CN105759207B - A kind of ocean current generator insulation system fault detection method based on M-EKF algorithms - Google Patents
A kind of ocean current generator insulation system fault detection method based on M-EKF algorithms Download PDFInfo
- Publication number
- CN105759207B CN105759207B CN201610298086.XA CN201610298086A CN105759207B CN 105759207 B CN105759207 B CN 105759207B CN 201610298086 A CN201610298086 A CN 201610298086A CN 105759207 B CN105759207 B CN 105759207B
- Authority
- CN
- China
- Prior art keywords
- formula
- matrix
- current
- insulation system
- model structure
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/34—Testing dynamo-electric machines
- G01R31/343—Testing dynamo-electric machines in operation
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/34—Testing dynamo-electric machines
- G01R31/346—Testing of armature or field windings
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16Z—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS, NOT OTHERWISE PROVIDED FOR
- G16Z99/00—Subject matter not provided for in other main groups of this subclass
Abstract
The invention discloses a kind of ocean current generator insulation system fault detection methods based on M EKF algorithms.The first step establishes rational state space equation according to the insulation system preference pattern structure to be monitored.Second step utilizes improved continuous Extended Kalman filter (M EKF) method based on continuous EKF so that the parameter of its more acurrate identification model structure.Include mainly using M EKF methods：The model structure of fixed subcoil insulating system is obtained, parametric variable needed for state space modeling is chosen, carries out state space modeling using traditional continuous Extended Kalman filter, M EKF methods is utilized to improve state space modeling, obtain current parameters identification result in model structure.
Description
Technical field：
The present invention relates to the ocean current generator insulation system fault detects in new energy field, more particularly to one kind being based on M-
The ocean current generator insulation system fault detection method of EKF algorithms.
Background technology：
As traditional energy gradually uses up totally, the resource and environmental problem of facing mankind sternness.Whole world energy by ocean current
Theoretical assessment is about 108kW magnitudes, if energy by ocean current can be utilized reasonably, it will can not to the survival and development offer of the mankind
The energy of appraisal.Current, regenerable marine energy has become an important participant of society, but in severe marine environment
In, such as high humility, high salt fog and thermal cycle etc. bring a series of stern challenges to the safe operation of ocean current generator.It is special
It is not that it is likely to result in the insulation system damage of stator and rotor, effective method is the health status of continuous monitoring insulation,
So the state of real time on-line monitoring insulation system is critically important.
Mainly have to the method for insulation system fault diagnosis at present：Shelf depreciation method and motor current signal analytic approach.Office
Portion's electric discharge has more than the fault detect of insulation system 80 years history, in the 1940s, Johnson is considered as
Shelf depreciation method is used high-tension apparatus by first, but this testing result is affected by environment bigger.Current of electric is believed
Number analytic approach is considered as applying a kind of popular method, but its precision is not high.When using intelligent algorithm, example
What such as Kalman Filtering for Discrete was extremely successful is applied to communication, positioning and navigation other field.But its real-time
It is not high, it generally requires first to collect gathered data then again to data progress batch processing.It insulate at present for machine winding coil
The fault detection problem of system is primarily present following difficulty：(1) instrument excessive cost needed for is very high；(2) detection real-time compared with
Low, processing is more difficult and computationally intensive；(3) it is not high often to export diagnostic accuracy for current detection method.
It is accounted between the 21% of total failare and 40% with the relevant failure of insulation system according to the type and size of motor.
