CN109858693A - A kind of prediction technique for declaring situation towards satellite network data - Google Patents
A kind of prediction technique for declaring situation towards satellite network data Download PDFInfo
- Publication number
- CN109858693A CN109858693A CN201910072478.8A CN201910072478A CN109858693A CN 109858693 A CN109858693 A CN 109858693A CN 201910072478 A CN201910072478 A CN 201910072478A CN 109858693 A CN109858693 A CN 109858693A
- Authority
- CN
- China
- Prior art keywords
- data
- satellite
- sample
- declaring
- network model
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of prediction techniques that situation is declared towards satellite network data, which comprises counts satellite declaration material over the years as sample data；Sample data is pre-processed；The RBF neural network model that Tendency Prediction is declared for satellite network data is established according to pretreated sample data；It is declared and is predicted using satellite network data of the RBF neural network model to the specified time.Method of the invention is when declaring situation to satellite network data and predicting, without refining the influence factor for declaring situation, the historical sample data for relying only on ITU database is fitted, and can be obtained preferable prediction result.
Description
Technical field
The present invention relates to satellite network frequencies and rail position protection of resources field, and in particular to one kind is towards satellite network data
Declare the prediction technique of situation.
Background technique
Present satellites network data declares process and relies on authorities or the artificial judgment of domain expert, artificial judgment substantially
Basis be that simple mathematical statistics is carried out to historical satellite network declaration material using the database of ITU, not further to defending
The formulation of star network data declared trend, declare strategy carries out the analysis of quantification, often on the basis of statistical data
Carry out artificial perception judgement.
Summary of the invention
It is an object of the present invention to which the present invention passes through algorithm neural network based, using the thinking of models fitting, to go through
History data quantitativeization prediction Future Satellite network data declares trend.
To achieve the goals above, a kind of prediction technique that situation is declared towards satellite network data of the present invention, the side
Method includes:
Satellite declaration material over the years is counted as sample data；
Sample data is pre-processed；
The RBF neural mould that Tendency Prediction is declared for satellite network data is established according to pretreated sample data
Type；
It is declared and is predicted using satellite network data of the RBF neural network model to the specified time.
As a kind of improvement of the above method, the step of satellite network data declares situation, is specifically included:
Step S1) sample data is serialized, the sample data serialized；
The time is declared according to satellite declaration material, is counted by month, obtains with " part/moon " being the complete of statistical unit
Ball satellite declares N data bank table；" moon in year-" field is serialized, the form of " moon in year-" is replaced with sequence of natural numbers, is obtained
To the sample data of serializing；
Step S2) in fixed time period data cleansing is carried out to the sample data of serializing；
Step S3) to step S2) treated sample data carry out queue resist it is bad.
As a kind of improvement of the above method, described established according to pretreated data sample is used for satellite network data
The step of declaring the RBF neural network model of Tendency Prediction specifically includes；
Step T1) initialize the RBF neural network model that Tendency Prediction is declared for satellite network data；
The RBF neural network model uses excitation function of the Gaussian bases as neuron；Its node is single input
With single output；
Step T2) in all sample datas of RBF neural network model input terminal input, then calculate all nodes of output end
Output as a result, choose output result and sample results label deviation it is maximum be worth as in the Gaussian bases of neuron
The heart calculates variance of the mean square error MSE of output result and sample results label as neuron Gaussian bases；
Step T3) neuron of step T2) is added in the hidden layer of RBF neural network model；Construct new RBF
Neural network model is transferred to step 2), until mean square error MSE is less than threshold value or neuronal quantity reaches the upper limit, the RBF
Neural network model building finishes.
As a kind of improvement of the above method, the satellite network data declare include: total satellite data declare number,
The satellite data of some country declares number, the satellite data of a certain frequency range declares number and the satellite data of a certain rail position is declared
Number.
