CN109145434A - A method of broadcast ephemeris orbit error is predicted using improved BP - Google Patents
A method of broadcast ephemeris orbit error is predicted using improved BP Download PDFInfo
- Publication number
- CN109145434A CN109145434A CN201810932222.5A CN201810932222A CN109145434A CN 109145434 A CN109145434 A CN 109145434A CN 201810932222 A CN201810932222 A CN 201810932222A CN 109145434 A CN109145434 A CN 109145434A
- Authority
- CN
- China
- Prior art keywords
- neural network
- broadcast ephemeris
- orbit error
- data
- error
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/01—Satellite radio beacon positioning systems transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
- G01S19/13—Receivers
- G01S19/23—Testing, monitoring, correcting or calibrating of receiver elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- 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
- G06N3/045—Combinations of networks
Abstract
A kind of method using improved BP prediction broadcast ephemeris orbit error disclosed by the invention, belongs to satellite navigation data processing technology field.Implementation method of the present invention is data needed for predicting broadcast ephemeris orbit error for BP neural network by acquisition；And data preprocessing operation is executed to data needed for BP neural network；Then BP neural network model is constructed using pretreated data；Pass through the initial weight and threshold value of PSO algorithm optimization BP neural network model again；Improved BP neural network model is trained again；The precision that improved BP neural network model is finally assessed using test data, is predicted and is compensated to broadcast ephemeris orbit error.The present invention can reduce the susceptibility to initial parameter, avoids falling into local minimum, and can effectively improve Satellite Orbit Determination precision, reduces the system-level error of satellite navigation system.
Description
Technical field
The present invention relates to a kind of methods using improved BP prediction broadcast ephemeris orbit error, belong to satellite and lead
Boat technical field of data processing.
Background technique
Beidou satellite navigation system (BeiDou Navigation Satellite System, BDS) is that country is important
Space fundamental facilities, master station are extrapolated using the determining value of track and form broadcast ephemeris, and receiver is joined according to broadcast ephemeris
Number can calculate the real time position of satellite orbit, and user is just able to carry out self poisoning by the position of several satellites and leads
Boat, therefore trajectory accuracy is one of the key factor for restricting the urban satellite navigation service performance.
A kind of method for improving orbit determination accuracy is to establish more accurate system mathematic model, and satellite in actual operation can
Influenced by various perturbative forces, it is known that some perturbation factors include earth gravitational field, tide, atmospheric drag, the theory of relativity effect
Answer, UT1 short-period term, earth tide, solar light pressure, N body perturbation etc., theoretically, the more accurate then orbit determination accuracy of perturbative force model
It is higher, but the characteristics such as complexity due to these space perturbation factors and model parameter high dynamic, it understands between variable completely
Causality, it is also relatively difficult to improve orbit determination accuracy by refining kinetic model.
In navigation data processing practice, it is found that there are some periodic regular phenomenons in broadcast ephemeris orbit error,
So breakthrough can also be found from the rule of orbit error in addition to further refining other than kinetic model, to it is uncertain because
Element is studied, and provides prediction and compensation under certain precision meaning for model trajectory error.Neural network is as a kind of adaptive
The modeling tool answered does not need the structural parameters for accurately knowing input-output function when handling non-linear and higher-dimension problem,
As long as the inner link between them can be grasped by training.In existing neural network, BP (Back Propagation)
Neural network is a kind of multitiered network " backstepping " learning algorithm, but BP neural network is one based on gradient information non-linear
Optimization problem is easily trapped into local minimum；And poor robustness, it is more sensitive to parameter initial setting up.
In summary problem establishes more accurate mathematical model undoubtedly right and wrong for improving broadcast ephemeris orbit determination accuracy
Often it is important, but be also it is extremely complex, need the research that deepens continuously for a long time.
Summary of the invention
For the complex problem of the method for improving orbit determination accuracy by kinetic model of refining in the prior art, utilize
BP neural network trains the Nonlinear Mapping relationship of broadcast ephemeris orbit error and relevant parameter, but the BP neural network
There are following defects: being easily trapped into local minimum；And poor robustness, it is more sensitive to parameter initial setting up.It is of the invention public
A kind of method technical problems to be solved using improved BP prediction broadcast ephemeris orbit error opened are: realizing benefit
Broadcast ephemeris orbit error is predicted with improved BP.This method has the advantages that (1) passes through particle swarm algorithm
The initial weight and threshold value of (Particle Swarm Optimization, PSO) Optimized BP Neural Network are reduced to initial ginseng
Several susceptibilitys avoids falling into local minimum；(2) by training PSO algorithm optimization BP neural network model, realize that track misses
The prediction and compensation of difference, can effectively improve Satellite Orbit Determination precision, reduce system-level error.
The purpose of the present invention is what is be achieved through the following technical solutions.
A kind of method using improved BP prediction broadcast ephemeris orbit error disclosed by the invention, passes through acquisition
Data needed for predicting broadcast ephemeris orbit error for BP neural network；And it is pre- to execute data to data needed for BP neural network
Processing operation；Then BP neural network model is constructed using pretreated data；Pass through PSO algorithm optimization BP neural network again
The initial weight and threshold value of model；Improved BP neural network model is trained again；Finally commented using test data
The precision for estimating improved BP neural network model predicted and compensated to broadcast ephemeris orbit error, and this method can have
Effect improves Satellite Orbit Determination precision, reduces the system-level error of satellite navigation system.
A kind of method using improved BP prediction broadcast ephemeris orbit error disclosed by the invention, including it is as follows
Step:
Step 1: acquisition is for data needed for BP neural network prediction broadcast ephemeris orbit error, comprising: epoch reference
Moment, satellite position and speed, perturbation correction, broadcast ephemeris orbit error.
Step 1.1: downloading Beidou satellite navigation system (BeiDou Navigation Satellite System, BDS)
Broadcast ephemeris, it is fast with reference to the satellite three-dimensional position vector (X, Y, Z) and three-dimensional of moment T to resolve epoch using radio news program
Spend vector (Vx,Vy,Vz), and extract corresponding perturbation correction in broadcast ephemeris, including orbit inclination angle change rateAscending node
Right ascension change rateTo the correction value delta n of mean angular velocity, to the corrected value C of latitude amplitude cosineuc, sinusoidal to latitude argument
Corrected value Cus, to the corrected value C of orbit radius cosinerc, to the corrected value C of orbit radius siners, to the school of inclination angle cosine
Positive value Cic, to the corrected value C of inclination angle sineis；
Step 1.2: downloading the precise ephemeris of BDS, acquire corresponding broadcast ephemeris epoch with reference to the satellite position (X at moments,
Ys,Zs), it, will be in the satellite position (X, Y, Z) and precise ephemeris in step 1.1 under unified space-time datum and reference frame
Satellite position (Xs,Ys,Zs) make difference to get BDS broadcast ephemeris orbit error (Δ X, Δ Y, Δ Z).
Step 2: data preprocessing operation is executed to data needed for BP neural network.
Step 2.1: deleting the clear data and abnormal data in step 1 broadcast ephemeris orbit error data, and formed such as
Data matrix shown in formula (1):
Wherein j is the total sample number in data set.
Step 2.2: the data matrix in step 2.1 being normalized into section [- 1,1], so that each of data matrix
Parameter is all unified to section [- 1,1], and new data matrix is obtained；
Step 2.3: data matrix new in step 2.2 being divided into training sample set and test sample collection, i.e. completion number
Data preprocess operation.
Step 3: BP neural network model is constructed based on step 1 and step 2.
Step 3.1: determining input layer and output layer neuron.Inputting layer parameter includes referring to the moment epoch, and three
Direction position and speed parameter and nine perturbation corrections amount to 16 parameters, and output parameter is that three direction tracks miss
Difference；
Step 3.2: determining hidden neuron.BP neural network hidden neuron number l is determined using formula (2):
Wherein, m is the number of input layer, and n is the number of output layer neuron,It indicates to be rounded symbol downwards,
A is the integer of [1,10]；
Step 3.3: determining transmission function.Hidden layer transmission function uses S type tangent function, and expression formula isOutput layer transmission function uses linear function, expression formula fo(x)=kx.
So far, building BP neural network model is completed.
Step 4: executing particle swarm algorithm (Particle Swarm Optimization, PSO) Optimized BP Neural Network mould
Type.
Step 4.1: determining PSO algorithm optimization object；Using the set of weight and threshold value in BP neural network model as
The PSO algorithm parameter to be optimized, number of parameters are the dimension d of each particle, are determined using formula (3):
D=ml+ln+l+n (3)
Step 4.2: primary group is constructed according to the particle dimension that step 4.1 determines；N number of d dimension particle structure is generated at random
Primary group is built, the position vector of i-th of particle is expressed as xi=[xi1,xi2,…,xid]T, velocity vector is expressed as vi=
[vi1,vi2,…,vid]T, particle personal best particle is denoted as p until current iterationi=[pi1,pi2,…,pid]T, global optimum
Position is denoted as pg=[pg1,pg2,…,pgd]T；
Step 4.3: the fitness function of particle is calculated, is determined using formula (4):
Wherein M indicates the number of training sample, Ppx,Ppy,PpzP-th of sample is respectively indicated to export in X/Y/Z direction prediction
Broadcast ephemeris orbit error value, Tpx,Tpy,TpzP-th of sample is respectively indicated in the actual broadcast ephemeris track in the direction X/Y/Z
Error amount；
Step 4.4: according to all individuals in the fitness value evaluation population in step 4.3, and updating current particle
Individual optimal value piWith global optimum pg；
Step 4.5: use formula (5), the speed of (6) more new particle and position:
V (t+1)=wv (t)+c1·r1·(pid-x(t))+c2·r2·(pgd-x(t)) (5)
X (t+1)=x (t)+v (t+1) (6)
Wherein t indicates that the t times iteration updates, and w indicates inertia weight, c1And c2It indicates accelerated factor (learning coefficient), r1With
r2For the random number between [0,1]；
Step 4.6: iterative calculation output optimal particle；When fitness value Fit is less than setting value or reaches the number of iterations
PSO algorithm terminates, by global optimum pgIt is mapped to the weight and threshold value of BP neural network model, i.e. completion PSO algorithm improvement
BP neural network model afterwards.
Step 5: the improved BP neural network model of training.The improved BP neural network model that step 4 is obtained
Suboptimization again is carried out according to gradient descent method, setting learning rate makes improved BP neural network model further small range
Search, obtains final broadcast ephemeris orbit error prediction model.
Step 6: the precision of final broadcast ephemeris orbit error prediction model in appraisal procedure 5.By the test in step 2
Collect broadcast ephemeris orbit error prediction model final in data steps for importing 5, obtain the predicted value of broadcast ephemeris orbit error,
Predicted value and actual broadcast ephemeris orbit error value are compared, assessment models precision, when model accuracy meets default comment
When estimating required precision, broadcast ephemeris orbit error is compensated using the model, effectively improves Satellite Orbit Determination precision, reduction is defended
The system-level error of star navigation system.
The utility model has the advantages that
1, a kind of method using improved BP prediction broadcast ephemeris orbit error disclosed by the invention, passes through
The initial weight and threshold value of PSO algorithm optimization BP neural network can reduce BP neural network to the susceptibility of initial parameter, keep away
Exempt from BP neural network and falls into local minimum.
2, a kind of method using improved BP prediction broadcast ephemeris orbit error disclosed by the invention, passes through instruction
Practice the BP neural network model of PSO algorithm improvement, there can be preferable capability of fitting to broadcast ephemeris orbit error, realize wide
The prediction and compensation for broadcasting ephemeris orbit error, effectively improve Satellite Orbit Determination precision, reduce the system-level error of satellite navigation system.
Detailed description of the invention
Fig. 1 is the implementation process in the embodiment of the method 1 using improved BP prediction broadcast ephemeris orbit error
Flow chart；
Fig. 2 is the BP nerve net in the embodiment of the method 1 using improved BP prediction broadcast ephemeris orbit error
The structure chart of network；
Fig. 3 is that the PSO algorithm in the embodiment of the method 1 using improved BP prediction broadcast ephemeris orbit error is excellent
Change the flow diagram of BP neural network parameter；
Fig. 4 is three directions in the embodiment of the method 1 using improved BP prediction broadcast ephemeris orbit error
The contrast simulation result figure of model predictive error and actual track error；
Fig. 5 is rail after the compensation in the embodiment of the method 1 using improved BP prediction broadcast ephemeris orbit error
The simulation result diagram of road error level comparison.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and examples and detailed description.
Embodiment 1
The present embodiment illustrate the present invention is based on improved BP prediction broadcast ephemeris orbit error method
Specific embodiment in the practical BDS broadcast ephemeris of application.Fig. 1 is implementation process flow chart of the invention, implements step
It is as follows:
Step 1: acquisition is for data needed for BP neural network prediction broadcast ephemeris orbit error, comprising: epoch reference
Moment, satellite position and speed, perturbation correction, broadcast ephemeris orbit error.
Step 1.1: the present embodiment acquires the data in total 10 days of 1 to 10 October in 2017, and the sampling interval is 1 small
When, from NASA crustal dynamics data information system (Crustal Dynamics Data Information System,
CDDIS BDS broadcast ephemeris) is downloaded, the satellite three-dimensional position vector (X, Y, Z) of epoch-making moment T is calculated using radio news program
With three dimensional velocity vectors (Vx,Vy,Vz), and extract corresponding perturbation correction in broadcast ephemeris, including orbit inclination angle change rateRight ascension of ascending node change rateTo the correction value delta n of mean angular velocity, to the corrected value C of latitude amplitude cosineuc, to latitude
Spend the corrected value C of argument sineus, to the corrected value C of orbit radius cosinerc, to the corrected value C of orbit radius siners, to inclining
The corrected value C of angle cosineic, to the corrected value C of inclination angle sineis；
Step 1.2: equally acquisition amounts to 10 days BDS precise ephemeris data, sampling interval in 1 to 10 October in 2017
It is 1 hour, the present embodiment is from international GNSS monitoring and evaluating system (International GNSS Monitoring&
Assessment System, IGMAS) downloading, corresponding broadcast ephemeris epoch is extracted with reference to the satellite position (X at moments,Ys,Zs),
Under unified space-time datum and reference frame, by the satellite position in the satellite position (X, Y, Z) and precise ephemeris in step 1.1
Set (Xs,Ys,Zs) make difference to get BDS broadcast ephemeris orbit error (Δ X, Δ Y, Δ Z).
Step 2: data preprocessing operation is executed to data needed for BP neural network.
Step 2.1: deleting the clear data and abnormal data in step 1 broadcast ephemeris orbit error data, and formed such as
Data matrix shown in formula (1):
Wherein j is the total sample number in data set.
Step 2.2: the data matrix in step 2.1 being normalized, the mapminmax carried using MATLAB
Data matrix is normalized to section [- 1,1] and obtains new data matrix by function；
Step 2.3: data matrix new in step 2.2 is divided into training sample set and test sample collection, the present embodiment
The total sample number j of middle selection is 24*10=240, wherein training sample set 216, test sample collection 24.
Step 3: BP neural network model is constructed based on step 1 and step 2.If Fig. 2 is the structure chart of the BP neural network,
It is mainly made of input layer, hidden layer and output layer, the training sample set data that step 2 is obtained input BP neural network model.
Step 3.1: determining input layer and output layer neuron.Inputting layer parameter includes one with reference to moment, three directions
Position and speed parameter and nine perturbation corrections amount to 16 parameters, and output parameter is three direction orbit errors；
Step 3.2: determining hidden neuron.BP neural network hidden neuron number l is determined using formula (2):
Wherein, m is the number of input layer, corresponding to be equal to the number that 16, n is output layer neuron, and correspondence is equal to
3,It indicates to be rounded symbol downwards, a is the integer of [1,10], and a takes 5 in the present embodiment, and can be calculated hidden neuron number is
9, then three layers of BP neural network that network structure is 16-9-3 in the present embodiment；
Step 3.3: determining transmission function.Hidden layer transmission function uses S type tangent function, expression formula are as follows:Output layer transmission function uses linear letter, and expression formula is fo (x)=kx.
So far, building BP neural network model is completed.
Step 4: executing particle swarm algorithm (Particle Swarm Optimization, PSO) Optimized BP Neural Network mould
Type, Fig. 3 are the flow diagrams of PSO algorithm optimization BP neural network parameter, and specific step is as follows for algorithm:
Step 4.1: determining PSO algorithm optimization object.Using the set of weight and threshold value in BP neural network model as
The particle swarm algorithm parameter to be optimized, number of parameters are the dimension d of each particle, are determined using formula (3):
D=ml+ln+l+n=16*9+9*3+9+3=183 (3)
Step 4.2: primary group is constructed according to the particle dimension that step 4.1 determines.Generate 200 grains at random by system
Son building primary group, the position vector of i-th of particle are expressed as xi=[xi1,xi2,…,xid]T, velocity vector is expressed as vi
=[vi1,vi2,…,vid]T, particle personal best particle is denoted as p until current iterationi=[pi1,pi2,…,pid]T, the overall situation is most
Excellent position is denoted as pg=[pg1,pg2,…,pgd]T；
Step 4.3: the fitness function of particle is calculated, is determined using formula (4):
Wherein M indicates the number of training sample, Ppx,Ppy,PpzP-th of sample is respectively indicated to export in X/Y/Z direction prediction
Broadcast ephemeris orbit error value, Tpx,Tpy,TpzP-th of sample is respectively indicated in the actual broadcast ephemeris track in the direction X/Y/Z
Error amount；
Step 4.4: according to all individuals in the fitness value evaluation population in step 4.3, and updating current particle
Individual optimal value piWith global optimum pg；
Step 4.5: use formula (5), the speed of (6) more new particle and position:
V (t+1)=wv (t)+c1·r1·(pid-x(t))+c2·r2·(pgd-x(t)) (5)
X (t+1)=x (t)+v (t+1) (6)
Wherein t indicates the number of iterations, and w indicates inertia weight, c1And c2It indicates accelerated factor, is nonnegative constant, is generally set to
2, r1And r2For the random number between [0,1]；
Step 4.6: iterative calculation output optimal particle.When fitness value Fit is less than setting value or reaches the number of iterations
PSO algorithm terminates, by global optimum pgIt is mapped to the weight and threshold value of BP neural network, i.e., after completion PSO algorithm improvement
BP neural network model.