Invention content：
Technical problem to be solved by the present invention lies in for the above-mentioned prior art the technical issues of and a kind of base for proposing
In the ocean current generator insulation system fault detection method of M-EKF algorithms, purpose is exactly to exist to overcome in prior art
Real-time it is poor, it is of high cost, convergence it is slow the shortcomings of, by using output error minimize technology and its relative sensitivity function it is true
Vertical model structure, and the state variable of selected model structure, establish state space equation, secondly by improvement for model structure
Continuous Extended Kalman filter (M-EKF) method carry out parameter identification.Parameter identification is the mould according to experimental data and foundation
Type determines one group of parameter value so that fitting test data that can be best by the numerical result that model is calculated.Experimental provision
Including coupling box, impulse generator, high-frequency signal collector and Roebel coil insulation system.Impulse generator passes through coupling box
It connect to form closed circuit with Roebel coil insulation system, high-frequency signal collector is connected to the resistance R of coupling boxm, it is used for
Acquire current signal.Roebel coil insulation system is exactly using bar made of Roebel transposition mode, i.e., inside bar
More stock conducting wires carry out coordinated transposition in bar straightway, by the change of cable space position, make the magnetic flux of each strand interlinkage
It is as balanced as possible, of substantially equal induced potential is generated, to eliminate the inside circulation in bar, bar loss is reduced, as ocean current
The generator of the large capacity of generator is all made of this bar.First, Roebel coil insulation system passes through output error minimum
Change technology and its relative sensitivity Function identification method obtain a five parameter R1L1C1/R2C2Model structure, wherein R1, L1, C1
It is concatenated, position is arranged in order from top to bottom, R2And C2It is concatenated, position is arranged in order from top to bottom, and and R1L1C1And
Connection.R1L1C1And R2C2After parallel connection, it is divided into the resistance R of other series coupled casemWith inductance lc, resistance RmIt connects from top, inductance lcFrom
R1L1C1And R2C2The following series connection of parallel-connection structure.Rm=47 Ω are to measure resistance, lc=0.8 × 10-6H is the inductance of coupler.
Then it is connected to form closed circuit with impulse generator.Its five parameter values are respectively：R1=908.9388 Ω, L1=
0.019859H, C1=2.0871e-9F, R2=20.484 Ω and C2=5.6612e-10F.M-EKF algorithms are in continuous expansion card
It improves and obtains on the basis of Kalman Filtering.Because traditional continuous Extended Kalman filter is when linearizing nonlinear system,
During Taylor expansion, Jacobian matrix is utilizedObtained lienarized equation, traditional continuous expansion
Exhibition Kalman filtering only accounts for the single order information of Taylor expansion, so error can be very big, this can reduce the essence of parameter identification
Exactness.And improved continuous Extended Kalman filter (Modified-ExtendedKalmanFilter, M-EKF) method is base
In continuous Extended Kalman filter (EKF) and second order information, by calculating the second order information of Taylor expansion, and in second order
Regulatory factor (M) is carried before information, regulatory factor M is obtained according to grid data service, i.e. F=F0+M1×F1+M2×F2
+......+Mn×Fn, F1It is state transition matrix f (x, t) partial differential x1Hessian matrix, x1It is the rendezvous value x of state variable
First component, M1It is F1Preceding coefficient.Subsequent M2, until MnAnd so on.Covariance P and karr can be calculated after obtaining F
Graceful gain K.And pass throughThe set x for constantly updating state variable, is exactly constantly updated
The first component value and second component value of value set x are exported, state prison can be reached by obtaining the sum of the first component and second component
Control the purpose of total current i.
The present invention is as follows to Roebel coil insulation system model Identification of Structural Parameters method and step using M-EKF algorithms：
Step 1 chooses state variable from model structure；
State variable is exactly the value for needing to estimate in Kalman filtering system.State variable takes the original randomly selected
Then, but choose state variable will for the purpose of monitoring total current i, because electric current i can reflect the variation of model structure parameter, into
And reflect the health status of stator insulation coil.The number n of state variable has to consistent with the exponent number of Jacobian matrix.
Step 2 carries out state space modeling using continuous Extended Kalman filter to model structure；
On the basis of step 1, after selected state variable, the collection of state variable shares vector x expression, continuously extends karr
The purpose of graceful filtering algorithm is exactly to constantly update the set x of state variable, including output valve electric current.Traditional continuous extension karr
Graceful filtering algorithm is described as follows：
Nonlinear dynamical model can be written as follows form：
Nonlinear measurement models are as follows：
Z (t)=h (x (t), t)+v (t) v (t)~N (0, R (t)) (2)
Wherein w (t) and v (t) is process and observation noise respectively, and is considered as white Gaussian noise, covariance point
It Wei not Q (t) and R (t).