Present invention has an advantage that
1, method of the invention uses RBF neural network algorithm, by historical sample data, realizes to satellite network data
Declare the blind recognition of Tendency Prediction model；
2, method of the invention does not need to determine satellite network data and declares the influence factor of situation (some factors even can not
Quantization, such as international space flight development situation), it is only necessary to historical sample data is fitted, it is accurate to can be obtained higher prediction
Rate；
3, method of the invention is when declaring situation to satellite network data and predicting, without refining the shadow for declaring situation
The factor of sound, the historical sample data for relying only on ITU database are fitted, and can be obtained preferable prediction result.
Detailed description of the invention
Fig. 1 it is of the invention declare situation trend prediction flow chart；
Fig. 2 is whole world GSO satellite declaration material database table structure figure；
Fig. 3 is serialized data library schematic diagram；
Fig. 4 is that 1986-2018 global satellite network N data declares statistical chart；
Fig. 5 is that 2008-2018 global satellite network N data declares statistical chart；
Fig. 6 is stubborn and stupid treated the sample schematic diagram of queue；
Fig. 7 is RBF neural network structure figure；
Fig. 8 is typical radial basic function curve graph；(1 is Gaussian function, and 2 be Reflected sigmoidal letter
Number；3 be inverse Multiquadric function)；
Fig. 9 is neural metwork training curve；
Figure 10 is matched curve of the RBF neural to original sample data after training.
Specific embodiment
Method of the invention is described in detail with reference to the accompanying drawing.
The situation of declaring of satellite network data refers under different observation dimensions (whole world, country, operator etc.), data quantity
The state and trend to change with time.For satellite network data the statistics for declaring state, trend development is declared to it
Forecast assessment can instruct relevant departments or satellite operations person to judge following aspect content, and mainly assessment collection has: satellite network
The growth trend of data quantity, the Competitive Trend of GSO track resources and growth trend of NGSO data etc..The satellite network money
It includes: that total satellite data declares number, the satellite data of some country declares the satellite data of number, a certain frequency range that material, which is declared,
The satellite data for declaring number and a certain rail position declares number.
Declaring situation can be predicted using neural network model, based on RBF neural, specifically declare trend
Prediction process is as shown in Figure 1.
Method of the invention is carried out by taking the trend prediction declared 01 month 2017 whole world GSO satellite network N data as an example
Method validation predicts that such data declared number at 2017 01 month, specific to predict that procedure declaration is as follows:
Step 1) sample sequence
According to the rule searching of central integral platform database, annual N data is extracted from database, according to data
The time is declared, is counted by month, obtains declaring N data bank table with the global satellite that " part/moon " is statistical unit, specifically such as
Library table is as shown in Figure 2.
" moon in year-" field is serialized, the form of " moon in year-" is replaced with sequence of natural numbers, because neural network
Input node generally requires as continuous or discrete numeric form, and the library table after serializing is as shown in Figure 3.
The library table data after serializing are imported in a program, with " moon in year-" after sequence for abscissa, with GSO satellite network
It is ordinate that network N data, which monthly declares number (part/moon), can obtain declaration material statistic curve (1986~2018 years) such as Fig. 4 institute
Show.
Step 2) data cleansing
Since time span is too long, influence factor is excessive, considers to carry out data cleansing with 10 years for one group of data, right
01 month 2017 GSO satellite network N declares quantity and is predicted, i.e., was starting point with 2008 06, counts in December, 2016 and is
Terminal, GSO satellite network N data declare situation, it is as shown in Figure 5 can to obtain declaration material statistic curve.
Step 3) queue resists bad
Queue is carried out to the data after cleaning and resists bad processing, and the data for intercepting 06 month in December, 2016 in 2008 are made
For the training sample of RBF network model, sample situation is as shown in Figure 6.
Step 4) netinit
After the acquisition for completing data sample, the design of RBF neural network structure can be started, since sample point is more,
The building of RBF network structure uses RAN (resource allocation network) learning algorithm, i.e., is no RBF neuron at the beginning, only
There are single input, single output.
RBF neural is a kind of feedforward neural network, generally 3-tier architecture, as shown in Figure 7.
