic diagram of the internal unit configuration of the LSTM network of the present invention.
10. An input layer; 20. a hidden layer; 30. an output layer; 40. neurons.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar symbols indicate like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are exemplary only for explaining the present invention and are not to be construed as limiting the present invention.
As shown in fig. 1-3, the invention discloses a satellite data detection method by constructing an LSTM network by using preprocessed data, which comprises the following steps:
s1, carrying out data preprocessing on satellite telemetry data;
s2, establishing an LSTM network, and training the LSTM network by utilizing the preprocessed data;
s3, generalizing the LSTM network by using the preprocessed data for the trained LSTM network;
s4, generalizing the LSTM network; inspecting the LSTM network using the preprocessed data; if the LSTM network is not up to the inspection requirement after inspection, repeating the steps S2 to S3;
s5, satellite telemetry data anomaly detection is carried out by using the LSTM network after detection.
The preprocessing step for satellite telemetry data comprises the following steps:
s11, performing outlier rejection processing on satellite telemetry data, wherein during the receiving and processing of satellite downlink telemetry data, the satellite self-sensor, the space electromagnetic environment, a measurement and control station data processing server and the like are interfered to form abnormal inflection point data, and if the data are not processed, the abnormal inflection point data are called as outliers, and the subsequent data anomaly detection analysis and the training precision of an LSTM network model are uncontrollably influenced; the difference between the outlier and the outlier is needed to be distinguished, and the outlier usually occurs in a period of time but does not exceed the normal range too much; the wild value appears randomly, the value is very large or very small, and is obviously different from the normal value; calculating the median, maximum value and minimum value of a certain telemetry parameter in a period of time by adopting a statistical method, drawing a box diagram, manually judging the telemetry parameter wild value, and directly removing; and filling the data loss caused by field elimination in the satellite telemetry data by adopting numerical interpolation methods such as cubic spline interpolation, lagrange interpolation and the like.
S12, performing dimension reduction on satellite telemetry data subjected to outlier rejection processing; according to different transmission periods of satellite telemetry data, the satellite telemetry data can be subdivided into quick change telemetry and slow change telemetry, wherein the quick change telemetry polling period is generally 1 second, and the slow change telemetry polling period is generally 8 seconds; in order to solve the problem that the accuracy of the training result is affected due to the fact that training data sets of the LSTM network model are not aligned due to different sampling rates, and in order to accelerate the training process of the LSTM network model, data dimension reduction is needed to be carried out on original satellite telemetry data through setting the sampling rates, and a new data set is formed.
S13, carrying out normalization transformation processing on the satellite telemetry data scale after the dimension reduction processing; the different and magnitude of various telemetry parameters in the satellite telemetry data also have obvious difference, in order to eliminate the influence of the telemetry parameter dimension on LSTM network training, solve the comparability problem between telemetry parameter data, normalize the data of various telemetry parameters to the same dimension, and preferably select the maximum and minimum normalization method for the dimension transformation of the satellite telemetry data to process, namelyThe method comprises the steps of carrying out a first treatment on the surface of the Wherein x is the data value to be normalized,for the telemetry data to be at its maximum value,for the telemetry data to be at a minimum value,is the normalized data value. After satellite telemetry data is normalized, in the process of training an LSTM network, the optimal solution solving process can be quickened, and the convergence rate is rapidly increased.
S14, carrying out feature labeling on satellite telemetry data after scale normalization transformation; the data of each satellite telemetry parameter is required to be added with a label by adding a data label in the form of a satellite telemetry data labeling feature, wherein the label is a telemetry parameter abnormality degree value (0 represents normal telemetry, 1 represents abnormal telemetry, and values between 0 and 1 represent abnormal telemetry of different degrees), so that a telemetry data set X= { X1, X2,..x n }, wherein X is a two-dimensional matrix, and xi represents a vector.
The step of establishing the long-short-term memory neural network LSTM comprises the following steps:
s21, initializing an LSTM network, and determining the network layer numbers and the number of 40 neurons in an input layer 10, a hidden layer 20 and an output layer 30 of the LSTM network; as shown in fig. 2, for the characteristics of satellite telemetry data anomaly detection, the number of network layers and the number of neurons 40 in an input layer 10, a hidden layer 20 and an output layer 30 of an LSTM network are defined; the number of the neurons 40 in the input layer 10 is 2, the number of the neurons 40 in the output layer 30 is p (p is more than or equal to 2), the hidden layer 20 is 3, the number q (q is more than or equal to 5) of the neurons 40 in each layer is variable, p and q are positive integers, and the reasonable network layer number and the number of the neurons 40 can ensure the precision of the LSTM network without excessive iteration times, so that the complex efficiency of the LSTM network is reduced.