The differential equation of state estimation can be write as：
The measurement amount of prediction is：
Calculate Jacobian matrix：
Covariance formula calculates as follows：
Kalman gain equation calculation formula is as follows:
K (t)=P (t) HT(t)R-1(t) (8)
(1) sieve shellfish must be fully described according to (1) formula in the nonlinear dynamical model that formula represents, the differential equation to be established
The model structure of your coil insulation system.(2) nonlinear measurement models of formula seek to be represented with state variable required
Observed quantity, i.e. output valve.(7) in formula it is the differential equation about covariance, passes through (8) formula and constantly update kalman gain and substitute into
(7) formula constantly updates covariance, and the value of covariance substitutes into (8) formula update kalman gain matrix again.(8) Kalman of formula increases
Beneficial matrix substitutes into (3) formula to update the set x of state variable, including the output current value to be monitored.
Step 3 corrects modeling error using M-EKF algorithms；
(5) formula is calculating in step 2When, the single order information of Taylor expansion is only only accounted for,
Have ignored second order and its more than order of information.Therefore, larger error can be introduced, the accurate of parameter identification is greatly reduced
Property.In other words, estimated accuracy cannot be guaranteed.The improved continuous expanded Kalman filtration algorithm of the present invention incorporates second order information.
In order to reduce parameter identification error, the present invention determines to make up above-mentioned parameter Identification Errors, the following institute of principle using M-EKF algorithms
Show：(5) when linearisation has been calculated in formula on the basis of Jacobian matrix, it is contemplated that the second order information of Taylor expansion, and counting
Regulatory factor (M) is carried before calculating second-order matrix.In the present invention, M-EKF be based on traditional continuous Extended Kalman filter, and
And carry regulatory factor (M).Linear approximation equation can be written as follows form again：
F=F0+M1×F1+M2×F2+......+Mn×Fn (9)
Wherein M1,M2And MnIt is referred to as regulatory factor, F0It is exactly first approximation Jacobian matrix, is equal to (5) formula, the numerical value of n
It is the number of state variable, and consistent with formula (5) order of matrix number.In actual mechanical process, the selection of regulatory factor can
To be obtained by grid data service approximation.
Wherein FiState transition matrix f (x (t), u (t)) partial differential xiHessian matrix, Fi(i=1,2 ..., n).
xiIt is exactly i-th of component of state variable set x, i.e. i-th of state variable.By (10) formula, Taylor expansion can be calculated
(10) formula is substituted into (9) formula by the information of formula second order, and here it is the Jacobian matrix after making up, error will smaller.Then it substitutes into
Operation is participated in (7) and (8) formula.Obtaining (8) formula can substitute into obtain the rendezvous value x of the state variable of (3) formula, wherein rendezvous value
The first components of x i1With second component i2The sum of be exactly to export result：Electric current i, as the status monitoring amount in the present invention.
According to the input of impulse generator, the measurement electricity in Roebel coil insulation system can be obtained with impulse generator
Flow valuve, this current value is as the z (t) in (3).Then according to the original of improved continuous Extended Kalman filter (M-EKF) algorithm
Reason is carried out the parameter identification of total current i, is achieved the effect that ocean current generator insulation system fault detect with this.
Description of the drawings
The holistic approach flow chart that Fig. 1 needs for institute's research experiment system；
Fig. 2 is main the experimental system platform explanation of the present invention, includes mainly oscillograph, impulse generator and is studied
Roebel coil insulation system；
Fig. 3 is of the invention from Roebel coil insulation system gained model structure；
Specific implementation mode：
In order to make the technical means, the creative features, the aims and the efficiencies achieved by the present invention be easy to understand, tie below
Conjunction is specifically illustrating, and further illustrates the present invention.