In Fig. 7, input layer is made of P signal source node.If N is currently trained total sample number, for training set
Each sample be input vector: X=(X1,X2,……Xp), wherein Xi(i=1 ... ..., P) is i-th of input of network,
Shared M hidden node forms hidden layer.Wherein the activation primitive of each hidden layer node is a radial basis function, and radial base is
A kind of non-negative nonlinear function of central point radial symmetric of the local distribution of decaying.The RBF neural network model that this method uses
Structure, be single input and single output；Radial basis function Φ can take diversified forms, and curve shape is as shown in figure 8, wherein 1
It represents
Since Gaussian bases have, representation is simple, radial symmetric, slickness are good, it is excellent to be easy to carry out theory analysis etc.
Point, so general RBF neural uses Gaussian bases, expression-form is as follows:
The output of RBF neural is the linear combination of hidden layer node output, and output form may be expressed as:
The initialization procedure of RBF neural, comprising:
RBF neural uses excitation function of the Gaussian bases as neuron, sees formula (1)；
Each layer network number of nodes is 1-0-1；
It is 1/time that hidden node, which increases training speed Speed,；(value can also determine any integer value greater than 1, depending on tool
Depending on body forecasting problem)
Output end mean square error MSE target value is set to 0；(value can also be set to the arbitrary value close to 0, depending on specific prediction
Depending on problem)
The adjustment of step 5) center
First the sample of worst error is set about from input data, is increased a RBF neuron, is exported accordingly, so
Redesign network linear layer afterwards to gradually reduce error.
RBF neural needs once to input all sample datas in RBF network input in each training, then calculates
The output of all nodes of output end is as a result, the result label with sample data calculates mean square error MSE, if MSE reaches target
Value then stops, less than then adding new neuron according to training speed Speed and calculate new weight, until MSE reaches target value,
Or maximum frequency of training:
(1) output end and sample results label deviation maximum are chosen as the data center of newly-increased neuron；
(2) according to new hidden layer configuration, each output weight is recalculated with gradient descent method.
Step 6) error calculation
After the adjustment of the center of progress, the sample according to next worst error is repeated in turn, and increase a nerve
Member reduces error so always repeatedly, reach until error as defined in error performance or neuronal quantity reach
In limited time, entire networking just terminates.Specific calculating process is as shown in Figure 9.RBF neural after training is to original sample data
Fitting it is as shown in Figure 10.
Step 7) Tendency Prediction
After the RBF neural network model after being trained, input serial number 103 (corresponding " moon in year-" field 2017 01
Month) 01 month 2017 whole world GSO satellite network N data can be obtained declare trend prediction as 24.3 parts/month.
The comparison of step 8) error
The global GSO satellite network N data Shen of 01 month 2017 actual count is 19 parts, with 24.3 part/month of prediction result
Comparison, relative error is about 27%.
According to the trend prediction process for declaring situation, computing repeatedly above, to 2017 year GSO/NGSO global network
A, shown in the prediction proof list 1 of C, N data:
Table 1:GSO global satellite network/N data declares trend prediction table
It should be noted last that the above examples are only used to illustrate the technical scheme of the present invention and are not limiting.Although ginseng
It is described the invention in detail according to embodiment, those skilled in the art should understand that, to technical side of the invention
Case is modified or replaced equivalently, and without departure from the spirit and scope of technical solution of the present invention, should all be covered in the present invention
Scope of the claims in.
Claims (4)
1. a kind of prediction technique for declaring situation towards satellite network data, which comprises
Satellite declaration material over the years is counted as sample data；
Sample data is pre-processed；
The RBF neural network model that Tendency Prediction is declared for satellite network data is established according to pretreated sample data；
It is declared and is predicted using satellite network data of the RBF neural network model to the specified time.
2. the prediction technique according to claim 1 for declaring situation towards satellite network data, which is characterized in that described to defend
The step of star network data declares situation specifically includes:
Step S1) sample data is serialized, the sample data serialized；
The time is declared according to satellite declaration material, is counted by month, obtains defending with the whole world that " part/moon " is statistical unit
Star declares N data bank table；" moon in year-" field is serialized, the form of " moon in year-" is replaced with sequence of natural numbers, obtains sequence
The sample data of columnization；
Step S2) in fixed time period data cleansing is carried out to the sample data of serializing；
Step S3) to step S2) treated sample data carry out queue resist it is bad.
3. the prediction technique according to claim 1 for declaring situation towards satellite network data, which is characterized in that described
Data sample after Data preprocess establishes the step of declaring the RBF neural network model of Tendency Prediction for satellite network data tool
Body includes；
Step T1) initialize the RBF neural network model that Tendency Prediction is declared for satellite network data；
The RBF neural network model uses excitation function of the Gaussian bases as neuron；Its node is single input and list
Output；
Step T2) in all sample datas of RBF neural network model input terminal input, then calculate the defeated of all nodes of output end
Out as a result, choosing the center of the Gaussian bases of output result and the maximum value of sample results label deviation as neuron, meter
Calculate variance of the mean square error MSE of output result and sample results label as neuron Gaussian bases；
Step T3) neuron of step T2) is added in the hidden layer of RBF neural network model；Construct new RBF nerve net
Network model, is transferred to step 2), until mean square error MSE is less than threshold value or neuronal quantity reaches the upper limit, the RBF nerve net
Network model construction finishes.
4. the prediction technique according to claim 1 for declaring situation towards satellite network data, which is characterized in that described to defend
It includes: that total satellite data declares number, the satellite data of some country declares number, a certain frequency range that star network data, which is declared,
Satellite data declares number and the satellite data of a certain rail position declares number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910072478.8A CN109858693A (en) | 2019-01-25 | 2019-01-25 | A kind of prediction technique for declaring situation towards satellite network data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910072478.8A CN109858693A (en) | 2019-01-25 | 2019-01-25 | A kind of prediction technique for declaring situation towards satellite network data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109858693A true CN109858693A (en) | 2019-06-07 |
Family
ID=66896151
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910072478.8A Pending CN109858693A (en) | 2019-01-25 | 2019-01-25 | A kind of prediction technique for declaring situation towards satellite network data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109858693A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111222776A (en) * | 2019-12-31 | 2020-06-02 | 中国科学院国家空间科学中心 | Satellite network coordination situation assessment method and system based on convolutional neural network |
| CN113162676A (en) * | 2021-03-26 | 2021-07-23 | 天津(滨海)人工智能军民融合创新中心 | GSO rail position efficiency evaluation method based on rail position multistage joint risk |
| CN113296743A (en) * | 2021-05-17 | 2021-08-24 | 天津(滨海)人工智能军民融合创新中心 | SpaceCap software mapping method based on GSO C data declaration |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102487343A (en) * | 2010-12-06 | 2012-06-06 | 中国人民解放军理工大学 | Diagnosis and prediction method for hidden faults of satellite communication system |
| CN105825271A (en) * | 2016-03-21 | 2016-08-03 | 南京邮电大学 | Satellite fault diagnosis and prediction method based on evidential reasoning (ER) |
| CN106446506A (en) * | 2016-08-26 | 2017-02-22 | 航天东方红卫星有限公司 | Multi-method-combined automatic and intelligent prediction method of moonlet state |
| CN107945146A (en) * | 2017-11-23 | 2018-04-20 | 南京信息工程大学 | A kind of space-time Satellite Images Fusion method based on depth convolutional neural networks |
| CN108881250A (en) * | 2018-06-28 | 2018-11-23 | 广东电网有限责任公司 | Powerline network security postures prediction technique, device, equipment and storage medium |
-
2019
- 2019-01-25 CN CN201910072478.8A patent/CN109858693A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102487343A (en) * | 2010-12-06 | 2012-06-06 | 中国人民解放军理工大学 | Diagnosis and prediction method for hidden faults of satellite communication system |
| CN105825271A (en) * | 2016-03-21 | 2016-08-03 | 南京邮电大学 | Satellite fault diagnosis and prediction method based on evidential reasoning (ER) |