S22, setting an objective function of an LSTM network: the internal units of the LSTM network form as shown in fig. 3, and the LSTM network is characterized in that the forgetting and updating of the memory are controlled by three gating units of a forgetting gate, an input gate and an output gate, and the objective function of the LSTM network can be set as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,in order to be able to input the input,in the state of the cell,for the current cell state,to conceal the output of the layer 20 state or instant t,、、、the weight matrix of the forgetting gate, the input gate, the current unit state input part calculating module and the output gate are respectively,、、、for the corresponding offset vector to be used,activating functions for sigmoid, i.e.Tanh is a hyperbolic tangent function, i.e；
The objective function adopts a Root Mean Square Error (RMSE) as a standard of measurement, the RMSE is the square root of the ratio of the sum of squares of the LSTM network predicted value and the actual value error to the number of times of prediction m, the RMSE is used for measuring the deviation between the network predicted value and the actual value, and the calculation formula of the RMSE is as follows:
wherein the method comprises the steps ofAndthe actual real value and the network predicted value of the satellite telemetry parameter at the time t are respectively, and n is the number of data.
S23, setting training parameters of an LSTM network; such as maximum number of iterations of training, learning rate penalty, output layer 30 neuron 40 encoding, data size to engage in training, etc.
The training of the long-term memory neural network LSTM network by the preprocessed data comprises the following steps:
s31, forward calculating the output value of each nerve unit in the input layer 10, the hidden layer 20 and the output layer 30 in the LSTM network by utilizing the preprocessed data;
s32, according to the LSTM network error back propagation principle, the output values of the nerve units are utilized to reversely calculate the error values of the nerve cells 40 in the LSTM network.
S33, calculating the LSTM network objective function descending gradient according to the LSTM network gradient descending principle, and updating each weight matrix and bias vector of each neuron 40 in the LSTM network according to the objective function descending gradient.
S34, judging whether the LSTM network training reaches a preset stopping condition, wherein the stopping condition comprises preset training precision and training iteration step length, judging whether the LSTM network training reaches the stopping condition, stopping training the LSTM network if the training stopping condition is met, and otherwise, continuing training by using the preprocessing data.
And generalizing the trained LSTM network, selecting a preprocessed data set which does not participate in the LSTM network training as an input data generalization LSTM network according to a data preprocessing result, counting the difference value between the output data and the expected output data of the generalization LSTM network, and after carrying out inverse normalization on the output of the network, counting the difference value between the output and the expected output data of the generalization LSTM network, and analyzing and evaluating the generalization capability of the network.
And continuously checking the generalized LSTM network to ensure the precision of the LSTM network, and selecting the LSTM network with strong generalization capability and better indexes as a checked network. According to the real satellite telemetry data, performing data preprocessing such as outlier rejection, data dimension reduction, scale transformation and the like on the satellite telemetry data, then using the data preprocessing as input to test an LSTM network, performing inverse normalization on the output of the network, counting the difference value between the output of the test LSTM network and the actual output data, analyzing and evaluating the test capability of the network, and performing alarm display on an abnormal result to generate an abnormal detection report. If the network has weaker checking capability, the LSTM network can not meet the checking requirement after being checked, namely, the fault detection false alarm rate based on satellite telemetry data is high, the fault missing report rate is high, the training sample size is enlarged, and the LSTM network establishment, LSTM network generalization and LSTM network checking processes are repeatedly carried out.
The foregoing is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art, who is within the scope of the present invention, should make equivalent substitutions or modifications according to the technical scheme of the present invention and the inventive concept thereof, and should be covered by the scope of the present invention.
It is to be understood that the terms "center," "longitudinal," "transverse," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counter-clockwise," "axial," "radial," "circumferential," and the like are directional or positional relationships as indicated based on the drawings, merely to facilitate describing the invention and to simplify the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be configured and operated in a particular orientation, and therefore should not be construed as limiting the invention.
Furthermore, the terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include at least one such feature. In the description of the present invention, the meaning of "plurality" means at least two, for example, two, three, etc., unless specifically defined otherwise.
In the present invention, unless explicitly specified and limited otherwise, the terms "mounted," "connected," "secured," and the like are to be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally formed; may be mechanically connected, may be electrically connected or may be in communication with each other; either directly or indirectly, through intermediaries, or both, may be in communication with each other or in interaction with each other, unless expressly defined otherwise. The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art according to the specific circumstances.