A kind of ocean current generator insulation system fault detection method based on M-EKF algorithms, main includes obtaining stator line
Enclose insulation system model structure, choose state space modeling needed for parametric variable, utilize traditional continuous spreading kalman
Filtering is carried out state space modeling, state space modeling is improved using M-EKF methods, obtains current parameters identification in model structure
As a result, obtaining being described as follows for the model structure of fixed subcoil insulating system：
Fig. 1 describes the flow chart of this method, determines model structure according to the insulation system to be studied first, then selects
The parametric variable in model structure is taken, and state space modeling is carried out to model structure on this basis, finally M-EKF is utilized to calculate
Method emulates to obtain output valve, experimental provision such as Fig. 2, including coupling box, impulse generator, high-frequency signal collector and Roebel
Insulation system is enclosed, impulse generator is connected to form closed circuit by coupling box with Roebel coil insulation system, high frequency letter
Number collector is connected to the resistance R of coupling boxm, for acquiring current signal.When impulse generator input signal, high-frequency signal
The current signal of collector acquisition enters M-EKF algorithms for recognizing current value i as actual measured value z (t).Fig. 3 is this hair
The bright model structure obtained by the Roebel coil insulation system.Far Left is high frequency voltage excitation module, by coupling box and most
The Roebel coil insulation system on the right connects.Roebel coil insulation system be exactly by output error minimum technology and its
The five parameter R that relative sensitivity function is established1L1C1/R2C2Model structure.Model structure must show its " fine " and its object
Manage the optimal compromise between the quantity of parameter.Actual conditions are must take into account in whole system modeling process, parameter Estimation
Limitation, the especially excitation of motor electrical insulation system should not upset drive control.More several moulds are being analyzed according to priori
On the basis of type, technology and its relative sensitivity function, five parameter R are minimized by output error1L1C1/R2C2The net of (Fig. 3)
Network topology is the model structure with Roebel coil insulation system best match, wherein R1, L1, C1It is concatenated, R2And C2It is string
Connection, and and R1L1C1It is in parallel.R1L1C1And R2C2After parallel connection, it is divided into the resistance R of other series coupled casemWith inductance lc, Rm=47 Ω
It is to measure resistance, lc=0.8 × 10-6H is the inductance of coupler.Then it is connected with impulse generator.Its five parameter values point
It is not：R1=908.9388 Ω, L1=0.019859H, C1=2.0871e-9F, R2=20.484 Ω and C2=5.6612e-10F.
Obtain model structure after, using M-EKF methods carry out current parameters identification specific steps are as follows：Step 1：From
State variable is chosen in model structure
Choose state variable will for the purpose of monitoring total current i, because electric current i can reflect the variation of model structure parameter,
And then reflect the health status of stator insulation coil.When the environment changes, such as larger temperature change, insulation system perhaps can
It is destroyed.Founding mathematical models describe model structure in Fig. 3, and to current variation into line trace.State variable take with
The principle that machine is chosen, but must include the quantity of state to be monitored, the state variable to be monitored in the present invention is exactly current value,
And the number of state variable is consistent with the exponent number of (8) formula Jacobian matrix, in model structure in fig. 3, i1, i2, uc, v
And R1It is selected as five state variables, i1It is R1L1C1Electric current in branch, i2It is R2C2Electric current in branch, i are R1L1C1With
R2C2Total current after parallel connection, target monitoring amount are total current i, and total current i is a point electric current i1And i2The sum of.