| CN106446506A (en) * | 2016-08-26 | 2017-02-22 | 航天东方红卫星有限公司 | Multi-method-combined automatic and intelligent prediction method of moonlet state |
| CN107945146A (en) * | 2017-11-23 | 2018-04-20 | 南京信息工程大学 | A kind of space-time Satellite Images Fusion method based on depth convolutional neural networks |
| CN108881250A (en) * | 2018-06-28 | 2018-11-23 | 广东电网有限责任公司 | Powerline network security postures prediction technique, device, equipment and storage medium |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111222776A (en) * | 2019-12-31 | 2020-06-02 | 中国科学院国家空间科学中心 | Satellite network coordination situation assessment method and system based on convolutional neural network |
| CN113162676A (en) * | 2021-03-26 | 2021-07-23 | 天津(滨海)人工智能军民融合创新中心 | GSO rail position efficiency evaluation method based on rail position multistage joint risk |
| CN113162676B (en) * | 2021-03-26 | 2022-07-29 | 天津(滨海)人工智能军民融合创新中心 | GSO rail position efficiency evaluation method based on rail position multistage joint risk |
| CN113296743A (en) * | 2021-05-17 | 2021-08-24 | 天津(滨海)人工智能军民融合创新中心 | SpaceCap software mapping method based on GSO C data declaration |
| CN113296743B (en) * | 2021-05-17 | 2022-07-22 | 天津(滨海)人工智能军民融合创新中心 | SpaceCap software mapping method based on GSO C data declaration |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Asadi et al. | A new hybrid artificial neural networks for rainfall–runoff process modeling | |
| Athanassopoulos et al. | A comparison of data envelopment analysis and artificial neural networks as tools for assessing the efficiency of decision making units | |
| CN111563706A (en) | Multivariable logistics freight volume prediction method based on LSTM network | |
| CN109118013A (en) | A kind of management data prediction technique, readable storage medium storing program for executing and forecasting system neural network based | |
| CN103745273B (en) | Semiconductor fabrication process multi-performance prediction method | |
| Claveria et al. | Data pre-processing for neural network-based forecasting: does it really matter? | |
| CN109858693A (en) | A kind of prediction technique for declaring situation towards satellite network data | |
| CN104468413B (en) | A kind of network service method and system | |
| Thiesson et al. | ARMA time-series modeling with graphical models | |
| CN107403296A (en) | Conveyance equilibrium method and device | |
| CN107798426A (en) | Wind power interval Forecasting Methodology based on Atomic Decomposition and interactive fuzzy satisfying method | |
| CN115906954A (en) | Multivariate time sequence prediction method and device based on graph neural network | |
| CN110941902B (en) | Lightning stroke fault early warning method and system for power transmission line | |
| Lubis et al. | Optimization of MSE Accuracy Value Measurement Applying False Alarm Rate in Forecasting on Fuzzy Time Series based on Percentage Change | |
| Purnawansyah et al. | An inflation rate prediction based on backpropagation neural network algorithm | |
| Ansari et al. | Sequential combination of statistics, econometrics and Adaptive Neural-Fuzzy Interface for stock market prediction | |
| CN107704944B (en) | Construction method of stock market fluctuation interval prediction model based on information theory learning | |
| CN110322055A (en) | A kind of method and system improving data risk model scoring stability | |
| CN117114190B (en) | River runoff prediction method and device based on mixed deep learning | |
| Ghosh | Comparative study of financial time series prediction by artificial neural network with gradient descent learning | |
| Christina et al. | Gold price prediction using type-2 neuro-fuzzy modeling and ARIMA | |
| Dornelles et al. | Artificial neural network methods applied to forecasting river levels | |
| NAPITUPULU | Artificial neural network application in gross domestic product forecasting an Indonesia case | |
| CN109711848A (en) | A kind of matching system and its construction method, matching process of financial transaction | |
| Zhang et al. | Two stage prediction model of sunspots monthly value based on CEEMDAN and particle swarm optimization ELM |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190607 |