In the present invention, unless expressly stated or limited otherwise, a first feature "up" or "down" a second feature may be the first and second features in direct contact, or the first and second features in indirect contact via an intervening medium. Moreover, a first feature being "above," "over" and "on" a second feature may be a first feature being directly above or obliquely above the second feature, or simply indicating that the first feature is level higher than the second feature. The first feature being "under", "below" and "beneath" the second feature may be the first feature being directly under or obliquely below the second feature, or simply indicating that the first feature is less level than the second feature.
Claims (5)
1. A satellite data detection method by using preprocessed data to build an LSTM network is characterized in that: the method comprises the following steps:
s1, carrying out data preprocessing on satellite telemetry data;
s2, establishing an LSTM network, and training the LSTM network by utilizing the preprocessed data;
s3, generalizing the LSTM network by using the preprocessed data for the trained LSTM network;
s4, generalizing the LSTM network; inspecting the LSTM network using the preprocessed data; if the LSTM network is not up to the inspection requirement after inspection, repeating the steps S2 to S3;
s5, satellite telemetry data anomaly detection is carried out by using the LSTM network after detection.
2. The method for satellite data detection by using the preprocessed data to build the LSTM network according to claim 1, wherein: the step S1 of preprocessing satellite telemetry data comprises the following steps:
s11, performing outlier rejection processing on satellite telemetry data;
s12, performing dimension reduction on satellite telemetry data subjected to outlier rejection processing;
s13, performing scale transformation on the satellite telemetry data subjected to the dimension reduction treatment;
s14, performing feature labeling on the satellite telemetry data after the scale transformation.
3. The method for satellite data detection by using the preprocessed data to build the LSTM network according to claim 2, wherein: the method for performing scale transformation processing on the satellite telemetry data after the dimension reduction processing comprises the following steps:
normalizing the telemetry data to the same dimension, and processing by adopting a data normalization method such as a maximum normalization method and a minimum normalization method, wherein the formula is as follows:wherein x is a data value to be normalized, < ->For telemetry data maximum, +.>For telemetry data min->Is the normalized data value.
4. The method for satellite data detection by using the preprocessed data to build the LSTM network according to claim 1, wherein: the step of establishing the LSTM network by the S2 comprises the following steps:
s21, initializing an LSTM network, and determining the network layer numbers and the neuron (40) in an input layer (10), a hidden layer (20) and an output layer (30) of the LSTM network;
the number of the neurons (40) in the input layer (10) is 2, the number of the neurons (40) in the output layer (30) is p (p is more than or equal to 2), the hidden layer (20) is 3 layers, the number q (q is more than or equal to 5) of the neurons (40) in each layer is variable, and p and q are positive integers;
s22, setting an objective function of an LSTM network:wherein (1)>For input, & lt + & gt>For the state of the unit->For the current cell state->For hiding the state of layer (20) or the output at time t,/or>、/>、/>、/>Weight matrix of forgetting gate, input gate, current unit state input part calculation module, output gate, respectively, +.>、/>、/>、/>For the corresponding bias vector, +.>Activating a function for sigmoid, i.e.)>Tan h is a hyperbolic tangent function, i.e.>；
The objective function adopts a Root Mean Square Error (RMSE) as a standard of measurement, the RMSE is the square root of the ratio of the sum of squares of the LSTM network predicted value and the actual value error to the number of times of prediction m, the RMSE is used for measuring the deviation between the network predicted value and the actual value, and the calculation formula of the RMSE is as follows:wherein->And->The actual real value and the network predicted value of the satellite telemetry parameter at the time t are respectively, and n is the number of data;
s23, setting training parameters of the LSTM network.