Step 2：State space modeling is carried out to model structure using continuous Extended Kalman filter
After step 1 selectes model structure parameter variable, so that it may to establish state according to the basic principle of Kalman filtering
Space equation describes the model structure of insulation system, and the state space equation for describing the model structure in Fig. 3 is as follows：
State space equation after above-mentioned 5 equatioies extension can be written as form：
Observational equation can be indicated with following equation：
Linear approximation equation can be calculated with Jacobi below：
Observing matrix can be indicated with following equations：
Wherein U represents the input of impulse generator；ucRepresent capacitance C1On voltage；V represents R2And C2Upper total voltage；i
It is the total current of institute's research model structure.(1) formula to (5) formula is the state space equation of descriptive model structure, in order to calculate (8)
(1) formula, is written as the form of (6) formula by the Jacobian matrix of formula to (5) formula.(8) it is needed when formula is exactly linearizing non-linear equation
The Jacobian matrix of calculating, wherein x just represent i1, i2, uc, v and R1Five state variables are selected as, (6) formula seeks partial derivative to x
Obtain the Jacobian matrix of (8) formula.Observational equation (7) formula is exactly the total current i of model structure for obtaining output valve.Observation
Matrix (9) formula is for calculating kalman gain K.The kalman gain K being calculated simultaneously is on the one hand for covariance matrix P's
It calculates, on the other hand is used to update the value x of state variable.
When being calculated using Jacobian matrix, the first-order equation of nonlinear system is only considered, there is prodigious error, institute
With step 3 this error is made up using M-EKF methods.
Step 3：Modeling error is corrected using M-EKF algorithms
(8) formula only considered the single order information of Taylor expansion when calculating Jacobian matrix, have ignored most high-order letters
Breath makes up the second order information of loss now according to the principle of M-EKF methods, calculates following formula：
(10)~(13) formula is exactly the second order correlation matrix being calculated according to M-EKF algorithms, in this insulation system, when
When using improved continuous expanded Kalman filtration algorithm (M-EKF), F3And F5All it is full null matrix, so actually calculating
They can be ignored in journey.
So Jacobian matrix can be re-written as following form：
Obtain new matrix F=F0+M1×F1+M2×F2+M3×F3, F0The Jacobian matrix being equal in (8) formula.Its
Middle F1It is state transition matrix f (x, t) partial differential x1Hessian matrix, x1It is the first component of the rendezvous value x of state variable, M1
It is F1The preceding coefficient adjustment factor.F2It is state transition matrix f (x, t) partial differential x2Hessian matrix, x2It is the collection of state variable
The second component of conjunction value x, M2It is F2The preceding coefficient adjustment factor.F3It is state transition matrix f (x, t) partial differential x3The gloomy square in sea
Battle array, x3It is the third component of the rendezvous value x of state variable, M3It is F3The preceding coefficient adjustment factor.
It is as follows to arrange obtained Jacobian matrix equation：
The new Jacobian matrix F obtained by (15) formula will replace the Jacobian matrix of (8) formula again
Into participation cycle in continuous Extended Kalman filter.The Jacobian matrix F newly obtained is for calculating covarianceWith kalman gain K (t)=P (t) HT(t)R-1(t), to
It can be according to the differential equation of state estimationConstantly update state variable value include
Output valve.Output is exactly electric current i in the present invention.Electric current i is the curve there are six pulse, wherein three pulses are upward, three arteries and veins
Downward, pulse upward is caused by raised voltage, and pulse directed downwardly is caused by drop-out voltage for punching.First electric current arteries and veins
Upward, maximum amplitude is happened between 0.8A and 1A before 50us for punching.Downward, maximum amplitude exists for second current impulse
Between 0.6A and 0.8A, it is happened between 50us and 100us.Third current impulse upward, maximum amplitude 0.8A and 1A it
Between, it is happened between 50us and 100us, but the time is after second pulse.4th current impulse downward, maximum amplitude
Between 0.6A and 0.8A, it is happened between 100us and 150us.Upward, maximum amplitude is in 0.8A and 1A for 5th current impulse
Between, it is happened between 200us and 250us.Downward, maximum amplitude occurs between 0.6A and 0.8A for 6th current impulse
Between 400us and 450us.Described in table 1 using error amount (subtracting measured value with simulation value) obtain as a result, can be with
It is clearly seen at each different temperature, it is minimum using M-EKF method errors value.Illustrate that EKF can have with significantly smaller error
There is faster convergence rate.Than output-error method, discrete Extended Kalman filter method and continuous Extended Kalman filter method
It more can Fast Identification parameter.In the present invention, the variation of electric current i can reflect insulation status, and the present invention is building experiment system
The value of electric current i can be gone out after system with Fast Identification by M-EKF.
Table 1
Average error value under different temperatures
The basic principles, main features and advantages of the present invention have been shown and described above.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and the above embodiments and description only describe this
The principle of invention, various changes and improvements may be made to the invention without departing from the spirit and scope of the present invention, these changes
Change and improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appended claims and its
Equivalent defines.
Claims (1)
1. a kind of ocean current generator insulation system fault detection method based on M-EKF algorithms, ocean current generator insulation system packet
Include coupling box, impulse generator, high-frequency signal collector and Roebel coil insulation system, impulse generator by coupling box with
Roebel coil insulation system connects to form closed circuit, and high-frequency signal collector is connected to the resistance R of coupling boxm, for adopting
Colleeting comb signal；Roebel coil insulation system minimizes technology by output error and its relative sensitivity function model recognizes
Method obtains a five parameter R1L1C1/R2C2Circuit topology, each arrangement of elements of circuit topology is as follows, R1, L1, C1Be it is concatenated,
And R1, L1, C1It is arranged in order from top to bottom, R2And C2Be it is concatenated be arranged in order from top to bottom, and and R1L1C1It is in parallel；R1L1C1
And R2C2After parallel connection, the resistance R of series coupled case in back upper place in parallelm, the inductance l of lower section series coupled casec, Rm=47 Ω are to measure
Resistance, lc=0.8 × 10-6H；Roebel coil insulation system then is connected to form closed circuit with impulse generator；It five
A parameter value is respectively：R1=908.9388 Ω, L1=0.019859H, C1=2.0871e-9F, R2=20.484 Ω and C2=
5.6612e-10F；Enable Jacobian matrix F=F0+M1×F1+M2×F2+M3×F3, wherein F1It is state transition matrix f (x, t) inclined
Differential x1Hessian matrix, x1It is the first component of the rendezvous value x of state variable, M1It is F1The preceding coefficient adjustment factor；F2It is shape
State transition matrix f (x, t) partial differential x2Hessian matrix, x2It is the second component of the rendezvous value x of state variable, M2It is F2Preceding
The coefficient adjustment factor；F3It is state transition matrix f (x, t) partial differential x3Hessian matrix, x3It is the rendezvous value x of state variable
Third component, M3It is F3The preceding coefficient adjustment factor；Jacobian matrix F is used to calculate the rendezvous value x of state variable；Rendezvous value x's
5 elements are respectively from top to bottom：i1, i2, uc, v and R1；Wherein i1It is R1L1C1Electric current in branch, i2It is R2C2In branch
Electric current, i are R1L1C1And R2C2Total current after parallel connection, ucRepresent capacitance C1On voltage；V represents R2And C2Upper total voltage；Collection
The first component i of conjunction value x1With second component i2The sum of to be wanted monitoring current value i；
It is characterized in that, the ocean current generator insulation system fault detection method based on M-EKF algorithms includes the following steps：
Step 1：State variable is chosen from model structure
According to the R obtained from Roebel coil insulation system1L1C1/R2C2Circuit topology, target monitoring amount are total current i, total electricity
It is a point electric current i to flow i1And i2The sum of；Jacobian matrix F is 5 rank equations, five stochastic regime variable is1, i2, uc, v and R1It is selected
It selects, contains state variable i1, i2；
Step 2：State space modeling is carried out to model structure using continuous Extended Kalman filter
Five state variable i of model structure are selected from step 11, i2, uc, v and R1, it is used for the state space side of descriptive model structure
Journey is as follows：
State space equation after above-mentioned 5 equatioies extension can be written as form：
Observational equation can be indicated with following equation：
Linear approximation equation can be calculated with Jacobi below：
Observing matrix can be indicated with following equations：