Step 5: the improved BP neural network model of training.The improved BP neural network model that step 4 is obtained
Suboptimization again is carried out according to gradient descent method, it is 0.001 that learning rate is arranged herein, makes improved BP neural network model again
Further small range search, obtains final broadcast ephemeris orbit error prediction model.
Step 6: the precision of final broadcast ephemeris orbit error prediction model in appraisal procedure 5.
Step 6.1: broadcast ephemeris orbit error final in the test set data steps for importing 5 in step 2 is predicted into mould
Type obtains the predicted value of broadcast ephemeris orbit error, and predicted value and actual broadcast ephemeris orbit error value are compared, commented
Estimate the precision of final broadcast ephemeris orbit error prediction model, Fig. 4 is that three direction models predict that error and actual track are missed
The comparison of difference, abscissa indicate the time, and ordinate respectively indicates the orbit error in tri- directions X/Y/Z, and "+" line style indicates
The error amount of PSO algorithm optimization BP neural network model prediction, "○" line style indicate the actual error amount of the direction, it can be seen that this
Method in invention can have preferable capability of fitting and prediction effect to BDS broadcast ephemeris orbit error；
Step 6.2: providing compensation to broadcast ephemeris track Orbit Error using the model in the present embodiment, Fig. 5 is compensation
Orbit error is horizontal afterwards, and abscissa indicates the time, and ordinate indicates that whole orbit error is horizontal, and "+" line style indicates application
For model in the present embodiment to the compensated error level of Orbit Error, "○" line style indicates the actual error level of track, can
Find out and compensated using the present embodiment when resolving satellite position, can be improved BDS orbit determination accuracy, to the system-level error of reduction
It is of great significance.
The above is presently preferred embodiments of the present invention, and it is public that the present invention should not be limited to embodiment and attached drawing institute
The content opened.It is all not depart from the lower equivalent or modification completed of spirit disclosed in this invention, both fall within the model that the present invention protects
It encloses.
Claims (4)
1. a kind of method using improved BP prediction broadcast ephemeris orbit error, it is characterised in that: including walking as follows
Suddenly,
Step 1: acquisition for BP neural network prediction broadcast ephemeris orbit error needed for data, comprising: epoch with reference to the moment,
Satellite position and speed, perturbation correction, broadcast ephemeris orbit error；
Step 2: data preprocessing operation is executed to data needed for BP neural network；
Step 2.1: deleting the clear data and abnormal data in step 1 broadcast ephemeris orbit error data, and form such as formula
(1) data matrix shown in:
Wherein j is the total sample number in data set；
Step 2.2: the data matrix in step 2.1 being normalized into section [- 1,1], so that each parameter in data matrix
All unify to section [- 1,1], obtains new data matrix；
Step 2.3: data matrix new in step 2.2 being divided into training sample set and test sample collection, i.e. completion data are pre-
Processing operation；
Step 3: BP neural network model is constructed based on step 1 and step 2；
Step 4: executing particle swarm algorithm (Particle Swarm Optimization, PSO) Optimized BP Neural Network model；
Step 5: the improved BP neural network model of training；The improved BP neural network model that step 4 is obtained according to
Gradient descent method carries out suboptimization again, and setting learning rate makes improved BP neural network model, and further small range is searched
Rope obtains final broadcast ephemeris orbit error prediction model；
Step 6: the precision of final broadcast ephemeris orbit error prediction model in appraisal procedure 5；By the test set number in step 2
According to broadcast ephemeris orbit error prediction model final in steps for importing 5, the predicted value of broadcast ephemeris orbit error is obtained, it will be pre-
Measured value is compared with actual broadcast ephemeris orbit error value, assessment models precision, when model accuracy meets default assessment essence
When degree requires, broadcast ephemeris orbit error is compensated using the model, effectively improves Satellite Orbit Determination precision, satellite is reduced and leads
The system-level error of boat system.
2. a kind of method using improved BP prediction broadcast ephemeris orbit error as described in claim 1, special
Sign is: step 1 concrete methods of realizing is,
Step 1.1: downloading the wide of Beidou satellite navigation system (BeiDou Navigation Satellite System, BDS)
Ephemeris is broadcast, resolves satellite three-dimensional position vector (X, Y, Z) and three-dimensional velocity arrow of the epoch with reference to moment T using radio news program
Measure (Vx,Vy,Vz), and extract corresponding perturbation correction in broadcast ephemeris, including orbit inclination angle change rateRight ascension of ascending node
Change rateTo the correction value delta n of mean angular velocity, to the corrected value C of latitude amplitude cosineuc, to the school of latitude argument sine
Positive value Cus, to the corrected value C of orbit radius cosinerc, to the corrected value C of orbit radius siners, to the corrected value of inclination angle cosine
Cic, to the corrected value C of inclination angle sineis；
Step 1.2: downloading the precise ephemeris of BDS, acquire corresponding broadcast ephemeris epoch with reference to the satellite position (X at moments,Ys,
Zs), under unified space-time datum and reference frame, by the satellite position (X, Y, Z) in step 1.1 and defending in precise ephemeris
Championship sets (Xs,Ys,Zs) make difference to get BDS broadcast ephemeris orbit error (Δ X, Δ Y, Δ Z).
3. a kind of method using improved BP prediction broadcast ephemeris orbit error as claimed in claim 2, special
Sign is: step 3 concrete methods of realizing is,
Step 3.1: determining input layer and output layer neuron；Inputting layer parameter includes an epoch referring to moment, three directions
Position and speed parameter and nine perturbation corrections amount to 16 parameters, and output parameter is three direction orbit errors；
Step 3.2: determining hidden neuron；BP neural network hidden neuron number l is determined using formula (2):
Wherein, m is the number of input layer, and n is the number of output layer neuron,It indicates to be rounded symbol downwards, a is
[1,10] integer；
Step 3.3: determining transmission function；Hidden layer transmission function uses S type tangent function, and expression formula isIt is defeated
Layer transmission function uses linear function, expression formula f outo(x)=kx；
So far, building BP neural network model is completed.
4. a kind of method using improved BP prediction broadcast ephemeris orbit error as claimed in claim 3, special
Sign is: step 4 concrete methods of realizing is,
Step 4.1: determining PSO algorithm optimization object；It is calculated the set of weight and threshold value in BP neural network model as PSO
The method parameter to be optimized, number of parameters are the dimension d of each particle, are determined using formula (3):
D=ml+ln+l+n (3)
Step 4.2: primary group is constructed according to the particle dimension that step 4.1 determines；N number of d dimension particle building is generated at random just
The position vector of beginning population, i-th of particle is expressed as xi=[xi1,xi2,…,xid]T, velocity vector is expressed as vi=[vi1,
vi2,…,vid]T, particle personal best particle is denoted as p until current iterationi=[pi1,pi2,…,pid]T, global optimum position
It is denoted as pg=[pg1,pg2,…,pgd]T；
Step 4.3: the fitness function of particle is calculated, is determined using formula (4):
Wherein M indicates the number of training sample, Ppx,Ppy,PpzRespectively indicate p-th of sample exported in X/Y/Z direction prediction it is wide
Broadcast ephemeris orbit error amount, Tpx,Tpy,TpzP-th of sample is respectively indicated in the actual broadcast ephemeris orbit error in the direction X/Y/Z
Value；
Step 4.4: according to all individuals in the fitness value evaluation population in step 4.3, and updating the individual of current particle
Optimal value piWith global optimum pg；
Step 4.5: use formula (5), the speed of (6) more new particle and position:
V (t+1)=wv (t)+c1·r1·(pid-x(t))+c2·r2·(pgd-x(t)) (5)
X (t+1)=x (t)+v (t+1) (6)
Wherein t indicates that the t times iteration updates, and w indicates inertia weight, c1And c2It indicates accelerated factor (learning coefficient), r1And r2For
[0,1] random number between；
Step 4.6: iterative calculation output optimal particle；When fitness value Fit is less than setting value or reaches the number of iterations, PSO is calculated
Method terminates, by global optimum pgIt is mapped to the weight and threshold value of BP neural network model, i.e. BP after completion PSO algorithm improvement
Neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810932222.5A CN109145434A (en) | 2018-08-16 | 2018-08-16 | A method of broadcast ephemeris orbit error is predicted using improved BP |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810932222.5A CN109145434A (en) | 2018-08-16 | 2018-08-16 | A method of broadcast ephemeris orbit error is predicted using improved BP |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109145434A true CN109145434A (en) | 2019-01-04 |
Family
ID=64789498
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810932222.5A Pending CN109145434A (en) | 2018-08-16 | 2018-08-16 | A method of broadcast ephemeris orbit error is predicted using improved BP |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109145434A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109991634A (en) * | 2019-03-21 | 2019-07-09 | 上海埃威航空电子有限公司 | Satellite position speed data processing method based on satellite earth station |
| CN111047018A (en) * | 2020-01-14 | 2020-04-21 | 东方红卫星移动通信有限公司 | Intelligent scheduling method for mobile communication resources of low-earth-orbit satellite |
| CN111562599A (en) * | 2020-05-14 | 2020-08-21 | 中国北方工业有限公司 | Method for establishing RTD pseudo-range correction forecasting model of Beidou satellite network |
| CN112257341A (en) * | 2020-10-20 | 2021-01-22 | 浙江大学 | Customized product performance prediction method based on heterogeneous data difference compensation fusion |
| CN112560343A (en) * | 2020-12-15 | 2021-03-26 | 南京航空航天大学 | J2 perturbation Lambert problem solving method based on deep neural network and targeting algorithm |
| CN112797988A (en) * | 2020-11-18 | 2021-05-14 | 北京理工大学 | Unknown maneuvering spacecraft orbit determination method based on neural network |
| CN113155136A (en) * | 2020-12-29 | 2021-07-23 | 千寻位置网络有限公司 | Training method and device of GNSS satellite orbit prediction deviation analysis model |
| CN113608239A (en) * | 2021-07-13 | 2021-11-05 | 中国科学院国家空间科学中心 | GNSS occultation troposphere parameter correction method based on BP neural network |
| CN113780517A (en) * | 2021-08-10 | 2021-12-10 | 北京自动化控制设备研究所 | Data-driven satellite receiver fault prediction method |
| CN114742265A (en) * | 2022-03-09 | 2022-07-12 | 大连海事大学 | VHF signal propagation path error prediction method and system |
| CN116227561A (en) * | 2023-02-17 | 2023-06-06 | 电子科技大学 | Satellite orbit error prediction method based on digital twin and container technology |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103499349A (en) * | 2013-09-29 | 2014-01-08 | 桂林电子科技大学 | Medium-and-long-term forecasting method and medium-and-long-term forecasting system based on broadcast ephemeris parameter extrapolation |
| CN105403901A (en) * | 2015-12-30 | 2016-03-16 | 广州中海达卫星导航技术股份有限公司 | DGNSS satellite orbit deviation correction method |
| CN107547457A (en) * | 2017-09-15 | 2018-01-05 | 重庆大学 | A kind of approach for blind channel equalization based on Modified particle swarm optimization BP neural network |
-
2018
- 2018-08-16 CN CN201810932222.5A patent/CN109145434A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103499349A (en) * | 2013-09-29 | 2014-01-08 | 桂林电子科技大学 | Medium-and-long-term forecasting method and medium-and-long-term forecasting system based on broadcast ephemeris parameter extrapolation |
| CN105403901A (en) * | 2015-12-30 | 2016-03-16 | 广州中海达卫星导航技术股份有限公司 | DGNSS satellite orbit deviation correction method |
| CN107547457A (en) * | 2017-09-15 | 2018-01-05 | 重庆大学 | A kind of approach for blind channel equalization based on Modified particle swarm optimization BP neural network |
Non-Patent Citations (4)
| Title |
|---|
| 刘星 等: "GPS/GLONASS/BDS广播星历轨道误差分析", 《测绘科学》 * |
| 李娜等: "基于粒子群-BP神经网络算法的电价预测", 《武汉大学学报(工学版)》 * |
| 黄金等: "基于神经网络模型的地球同步卫星高精度轨道预报", 《空间科学学报》 * |
| 黄金等: "高轨卫星轨道预报中神经网络模型优化设计", 《测绘科学技术学报》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020186536A1 (en) * | 2019-03-21 | 2020-09-24 | 上海埃威航空电子有限公司 | Method for processing satellite position velocity data based on satellite earth station |
| CN109991634A (en) * | 2019-03-21 | 2019-07-09 | 上海埃威航空电子有限公司 | Satellite position speed data processing method based on satellite earth station |
| CN111047018A (en) * | 2020-01-14 | 2020-04-21 | 东方红卫星移动通信有限公司 | Intelligent scheduling method for mobile communication resources of low-earth-orbit satellite |
| CN111562599B (en) * | 2020-05-14 | 2021-07-16 | 中国北方工业有限公司 | Method for establishing RTD pseudo-range correction forecasting model of Beidou satellite network |
| CN111562599A (en) * | 2020-05-14 | 2020-08-21 | 中国北方工业有限公司 | Method for establishing RTD pseudo-range correction forecasting model of Beidou satellite network |
| CN112257341A (en) * | 2020-10-20 | 2021-01-22 | 浙江大学 | Customized product performance prediction method based on heterogeneous data difference compensation fusion |
| CN112257341B (en) * | 2020-10-20 | 2022-04-26 | 浙江大学 | Customized product performance prediction method based on heterogeneous data difference compensation fusion |
| CN112797988A (en) * | 2020-11-18 | 2021-05-14 | 北京理工大学 | Unknown maneuvering spacecraft orbit determination method based on neural network |
| CN112797988B (en) * | 2020-11-18 | 2023-04-07 | 北京理工大学 | Unknown maneuvering spacecraft orbit determination method based on neural network |
| CN112560343A (en) * | 2020-12-15 | 2021-03-26 | 南京航空航天大学 | J2 perturbation Lambert problem solving method based on deep neural network and targeting algorithm |
| CN112560343B (en) * | 2020-12-15 | 2024-04-02 | 南京航空航天大学 | J2 perturbation Lambert problem solving method based on deep neural network and targeting algorithm |
| CN113155136A (en) * | 2020-12-29 | 2021-07-23 | 千寻位置网络有限公司 | Training method and device of GNSS satellite orbit prediction deviation analysis model |
| CN113155136B (en) * | 2020-12-29 | 2022-08-05 | 千寻位置网络有限公司 | Training method and device of GNSS satellite orbit prediction deviation analysis model |
| CN113608239A (en) * | 2021-07-13 | 2021-11-05 | 中国科学院国家空间科学中心 | GNSS occultation troposphere parameter correction method based on BP neural network |
| CN113608239B (en) * | 2021-07-13 | 2023-10-17 | 中国科学院国家空间科学中心 | GNSS occultation troposphere parameter correction method based on BP neural network |
| CN113780517A (en) * | 2021-08-10 | 2021-12-10 | 北京自动化控制设备研究所 | Data-driven satellite receiver fault prediction method |
| CN114742265B (en) * | 2022-03-09 | 2023-07-18 | 大连海事大学 | VHF signal propagation path error prediction method and system |
| CN114742265A (en) * | 2022-03-09 | 2022-07-12 | 大连海事大学 | VHF signal propagation path error prediction method and system |
| CN116227561B (en) * | 2023-02-17 | 2023-09-12 | 电子科技大学 | Satellite orbit error prediction method based on digital twin and container technology |
| CN116227561A (en) * | 2023-02-17 | 2023-06-06 | 电子科技大学 | Satellite orbit error prediction method based on digital twin and container technology |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109145434A (en) | A method of broadcast ephemeris orbit error is predicted using improved BP | |
| CN110232471B (en) | Rainfall sensor network node layout optimization method and device | |
| Krasnopolsky et al. | A neural network nonlinear multimodel ensemble to improve precipitation forecasts over continental US | |
| CN107563574A (en) | A kind of earth's magnetic field Forecasting Methodology based on long memory models Recognition with Recurrent Neural Network in short-term | |
| CN107688906B (en) | Multi-method fused transmission line meteorological element downscaling analysis system and method | |
| CN105956709B (en) | A kind of modularization support vector machines tide prediction method based on GUI | |
| CN115935834B (en) | Historical fitting method based on deep autoregressive network and continuous learning strategy | |
| Ghajar et al. | An adaptive network-based fuzzy inference system for rock share estimation in forest road construction | |
| CN114861840B (en) | Multi-source precipitation data fusion method | |
| CN115471679A (en) | Method and intelligent system for synchronously assimilating water level and flow of natural river | |
| CN111797573A (en) | Ionized layer electron concentration total content time sequence prediction method based on deep learning technology | |
| Ha et al. | Error correction of meteorological data obtained with Mini-AWSs based on machine learning | |
| Adnan Ikram et al. | Novel evolutionary-optimized neural network for predicting landslide susceptibility | |
| CN116562414A (en) | Land water reserve change prediction method based on multisource hydrographic meteorological data | |
| CN115270608A (en) | Coastal zone ground settlement prediction method based on ARIMA and LSTM | |
| Han et al. | Novel Along‐Track Processing of GRACE Follow‐On Laser Ranging Measurements Found Abrupt Water Storage Increase and Land Subsidence During the 2021 March Australian Flooding | |
| Frame et al. | Post processing the US national water model with a long short-term memory network | |
| CN113011455B (en) | Air quality prediction SVM model construction method | |
| CN114676621A (en) | Method for improving accuracy of land water reserve abnormity based on deep learning weight load | |
| CN112307536B (en) | Dam seepage parameter inversion method | |
| CN116628442B (en) | Groundwater reserve change space downscaling method based on artificial neural network | |
| Pashova et al. | Gap filling of daily sea levels by artificial neural networks | |
| CN116822185A (en) | Daily precipitation data space simulation method and system based on HASM | |
| CN113642785B (en) | Method, system and equipment for long-term prediction of space debris track based on priori information | |
| CN113139327B (en) | Ionized layer TEC single-point prediction method and system based on GRU network model |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190104 |