5. The method for satellite data detection by using the preprocessed data to build the LSTM network according to claim 1, wherein: the step of training the LSTM network by the S2 through the preprocessing data comprises the following steps:
s31, forward calculating output values of all nerve units in an input layer (10), a hidden layer (20) and an output layer (30) in the LSTM network by utilizing the preprocessed data;
s32, reversely calculating error values of all neurons (40) in the LSTM network according to the output values of all the nerve units;
s33, calculating the descent gradient of the LSTM network objective function according to the gradient descent principle, and updating each weight matrix and bias vector of each neuron (40) in the LSTM network according to the descent gradient of the objective function;
s34, judging whether the LSTM network training reaches a preset stopping condition, and if the LSTM network training does not reach the training stopping condition, continuing to step S31 to step S33.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310974501.9A CN116680647A (en) | 2023-08-04 | 2023-08-04 | Method for detecting satellite data by constructing LSTM network by using preprocessed data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310974501.9A CN116680647A (en) | 2023-08-04 | 2023-08-04 | Method for detecting satellite data by constructing LSTM network by using preprocessed data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116680647A true CN116680647A (en) | 2023-09-01 |
Family
ID=87781389
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310974501.9A Pending CN116680647A (en) | 2023-08-04 | 2023-08-04 | Method for detecting satellite data by constructing LSTM network by using preprocessed data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116680647A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109859469A (en) * | 2019-02-15 | 2019-06-07 | 重庆邮电大学 | A kind of vehicle flowrate prediction technique based on integrated LSTM neural network |
| CN109934337A (en) * | 2019-03-14 | 2019-06-25 | 哈尔滨工业大学 | A kind of detection method of the spacecraft telemetry exception based on integrated LSTM |
| CN113255792A (en) * | 2021-06-01 | 2021-08-13 | 广东粤港澳大湾区硬科技创新研究院 | Data anomaly point detection method, device, system and storage medium |
| CN116205318A (en) * | 2021-12-21 | 2023-06-02 | 一拓通信集团股份有限公司 | LSTM-based time sequence landslide data prediction method |
| CN116451848A (en) * | 2023-03-29 | 2023-07-18 | 北京信息科技大学 | Satellite telemetry data prediction method and device based on space-time attention mechanism |
-
2023
- 2023-08-04 CN CN202310974501.9A patent/CN116680647A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109859469A (en) * | 2019-02-15 | 2019-06-07 | 重庆邮电大学 | A kind of vehicle flowrate prediction technique based on integrated LSTM neural network |
| CN109934337A (en) * | 2019-03-14 | 2019-06-25 | 哈尔滨工业大学 | A kind of detection method of the spacecraft telemetry exception based on integrated LSTM |
| CN113255792A (en) * | 2021-06-01 | 2021-08-13 | 广东粤港澳大湾区硬科技创新研究院 | Data anomaly point detection method, device, system and storage medium |
| CN116205318A (en) * | 2021-12-21 | 2023-06-02 | 一拓通信集团股份有限公司 | LSTM-based time sequence landslide data prediction method |
| CN116451848A (en) * | 2023-03-29 | 2023-07-18 | 北京信息科技大学 | Satellite telemetry data prediction method and device based on space-time attention mechanism |
Non-Patent Citations (1)
| Title |
|---|
| 山梦娜: ""基于深度学习的遥测数据异常检测"", 《中国优秀硕士学位论文全文数据库》, pages 15 - 20 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110751199A (en) | Satellite anomaly detection method based on Bayesian neural network | |
| CN108171259A (en) | A kind of wisdom tip water quality monitoring system and method based on NB-IoT | |
| CN105676833B (en) | Power generation process control system fault detection method | |
| CN112785091A (en) | Method for performing fault prediction and health management on oil field electric submersible pump | |
| KR101896015B1 (en) | Ai type remote meter reading system | |
| CN113155178B (en) | Sensor system and method | |
| CN111461187B (en) | Intelligent building settlement detection system | |
| CN111030889B (en) | Network traffic prediction method based on GRU model | |
| CN111611294B (en) | Star sensor data anomaly detection method | |
| CN107292040A (en) | Finite Element Model Validation based on Kriging response surfaces | |
| Li et al. | Collective efficacy of support vector regression with smoothness priority in marine sensor data prediction | |
| CN112100574A (en) | Resampling-based AAKR model uncertainty calculation method and system | |
| Yong et al. | Bayesian autoencoders for drift detection in industrial environments | |
| Li et al. | Reliability modeling of a series system with correlated or dependent component degradation processes | |
| CN111122811A (en) | Sewage treatment process fault monitoring method of OICA and RNN fusion model | |
| CN111832703A (en) | Sampling interval perception long-short term memory network-based process manufacturing industry irregular sampling dynamic sequence modeling method | |
| CN116680647A (en) | Method for detecting satellite data by constructing LSTM network by using preprocessed data | |
| CN117391463A (en) | River pollution tracing method | |
| CN114626482A (en) | Active digital array radar real-time fault diagnosis method based on deep learning | |
| CN115047853A (en) | Micro fault detection method based on recursion standard variable residual error and kernel principal component analysis | |
| CN107402559B (en) | A kind of batch process measurement data method for detecting abnormality based on the variation of dynamic Hyper-sphere Structure | |
| CN113901741A (en) | Ocean-ionosphere dissipation dynamics simulation method, simulation system, equipment and application | |
| US20210232132A1 (en) | Online fault localization in industrial processes without utilizing a dynamic system model | |
| Saranathan et al. | Multi-Parameter Retrieval of Water Quality Indicators from Bayesian and Mixture Density Networks | |
| CN117786586B (en) | Self-locking nut detection system and method thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |