A kind of lifetime estimation method and device of divide-shut brake coil
Technical field
The present invention relates to power electronics field more particularly to a kind of lifetime estimation methods and dress of divide-shut brake coil
It sets.
Background technology
Divide-shut brake coil is the core component of breaker, while being also the multiple component of failure, due to circuit breaker design, system
It makes, the reason of all various aspects such as quality of materials and operation, the failure of divide-shut brake coil happens occasionally, and breaker is in actual motion
When, under the effect of the environmental factors such as temperature, humidity, salt fog, divide-shut brake coil performance can gradually degrade, it may appear that dynamic iron core moves
Situations such as bite, wire loop poor contact, the safe and stable operation of power grid is seriously affected.It is united according to south electric network data
Meter, about 45% breaker tripping failure is due to caused by the failure of divide-shut brake coil.Currently, for the divide-shut brake of breaker
The degradation of coil there is no accurate and effective method, is easy in this way because divide-shut brake coil fails to replace and cause in time
The phenomenon that breaker tripping, causes heavy losses to power grid.
Invention content
The purpose of the embodiment of the present invention is to provide a kind of lifetime estimation method and device of divide-shut brake coil, can simulate gas
Time condition effectively obtains the life appraisal information of divide-shut brake coil.
To achieve the above object, an embodiment of the present invention provides a kind of lifetime estimation methods of divide-shut brake coil, including:
Obtain the raw current data of at least two divide-shut brake coils;
The operation of alternation salt spray test is executed to each divide-shut brake coil according to preset data;
The test current data that each divide-shut brake coil is obtained after the alternation salt spray test operates are executed;
On-board the circuit breaker by divide-shut brake coil installation, and judge whether the divide-shut brake coil can touch breaker
Close/open machine;
If so, continue to execute alternation salt spray test operation to the divide-shut brake coil until the divide-shut brake coil without
Method touches the close/open machine;
If it is not, being carried out being based on Support vector regression algorithm according to the raw current data and the test current data
Training operation, obtain training pattern.
Compared with prior art, a kind of lifetime estimation method of divide-shut brake coil disclosed by the invention passes through according to described pre-
The each divide-shut brake coil of setting parameter executes the operation of alternation salt spray test, and obtains test current data, then described in judgement
Whether divide-shut brake coil can touch the close/open machine of breaker;If can, continue to execute the friendship to the divide-shut brake coil
Become salt spray test operation;It is on the contrary then according to pre-selection obtain the divide-shut brake coil raw current data and the test current
Data carry out the training operation based on Support vector regression algorithm, and then obtain training pattern.Solves needle in the prior art
To the degradation of the divide-shut brake coil of breaker, accurate and effective method is there is no, is easy in this way because of divide-shut brake coil not
The phenomenon that capable of replacing in time and causing breaker tripping, can simulated climatic conditions, the service life for effectively obtaining divide-shut brake coil comments
Estimate information, replaces divide-shut brake coil in time.
As the improvement of said program, the training operation based on Support vector regression algorithm specifically includes:
Construct training set input matrix and regressand value corresponding with the training set input matrix;
According to the training set input matrix and the regressand value regression modeling is carried out using support vector machines.
It is described to be based on according to the raw current data and the test current data as the improvement of said program
The training of Support vector regression algorithm operates, and after obtaining training pattern, further includes:
It is right according to the training pattern using the historical current data of the divide-shut brake coil of the breaker as input data
The divide-shut brake coil carries out life appraisal.
As the improvement of said program, the preset data include default spray time, default spray amount, preset temperature,
Default humidity and default test period.
As the improvement of said program, the alternation salt spray test operation includes four test periods, each experiment
Period includes two hours spray cycles and seven hours storage periods.
The embodiment of the present invention additionally provides a kind of life appraisal device of divide-shut brake coil, including:
Chamber, for executing the operation of alternation salt spray test to divide-shut brake coil;
Upper computer module, for setting preset data, while for according to raw current data and test current data into
Training operation of the row based on Support vector regression algorithm, obtains training pattern;
Governor circuit module, the control parameter for adjusting the regulation and control module according to the preset data;
Data acquisition module, for acquiring raw current data and test current data;
Regulate and control module, for carrying out parameter regulation and control to the chamber according to the control parameter.
Compared with prior art, a kind of life appraisal device of divide-shut brake coil disclosed by the invention passes through master control electricity
Road module adjusts the control parameter of the regulation and control module according to the parameter preset that the upper computer module is set, so that the regulation and control
Module carries out parameter regulation and control according to the control parameter to the chamber, so that divide-shut brake coil is held in the chamber
Row alternation salt spray test operates, data collecting module collected raw current data and test current data, so that upper computer module
The training operation based on Support vector regression algorithm is carried out according to raw current data and test current data, obtains training mould
Type.The degradation for solving the divide-shut brake coil for being directed to breaker in the prior art, there is no accurate and effective method, in this way
Be easy because divide-shut brake coil fails to replace in time and the phenomenon that cause breaker tripping, can simulated climatic conditions, it is effective
To the life appraisal information of divide-shut brake coil, divide-shut brake coil is replaced in time.
As the improvement of said program, the training operation based on Support vector regression algorithm specifically includes:
Construct training set input matrix and regressand value corresponding with the training set input matrix;
According to the training set input matrix and the regressand value regression modeling is carried out using support vector machines.
As the improvement of said program, the upper computer module is additionally operable to:
It is right according to the training pattern using the historical current data of the divide-shut brake coil of the breaker as input data
The divide-shut brake coil carries out life appraisal.
As the improvement of said program, the preset data include default spray time, default spray amount, preset temperature,
Default humidity and default test period.
As the improvement of said program, the alternation salt spray test operation includes four test periods, each experiment
Period includes two hours spray cycles and seven hours storage periods.
Description of the drawings
Fig. 1 is a kind of flow chart of the lifetime estimation method of divide-shut brake coil provided in an embodiment of the present invention;
Fig. 2 be a kind of divide-shut brake coil provided in an embodiment of the present invention lifetime estimation method in alternation salt spray test operate
Flow chart;
Fig. 3 be a kind of divide-shut brake coil provided in an embodiment of the present invention lifetime estimation method in structure training pattern and make
With the flow chart of training pattern;
Fig. 4 is a kind of structure diagram of the life appraisal device of divide-shut brake coil provided in an embodiment of the present invention.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art are obtained every other without creative efforts
Embodiment shall fall within the protection scope of the present invention.
Embodiment one
It is a kind of flow chart of the lifetime estimation method of divide-shut brake coil provided in an embodiment of the present invention referring to Fig. 1, Fig. 1;
Including:
S1, the raw current data for obtaining at least two divide-shut brake coils;
S2, the operation of alternation salt spray test is executed to each divide-shut brake coil according to preset data;
S3, the test current data that each divide-shut brake coil is obtained after the alternation salt spray test operates have been executed;
S4, the divide-shut brake coil is installed on-board the circuit breaker, and judges whether the divide-shut brake coil can touch open circuit
The close/open machine of device;
S5, if so, continuing to execute alternation salt spray test operation until the division brake cable to the divide-shut brake coil
Circle can not touch the close/open machine;
S6, if it is not, according to the raw current data and the test current data carry out be based on Support vector regression
The training operation of algorithm, obtains training pattern.
Specifically, in step sl, before carrying out alternation salt spray test operation to the divide-shut brake coil, first obtaining institute
State the raw current data of divide-shut brake coil, it is preferred that the divide-shut brake coil includes at least two, each division brake cable
Circle has corresponding number.In embodiments of the present invention, it includes 10 that most preferred scheme, which is the divide-shut brake coil,.
Specifically, in step s 2, can be executed to each divide-shut brake coil according to preset data in chamber
Alternation salt spray test operates, it is preferred that the chamber is using the alternation salt spray test grade of GB/T2423.18 to each described
Divide-shut brake coil executes the operation of alternation salt spray test, and specific flow can refer to Fig. 2.Preferably, the preset data includes pre-
If spray time, default spray amount, preset temperature, default humidity and default test period.Preferably, the alternation salt spray test
Operation includes four test periods, and each test period includes two hours spray cycles and seven hours storage periods.
Preferably, spray for 5% sodium chloride solution, the preset temperature be 40 DEG C, the default humidity be 95%, when described default
Between i.e. four test periods of alternation salt spray test operation total duration.
Specifically, in step s3, each divide-shut brake coil is obtained after having executed the alternation salt spray test operation
Test current data, and record experimental period and corresponding coil number in detail.
Specifically, in step s 4, the divide-shut brake coil after the alternation salt spray test operates will have been executed and installed
On-board the circuit breaker, and judge whether the divide-shut brake coil can touch the close/open machine of breaker;If the divide-shut brake coil
The close/open machine of breaker can also be touched, then shows that the divide-shut brake coil can also work normally;If the divide-shut brake coil
The close/open machine of breaker cannot be touched, then shows that the performance of the divide-shut brake coil has deteriorated, can not work normally again.
Specifically, in step s 5, if the divide-shut brake coil can also touch the close/open machine of breaker, continue pair
The divide-shut brake coil executes the alternation salt spray test operation until the divide-shut brake coil can not touch the division gate
Structure.
Specifically, in step s 6, if the divide-shut brake coil cannot touch the close/open machine of breaker, according to institute
It states raw current data and the test current data carries out the training operation based on Support vector regression algorithm, trained
Model.Preferably, step S6 is specifically included:
S61, construction training set input matrix and regressand value corresponding with the training set input matrix;
S62, regression modeling is carried out using support vector machines according to the training set input matrix and the regressand value.
Specifically, in step S61, experiment of the divide-shut brake coil after alternation salt spray test operation is arranged
Current data, it is preferred that the test current data can be the current curve of the divide-shut brake coil, calculate the divide-shut brake
The current average of coil divide-shut brake coil after the operation of n times alternation salt spray test, obtains the current average array,
N is integer, and N is greater than or equal to 0;Wherein, the current average array is:
IN=[xN1, xN2 ..., xNn];
Wherein, iN indicates to carry out the current average array that alternation salt spray test operates the divide-shut brake coil after n times;n
Indicate the number of the testing site of the divide-shut brake coil;XNn indicates each divide-shut brake in the operation of n-th alternation salt spray test
The current average of n-th of testing site on coil.When N be equal to 0 when, expression do not carry out alternation salt spray test operation when described in point
The current curve average value of closing coil.
For example, a current curve, it is made of many testing sites, if the sample rate of current curve i is
15kHZ, sampling time 100ms, then a curve just has 15kHZ*0.1s=1500 testing site, i.e. n=1500.This
When, each divide-shut brake coil will adopt the current curve of 1500 testing site.If the divide-shut brake coil has 10
It is a, then corresponding to each testing site, current average is sought, the current average number of 1500 testing sites is finally obtained
Group.The current average array is the result after each alternation salt spray test operation.
Specifically, if N is greater than or equal to 5, i0~i4 composing training collection is taken, then i5~iN is as test set.The instruction
Practicing collection can be used for modeling, to obtain the training pattern, the test set be used for the accuracy to the training pattern into
Row test could be with the training pattern to the division brake cable on the basis of it is accurate and reliable to ensure the training pattern
Circle carries out life appraisal.If by test, accuracy is very low, that just needs to model again.
Specifically, current average array i0~i4 is connected into current curve one-dimension array, then, and the current curve one
Dimension group is expressed as:I=[i0, i1, i2, i3, i4], for convenience state, again give i number, enable i=[x1, x2 ...,
X5n], window function T [m]={ tk, tk+1 ..., tk+m-1 }, k=1,2 ..., 5n-m;M=3n, t are the time.Use window function T
[m] slides value on i, and each intercepted length is the data of m, regard T [m]={ tk, tk+1 ..., tk+m-1 } as input value,
With the regressand value that season tk+m is the sequence.Construct the training set input matrix X, and with the training set input matrix pair
The regressand value Y answered.
Y=[y1, y2 ..., yq]T=[x3n+1, x3n+2 ..., x5n]T;
Wherein, the training set input matrix X is the matrix of p*q, P=3n, q=2n.
Specifically, in step S62, the training set input matrix X is expressed as to the form of row vector, at this point, X=
[X1, X2 ..., Xq]T;Wherein, Xq indicates x2n ..., x5n-1;The regressand value is expressed as Y=[Y1, Y2 ..., Yq simultaneously
]T;Wherein, Y1=x3n+1, Yq=x5n.
Specifically, regression modeling is carried out using support vector machines according to the training set input matrix and the regressand value,
The regression equation (i.e. training pattern) that electric current changes over time is obtained, the expression formula of the regression equation is as follows:
Y=xw+b;
Wherein;W=[w1, w2 ..., wp]T, b=[b1, b2 ..., bq]T.Specifically, when obtaining the regression equation,
Input x=i2 (current average of the divide-shut brake coil after the 2nd alternation salt spray test operation), you can prediction division brake cable
The situation of change in loop current waveform future, i.e. predicted value y, at this point it is possible to the predicted value is compared with the test set,
To can determine the reliability of the regression equation.
Specifically, the process for solving w, b (is converted into minimize one with linear inequality constraint two with reference to following formula
Secondary function convex programming problem):
yi-((xi·w)+b)≤ε+ξiFormula (3);
Wherein,C is penalty factor,For pine
Relaxation variable, min R (w) refer to solved function R (w) minimum values, s.t. ((xiW)+b) it refers in (xiW) under conditions of+b
Solve the minimum value of R (w).
The penalty factor is a normal number, is the folding between the complexity of function regression model and sample fitting precision
Inner feelings, value is bigger, and fitting degree is higher;ε is to return the worst error allowed, controls the number and generalization ability of supporting vector,
Value is bigger, and supporting vector is fewer.Using the principle of duality, while Lagrange multiplier a and kernel function K are introduced, by formula (1)~public affairs
Formula (3) is converted into following formula:
Wherein,aiWithIt indicates to have inside a formula and add up twice, j is control
Cumulative process number processed, kernel function K select Radial basis kernel function K (xi,xj)=exp (- γ | | xi-xj| |) ^2), wherein γ is
For the parameter that Radial basis kernel function carries, impliedly determine that the distribution that data are mapped to after new feature space, γ are got over
Greatly, supporting vector is fewer, and γ values are smaller, and supporting vector is more.The number of supporting vector influences the speed of training and prediction.Formula
(4) a is to apply Solve problems when lagrange's method of multipliers solves in, and the Lagrange multiplier of introducing solves the pass of the formula
Key is solution Lagrange multiplier a, and many softwares of process for solving a can be achieved, inside libsvm, python
The libraries sklearn, it can thus be concluded that going out to find out w.Lagrange's method of multipliers is to solve for a kind of common methods of optimization problem.For training
For collection, y, x are it is known that w has been found out.Due to y=xw+b, b can be solved by substituting into this formula.
Further, further include step S7 after step S6 obtains the training pattern:
S7, according to the training pattern using the historical current data of the divide-shut brake coil of the breaker as input number
According to divide-shut brake coil progress life appraisal.Specifically, referring to Fig. 3, the division brake cable of the breaker in being used is taken
The historical current waveform of circle is as input value X, the situation of change in prediction divide-shut brake coil current waveform future.If current waveform
There is the trend for tapering into or becoming larger, then illustrates that divide-shut brake coil performance is gradually degrading, need replacing division
Brake cable circle.
Compared with prior art, a kind of lifetime estimation method of divide-shut brake coil disclosed by the invention passes through according to described pre-
The each divide-shut brake coil of setting parameter executes the operation of alternation salt spray test, and obtains test current data, then described in judgement
Whether divide-shut brake coil can touch the close/open machine of breaker;If can, continue to execute the friendship to the divide-shut brake coil
Become salt spray test operation;It is on the contrary then according to pre-selection obtain the divide-shut brake coil raw current data and the test current
Data carry out the training operation based on Support vector regression algorithm, and then obtain training pattern.Solves needle in the prior art
To the degradation of the divide-shut brake coil of breaker, accurate and effective method is there is no, is easy in this way because of divide-shut brake coil not
The phenomenon that capable of replacing in time and causing breaker tripping, can simulated climatic conditions, the service life for effectively obtaining divide-shut brake coil comments
Estimate information, replaces divide-shut brake coil in time.
Embodiment two
It is a kind of structural frames of the life appraisal device of divide-shut brake coil provided in an embodiment of the present invention referring to Fig. 4, Fig. 4
Figure;Including:
Chamber 1, for executing the operation of alternation salt spray test to divide-shut brake coil;
Upper computer module 2, for setting preset data, while for according to raw current data and test current data into
Training operation of the row based on Support vector regression algorithm, obtains training pattern;
Governor circuit module 3, the control parameter for adjusting the regulation and control module 5 according to the preset data;
Data acquisition module 4, for acquiring raw current data and test current data;
Regulate and control module 5, for carrying out parameter regulation and control to the chamber 1 according to the control parameter.
Specifically, the data acquisition module 4 includes temperature sensor 41, humidity sensor 42 and Hall current sensor
43.Before carrying out alternation salt spray test operation to the divide-shut brake coil, first obtained by the Hall current sensor 43
The raw current data of the divide-shut brake coil, it is preferred that the divide-shut brake coil includes at least two, each divide-shut brake
Coil has corresponding number.In embodiments of the present invention, it includes 10 that most preferred scheme, which is the divide-shut brake coil,.
Specifically, alternation salt can be executed to each divide-shut brake coil according to preset data in the chamber 1
Mist test operation, it is preferred that the chamber 1 is using the alternation salt spray test grade of GB/T2423.18 to each division
Brake cable circle executes the operation of alternation salt spray test, and specific flow can refer to Fig. 2.
Specifically, the upper computer module 2 mainly by computer, serial communication program and control program form, it is described on
Machine module 2 function in position is the startup and stopping for the life appraisal device for controlling entire divide-shut brake coil, and the parameter preset is arranged,
The data for receiving the data acquisition module 4 simultaneously, to be carried out based on branch according to raw current data and test current data
The training operation for holding vector machine regression algorithm, obtains training pattern.Preferably, the preset data include default spray time,
Default spray amount, preset temperature, default humidity and default test period.Preferably, the alternation salt spray test operation includes four
A test period, each test period include two hours spray cycles and seven hours storage periods.Preferably, it sprays
For 5% sodium chloride solution, the preset temperature is 40 DEG C, and the default humidity is 95%, the preset time, that is, friendship
Become the total duration of four test periods of salt spray test operation.
Specifically, the governor circuit module 3 includes signal conditioning circuit 31, microcontroller 32,33 and of serial communication circuit
Driving circuit 34;Wherein;The microcontroller 32 is separately connected the signal conditioning circuit 31, the serial communication circuit 33 and institute
Driving circuit 34 is stated, the signal conditioning circuit 31 connects the data acquisition module 4, and the serial communication circuit 33 connects institute
Upper computer module 2 is stated, the microcontroller 32 plays control action, and the driving circuit 34 is used to drive the regulation and control module 5
Work.The governor circuit module 3 receives the preset data of the upper computer module 1 by the serial communication circuit 33, and
The control parameter of the regulation and control module 5 is adjusted according to the preset data, while the governor circuit module 3 passes through the signal
Modulate circuit 31 receives the signal of the acquisition of the data acquisition module 4, and passes upper computer module 1 back and analyzed.
Preferably, the chamber 1 is connect with the regulation and control module 5;The regulation and control module 5 includes spraying device 51, temperature
Spend regulating device 52 and humidity control apparatus 53, wherein the spraying device 51 includes automatic nozzle and beaker, the temperature tune
Regulating device 52 includes resistance heater and compressor, and the humidity control apparatus 53 includes humidifier and air drying agent.With
Parameter regulation and control are carried out to the chamber 1 in the control parameter sent according to the governor circuit module 3.So that the chamber
Spray time in 1 meets the default spray time, spray amount meets the default spray amount, and temperature meets the default temperature
Degree, humidity meets the default humidity, test period meets the default test period.
Specifically, after having executed the alternation salt spray test operation, obtained by the Hall current sensor 43 each
The test current data of the divide-shut brake coil, and experimental period and corresponding coil number are recorded in detail.
Specifically, by the divide-shut brake coil installation after the alternation salt spray test operates has been executed on-board the circuit breaker,
And judge whether the divide-shut brake coil can touch the close/open machine of breaker;If the divide-shut brake coil can also touch open circuit
The close/open machine of device then shows that the divide-shut brake coil can also work normally;If the divide-shut brake coil cannot touch open circuit
The close/open machine of device then shows that the performance of the divide-shut brake coil has deteriorated, can not work normally again.
If specifically, the divide-shut brake coil can also touch the close/open machine of breaker, continue to the divide-shut brake
Coil executes the alternation salt spray test operation until the divide-shut brake coil can not touch the close/open machine.
If specifically, the divide-shut brake coil cannot touch the close/open machine of breaker, the upper computer module 1 is also
For constructing training set input matrix and regressand value corresponding with the training set input matrix;Square is inputted according to the training set
Battle array and the regressand value carry out regression modeling using support vector machines.
Specifically, test current data of the divide-shut brake coil after alternation salt spray test operation are arranged, it is excellent
Choosing, the test current data can be the current curve of the divide-shut brake coil, calculate the divide-shut brake coil and handed in n times
The current average for becoming the divide-shut brake coil after salt spray test operation, obtains the current average array, and N is integer, and N
More than or equal to 0;Wherein, the current average array is:
IN=[xN1, xN2 ..., xNn];
Wherein, iN indicates to carry out the current average array that alternation salt spray test operates the divide-shut brake coil after n times;n
Indicate the number of the testing site of the divide-shut brake coil;XNn indicates each divide-shut brake in the operation of n-th alternation salt spray test
The current average of n-th of testing site on coil.When N be equal to 0 when, expression do not carry out alternation salt spray test operation when described in point
The current curve average value of closing coil.
For example, a current curve, it is made of many testing sites, if the sample rate of current curve i is
15kHZ, sampling time 100ms, then a curve just has 15kHZ*0.1s=1500 testing site, i.e. n=1500.This
When, each divide-shut brake coil will adopt the current curve of 1500 testing site.If the divide-shut brake coil has 10
It is a, then corresponding to each testing site, current average is sought, the current average number of 1500 testing sites is finally obtained
Group.The current average array is the result after each alternation salt spray test operation.
Specifically, if N is greater than or equal to 5, i0~i4 composing training collection is taken, then i5~iN is as test set.The instruction
Practicing collection can be used for modeling, to obtain the training pattern, the test set be used for the accuracy to the training pattern into
Row test could be with the training pattern to the division brake cable on the basis of it is accurate and reliable to ensure the training pattern
Circle carries out life appraisal.If by test, accuracy is very low, that just needs to model again.
Specifically, current average array i0~i4 is connected into current curve one-dimension array, then, and the current curve one
Dimension group is expressed as:I=[i0, i1, i2, i3, i4], for convenience state, again give i number, enable i=[x1, x2 ...,
X5n], window function T [m]={ tk, tk+1 ..., tk+m-1 }, k=1,2 ..., 5n-m;M=3n, t are the time.Use window function T
[m] slides value on i, and each intercepted length is the data of m, regard T [m]={ tk, tk+1 ..., tk+m-1 } as input value,
With the regressand value that season tk+m is the sequence.Construct the training set input matrix X, and with the training set input matrix pair
The regressand value Y answered.
Y=[y1, y2 ..., yq]T=[x3n+1, x3n+2 ..., x5n]T;
Wherein, the training set input matrix X is the matrix of p*q, P=3n, q=2n.
Specifically, the training set input matrix X to be expressed as to the form of row vector, at this point, X=[X1, X2 ..., Xq
]T;Wherein, Xq indicates x2n ..., x5n-1;The regressand value is expressed as Y=[Y1, Y2 ..., Yq] simultaneouslyT;Wherein, Y1=
X3n+1, Yq=x5n.
Specifically, regression modeling is carried out using support vector machines according to the training set input matrix and the regressand value,
The regression equation (i.e. training pattern) that electric current changes over time is obtained, the expression formula of the regression equation is as follows:
Y=xw+b;
Wherein;W=[w1, w2 ..., wp]T, b=[b1, b2 ..., bq]T.Specifically, when obtaining the regression equation,
Input x=i2 (current average of the divide-shut brake coil after the 2nd alternation salt spray test operation), you can prediction division brake cable
The situation of change in loop current waveform future, i.e. predicted value y, at this point it is possible to the predicted value is compared with the test set,
To can determine the reliability of the regression equation.
Specifically, the process for solving w, b (is converted into minimize one with linear inequality constraint two with reference to following formula
Secondary function convex programming problem):
yi-((xi·w)+b)≤ε+ξiFormula (3);
Wherein,C is penalty factor,For pine
Relaxation variable, min R (w) refer to solved function R (w) minimum values, s.t. ((xiW)+b) it refers in (xiW) under conditions of+b
Solve the minimum value of R (w).
The penalty factor is a normal number, is the folding between the complexity of function regression model and sample fitting precision
Inner feelings, value is bigger, and fitting degree is higher;ε is to return the worst error allowed, controls the number and generalization ability of supporting vector,
Value is bigger, and supporting vector is fewer.Using the principle of duality, while Lagrange multiplier a and kernel function K are introduced, by formula (1)~public affairs
Formula (3) is converted into following formula:
Wherein,aiWithIt indicates to have inside a formula and add up twice, j is control
Cumulative process number processed, kernel function K select Radial basis kernel function K (xi,xj)=exp (- γ | | xi-xj| |) ^2), wherein γ is
For the parameter that Radial basis kernel function carries, impliedly determine that the distribution that data are mapped to after new feature space, γ are got over
Greatly, supporting vector is fewer, and γ values are smaller, and supporting vector is more.The number of supporting vector influences the speed of training and prediction.Formula
(4) a is to apply Solve problems when lagrange's method of multipliers solves in, and the Lagrange multiplier of introducing solves the pass of the formula
Key is solution Lagrange multiplier a, and many softwares of process for solving a can be achieved, inside libsvm, python
The libraries sklearn, it can thus be concluded that going out to find out w.Lagrange's method of multipliers is to solve for a kind of common methods of optimization problem.For training
For collection, y, x are it is known that w has been found out.Due to y=xw+b, b can be solved by substituting into this formula.
Further, after obtaining the training pattern, the upper computer module 1 is additionally operable to will according to the training pattern
The historical current data of the divide-shut brake coil of the breaker comment the divide-shut brake coil progress service life as input data
Estimate.
Specifically, according to the training pattern using the historical current data of the divide-shut brake coil of the breaker as input
Data carry out life appraisal to the divide-shut brake coil.Specifically, referring to Fig. 3, the divide-shut brake of the breaker in being used is taken
The historical current waveform of coil is as input value X, the situation of change in prediction divide-shut brake coil current waveform future.If current wave
There is the trend for tapering into or becoming larger in shape, then illustrates that divide-shut brake coil performance is gradually degrading, and needs replacing point
Closing coil.
Compared with prior art, a kind of life appraisal device of divide-shut brake coil disclosed by the invention passes through master control electricity
Road module 3 adjusts the control parameter of the regulation and control module 5 according to the parameter preset that the upper computer module 2 is set, so that described
Regulate and control module 5 and parameter regulation and control are carried out to the chamber 1 according to the control parameter, so that the chamber 1 is to division
Brake cable circle executes the operation of alternation salt spray test, and data acquisition module 4 acquires raw current data and test current data, so that on
Position machine module 2 carries out the training based on Support vector regression algorithm according to raw current data and test current data and operates,
Obtain training pattern.Solves the degradation for the divide-shut brake coil for being directed to breaker in the prior art, it is accurate and effective to there is no
Method, be easy in this way because divide-shut brake coil fails to replace in time and the phenomenon that cause breaker tripping, being capable of simulation climate
Condition effectively obtains the life appraisal information of divide-shut brake coil, replaces divide-shut brake coil in time.
The above is the preferred embodiment of the present invention, it is noted that for those skilled in the art
For, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also considered as
Protection scope of the present invention.