Wherein U represents the input of impulse generator, and (1) formula is to the state space equation that (5) formula is descriptive model structure, in order to count
(1) formula, is written as the form of (6) formula by the Jacobian matrix of (8) formula of calculation to (5) formula；(8) it is needed when formula is linearizing non-linear equation
The Jacobian matrix to be calculated, wherein x represent i1, i2, uc, v and R1Five state variables are selected as, (6) formula seeks partial derivative to x
Obtain the Jacobian matrix of (8) formula；Observational equation (7) formula is exactly R for obtaining output valve1L1C1/R2C2Total electricity of model structure
Flow i；Observing matrix (9) formula is for calculating kalman gain K；On the one hand the kalman gain K being calculated simultaneously is used for the side of association
On the other hand the calculating of poor matrix P is used to update the rendezvous value x of state variable；
Step 3：Modeling error is corrected using M-EKF algorithms
The Jacobian matrix of (8) formula of amendment, can be re-written as following form：
It is as follows to arrange obtained Jacobian matrix equation：
The new Jacobian matrix F obtained by (11) formula will replace the Jacobian matrix of (8) formula to reenter continuous extension
Cycle is participated in Kalman filtering；The value F newly got passes through K (t)=P (t) HT(t)R-1(t) kalman gain K is obtained, is passed throughCovariance matrix P is obtained, is on the other hand passed throughUpdate the rendezvous value x of state variable；When impulse generator input signal, high frequency
The current signal of signal picker acquisition enters M-EKF algorithms for recognizing current value i as actual measured value z (t)；Electric current i
It is the curve there are six pulse, wherein three pulses are upward, downward, pulse upward is caused by raised voltage for three pulses
, pulse directed downwardly is caused by drop-out voltage；Upward, maximum amplitude is between 0.8A and 1A, hair for first current impulse
Life is before 50us；Second current impulse downward, maximum amplitude between 0.6A and 0.8A, be happened at 50us and 100us it
Between；Upward, maximum amplitude is happened between 0.8A and 1A between 50us and 100us, but the time exists for third current impulse
After second pulse；Downward, maximum amplitude is happened at 100us and 150us between 0.6A and 0.8A for 4th current impulse
Between；Upward, maximum amplitude is happened between 0.8A and 1A between 200us and 250us for 5th current impulse；6th
Downward, maximum amplitude is happened between 0.6A and 0.8A between 400us and 450us for current impulse.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610298086.XA CN105759207B (en) | 2016-05-06 | 2016-05-06 | A kind of ocean current generator insulation system fault detection method based on M-EKF algorithms |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610298086.XA CN105759207B (en) | 2016-05-06 | 2016-05-06 | A kind of ocean current generator insulation system fault detection method based on M-EKF algorithms |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105759207A CN105759207A (en) | 2016-07-13 |
| CN105759207B true CN105759207B (en) | 2018-10-02 |
Family
ID=56323547
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610298086.XA Active CN105759207B (en) | 2016-05-06 | 2016-05-06 | A kind of ocean current generator insulation system fault detection method based on M-EKF algorithms |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105759207B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI634447B (en) | 2016-12-30 | 2018-09-01 | 財團法人工業技術研究院 | Heater condiction monitoring and accessing method and applications thereof |
| CN107704660B (en) * | 2017-09-12 | 2021-01-19 | 大连理工大学 | Error compensation method for industrial robot |
| CN108919116B (en) * | 2018-05-16 | 2020-08-28 | 上海海事大学 | Ocean current generator unbalanced stator current fault diagnosis method based on MCCKAF-FFT-Softmax |
| CN109991475B (en) * | 2019-03-26 | 2021-01-08 | 安徽贵博新能科技有限公司 | KF observer-based bridge type insulation detection circuit and method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3002154B2 (en) * | 1997-05-14 | 2000-01-24 | 川崎重工業株式会社 | Automatic large traveling equipment design system |
| CN102930171A (en) * | 2012-11-15 | 2013-02-13 | 北京理工大学 | Nonlinear filtering method based on polytope differential inclusion |
| CN105044531A (en) * | 2015-08-20 | 2015-11-11 | 河海大学 | Dynamic signal parameter identification method based on EKF and FSA |
| CN102163255B (en) * | 2010-02-17 | 2015-11-18 | 通用汽车环球科技运作有限责任公司 | The health forecast of the complication system of operational failure modeling |
-
2016
- 2016-05-06 CN CN201610298086.XA patent/CN105759207B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3002154B2 (en) * | 1997-05-14 | 2000-01-24 | 川崎重工業株式会社 | Automatic large traveling equipment design system |
| CN102163255B (en) * | 2010-02-17 | 2015-11-18 | 通用汽车环球科技运作有限责任公司 | The health forecast of the complication system of operational failure modeling |
| CN102930171A (en) * | 2012-11-15 | 2013-02-13 | 北京理工大学 | Nonlinear filtering method based on polytope differential inclusion |
| CN105044531A (en) * | 2015-08-20 | 2015-11-11 | 河海大学 | Dynamic signal parameter identification method based on EKF and FSA |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105759207A (en) | 2016-07-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105759207B (en) | A kind of ocean current generator insulation system fault detection method based on M-EKF algorithms | |
| CN107679768B (en) | Situation awareness system based on real-time data of power grid and construction method thereof | |
| CN102221651B (en) | Fault on-line diagnosis and early warning method of flameproof dry-type transformer for mine | |
| CN109165504B (en) | Power system false data attack identification method based on anti-generation network | |
| CN106353623B (en) | Low-frequency oscillation of electric power system mode on-line identification method based on random response signal | |
| CN105300692B (en) | A kind of bearing failure diagnosis and Forecasting Methodology based on expanded Kalman filtration algorithm | |
| CN109713685B (en) | Online positioning method suitable for VSC access-induced subsynchronous oscillation | |
| CN103728535A (en) | Extra-high-voltage direct-current transmission line fault location method based on wavelet transformation transient state energy spectrum | |
| CN106096298B (en) | Transducer parameters on-line identification method based on observer | |
| CN110943453B (en) | Power system fault sample generation and model construction method facing transfer learning | |
| CN103440497B (en) | A kind of GIS insulation defect shelf depreciation collection of illustrative plates mode identification method | |
| CN104392056B (en) | A kind of directly driven wind-powered unit modeling method and device | |
| CN106980044B (en) | A kind of Harmonious Waves in Power Systems current estimation method adapting to wind power integration | |
| CN109753689A (en) | A kind of online identifying approach of electric system electromechanical oscillations modal characteristics parameter | |
| CN105353226A (en) | EMI noise source impedance equivalent parameter extraction method based on scattering parameter and intelligent algorithm | |
| CN107402334A (en) | The frequency domain identification method and system of power system time/supersynchronous coupled impedance model | |
| CN105160058B (en) | A kind of network boundary equivalence method calculated for electric network electromagnet transient | |
| CN105353304A (en) | Validation method of low voltage ride-through characteristic of electric model of wind turbine generator | |
| CN104316768A (en) | Negative sequence impedance parameter estimation method for locating three-phase unbalanced disturbance source | |
| CN103023418A (en) | Online parameter identification method of synchronous generator based on wide-area measurement information | |
| CN108154223A (en) | Power distribution network operating mode recording sorting technique based on network topology and long timing information | |
| CN109494726A (en) | Stability of power system online evaluation method based on DLRNN neural network | |
| CN106443253A (en) | Power transmission line parameter identification method based on PMU (phasor measurement unit) data | |
| CN115329246A (en) | Super-capacitor equivalent circuit parameter identification method suitable for sparse data | |
| CN103746372B (en) | The on-line automatic identification method of power grid oscillation mode type and association dispatcher-controlled territory |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |