CN106707153A - FOA-RBF based high-voltage circuit breaker fault diagnosis method - Google Patents
FOA-RBF based high-voltage circuit breaker fault diagnosis method Download PDFInfo
- Publication number
- CN106707153A CN106707153A CN201611227922.1A CN201611227922A CN106707153A CN 106707153 A CN106707153 A CN 106707153A CN 201611227922 A CN201611227922 A CN 201611227922A CN 106707153 A CN106707153 A CN 106707153A
- Authority
- CN
- China
- Prior art keywords
- rbf
- foa
- fault diagnosis
- iron core
- represent
- 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
- 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
- G01R31/327—Testing of circuit interrupters, switches or circuit-breakers
- G01R31/3271—Testing of circuit interrupters, switches or circuit-breakers of high voltage or medium voltage devices
- G01R31/3275—Fault detection or status indication
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Testing Electric Properties And Detecting Electric Faults (AREA)
Abstract
The invention discloses an FOA-RBF based high-voltage circuit breaker fault diagnosis method, which comprises the steps of according to the waveform, which varies along with the time, of closing current of a high-voltage circuit breaker, extracting feature variables, and selecting 24 groups of typical data to act as a feature variable sample; performing pretreatment and normalization on the feature variable sample, carrying out binary coding on six groups of diagnosis types, performing iterative optimization on a spreading parameter spread of an RBF network in the global scope, reserving an optimal spread value, and building an FOA-RBF based fault diagnosis model; selecting sample feature variables to act as input of an FOA-RBF network, enabling the six groups of diagnosis types to act as output of the FOA-RBF neural network, and performing high-voltage circuit breaker fault diagnosis. According to the FOA-RBF based high-voltage circuit breaker fault diagnosis method, an artificial intelligence method of FOA-RBF is applied to fault diagnosis for the high-voltage circuit breaker, and quick and effective detection can be performed on a fault.
Description
Technical field
Broken the present invention relates to a kind of Fault Diagnosis for HV Circuit Breakers technology, more particularly to a kind of height based on FOA-RBF
Road device method for diagnosing faults.
Background technology
Primary cut-out can be switched on or switched off the switching device of high-tension circuit in the case of normal or failure, in power distribution network system
Play dual parts of to control in system and protect, the quality of its operation conditions directly decides that can whole power system normally transport
OK.Therefore, to carry out fault diagnosis to primary cut-out significant.
By inspection and repair system of regularly stopping transport, this method blindness is big, and specific aim is not strong, takes for the maintenance of past breaker
When it is laborious, and cause large-area power-cuts.The change of primary cut-out fault type is various, and phenomenon of the failure exists complicated with failure cause
Various non-linear relation, traditional method for diagnosing faults effect is not obvious.With flying for artificial intelligence and computer technology
Speed development, generates the intelligent diagnosing method of a collection of primary cut-out, such as the height based on error back propagation (BP) neutral net
Voltage breaker method for diagnosing faults, the Fault Diagnosis for HV Circuit Breakers method based on radial direction base (RBF) neutral net etc..But these
Method has that rate of convergence is slow, is easily trapped into local extreme points, spreading parameter is chosen.
The characteristic curve that the closing coil current waveform of primary cut-out is changed over time has reacted closing iron core and it is controlled
The running status of the associated mechanisms such as the chain contact of system, contains the mechanical characteristic of abundant circuit breaker operation mechanism.Carry
Take in closing coil current waveform for information about, can be inferred that the working condition of primary cut-out, such as operating voltage voltage, iron
Core idle stroke, operating mechanism operating state etc., therefore, closing coil current waveform profile is the one of Fault Diagnosis for HV Circuit Breakers
Individual important point of penetration.
Neutral net is, by a large amount of letter single-neuron modelling human brain behaviors, its input, outlet chamber to be realized by connection extensively
The complicated network system of Nonlinear Mapping.Neutral net can be to the function reality such as the abstract thinking of physiology human brain, associative memory
Simulation is applied abstract, so as to realize the information processing functions such as the study similar to human brain, supposition, identification, memory, it is adaptable to breaker
Fault diagnosis.
For primary cut-out fault characteristic there is non-linear and closing coil electric current to change over time characteristic curve, such as
What rapidly and accurately carries out intelligent diagnostics, the key as problem.
Prior art one:
Fault Diagnosis for HV Circuit Breakers is carried out using regular stoppage in transit inspection and repair system, is mainly combined using overhaul and light maintenance
Preventative maintenance is carried out, light maintenance is examination and repair for high-pressure breaker insulating properties, operating mechanism running-active status etc., in periodic replacement part
When carry out large repairs.
The shortcoming of prior art one:
The failure symptom of exception can not be in time found and investigate, maintenance has blindness, wastes time and energy, and brings unnecessary
Large-area power-cuts, while regular artificial inspection operation may increase new failure, reduces breaker service life.
Prior art two:
Fault Diagnosis for HV Circuit Breakers is carried out using BP neural network, by the characteristic quantity of primary cut-out switching current waveform
The input signal of BP neural network is converted into, BP neural network is by information forward-propagating and the two-way mistake of error back propagation
The weights and deviation of network are carried out adjusting training repeatedly by journey, make the vector of output and Mean Vector close to building
Mapping between the failure symptom of vertical primary cut-out system input and the failure cause of output.
The shortcoming of prior art two:
Have that rate of convergence is slow, the defect such as poor that is easily absorbed in local extreme points, inferential capability, convergence precision is high, not even
Convergence, so as to reduce the accuracy rate of Fault Diagnosis for HV Circuit Breakers.
Prior art three:
Using the Fault Diagnosis for HV Circuit Breakers method of RBF neural, closing coil Current Waveform Characteristics amount is extracted, built
The Fault Diagnosis for HV Circuit Breakers model of the RBF neural that is based on.RBF neural is a kind of feedforward counterpropagation network,
Using higher dimensional space interpolation method, with respect to BP neural network, faster, the degree of accuracy is higher for rate of convergence.
The shortcoming of prior art three:
Spreading parameter spread is RBF neural performance important factor in order, and spread is bigger, and Function Fitting is more flat
It is sliding, but approximate error can become big;Spread is smaller, and approaching for function can be more accurate, but approximate procedure can be unsmooth, network
Poor performance, it may appear that cross adaptation.In the existing circuit breaker failure diagnostic method based on RBF neural, generally select respectively
Different spread preset values are taken to be attempted, it is so cumbersome and time consuming, do not reach preferable effect also sometimes.
The content of the invention
It is an object of the invention to provide a kind of Fault Diagnosis for HV Circuit Breakers method based on FOA-RBF.
The purpose of the present invention is achieved through the following technical solutions:
Fault Diagnosis for HV Circuit Breakers method based on FOA-RBF of the invention, comprises the following steps:
A, waveform is changed over time according to primary cut-out switching current, extract characteristic quantity I1、I2、I3、t1、t2、t3、t4、
t5, 24 groups of typical datas are chosen as characteristic quantity sample;
B, characteristic quantity sample is pre-processed, wherein characteristic quantity I1、I2、I3、t1、t2、t3、t4、t5It is normalized place
Reason, diagnostic-type Y1、Y2、Y3、Y4、Y5、Y6Carry out binary coding;
C, using FOA algorithm search abilities are strong, the advantage that low optimization accuracy is high, to the extension of RBF networks in global scope
Parameter spread is iterated optimizing, retain optimal spread values, it is to avoid be manually set preset value tries to gather repeatedly, and foundation is based on
The fault diagnosis model of FOA-RBF;
D, selection characteristic quality of sample I1、I2、I3、t1、t2、t3、t4、t5As the input quantity of FOA-RBF networks, Y1、Y2、Y3、
Y4、Y5、Y6As the output quantity of FOA-RBF neutral nets, training sample of 18 groups of characteristic quantity samples as FOA-RBF networks is chosen
This, is left 6 groups as test samples, carries out Fault Diagnosis for HV Circuit Breakers;
In above-mentioned steps, each parameter is represented:
In t0-t1Stage, coil electricity, electric current rises to maximum I1;
In t1-t2Stage, iron core starts action, and load increases electric current and declines, until iron core strikes buckle electric current and reaches pole
Small value I2;
In t2-t3Stage, iron core stopping action, coil current is exponentially risen to close to maximum steady state value I3;
In t3-t4Stage, continuity on last stage, electric current reaches maximum steady state value I3;
In t4-t5Stage, auxiliary switch disjunction, it is zero that electric current is dropped rapidly to;
Y1、Y2、Y3、Y4、Y5、Y6The possible fault diagnosis type of primary cut-out is represented, wherein:Network output quantity Y1Represent
Breaker is normal, Y2Represent operation power too low, Y3Represent closing iron core initial time bite, Y4Represent operating mechanism bite, Y5
Represent iron core idle stroke too big, Y6Represent auxiliary switch action contact not good.
As seen from the above technical solution provided by the invention, the height based on FOA-RBF provided in an embodiment of the present invention
Voltage breaker method for diagnosing faults, is one during the artificial intelligence approach of FOA-RBF applied into the fault diagnosis of primary cut-out
The Fault Diagnosis for HV Circuit Breakers method based on FOA-RBF is planted, quick effective detection can be carried out to failure.
Brief description of the drawings
Fig. 1 is changed over time to gather primary cut-out switching current using mechanic character tester in the embodiment of the present invention
Waveform;
Fig. 2 is the flow chart of FOA optimization RBF neurals in the embodiment of the present invention;
Fig. 3 is the Fault Diagnosis for HV Circuit Breakers structure based on FOA-RBF in the embodiment of the present invention;
Fig. 4 is the performance error curve of network training in the embodiment of the present invention;
Fig. 5 is BP network training error curves in the embodiment of the present invention.
Specific embodiment
With reference to the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Ground description, it is clear that described embodiment is only a part of embodiment of the invention, rather than whole embodiments.Based on this
Inventive embodiment, the every other implementation that those of ordinary skill in the art are obtained under the premise of creative work is not made
Example, belongs to protection scope of the present invention.
Fault Diagnosis for HV Circuit Breakers method based on FOA-RBF of the invention, its preferably specific embodiment is:
Comprise the following steps:
A, waveform is changed over time according to primary cut-out switching current, extract characteristic quantity I1、I2、I3、t1、t2、t3、t4、
t5, 24 groups of typical datas are chosen as characteristic quantity sample;
B, characteristic quantity sample is pre-processed, wherein characteristic quantity I1、I2、I3、t1、t2、t3、t4、t5It is normalized place
Reason, diagnostic-type Y1、Y2、Y3、Y4、Y5、Y6Carry out binary coding;
C, using FOA algorithm search abilities are strong, the advantage that low optimization accuracy is high, to the extension of RBF networks in global scope
Parameter spread is iterated optimizing, retain optimal spread values, it is to avoid be manually set preset value tries to gather repeatedly, and foundation is based on
The fault diagnosis model of FOA-RBF;
D, selection characteristic quality of sample I1、I2、I3、t1、t2、t3、t4、t5As the input quantity of FOA-RBF networks, Y1、Y2、Y3、
Y4、Y5、Y6As the output quantity of FOA-RBF neutral nets, training sample of 18 groups of characteristic quantity samples as FOA-RBF networks is chosen
This, is left 6 groups as test samples, carries out Fault Diagnosis for HV Circuit Breakers;
In above-mentioned steps, each parameter is represented:
In t0-t1Stage, coil electricity, electric current rises to maximum I1;
In t1-t2Stage, iron core starts action, and load increases electric current and declines, until iron core strikes buckle electric current and reaches pole
Small value I2;
In t2-t3Stage, iron core stopping action, coil current is exponentially risen to close to maximum steady state value I3;
In t3-t4Stage, continuity on last stage, electric current reaches maximum steady state value I3;
In t4-t5Stage, auxiliary switch disjunction, it is zero that electric current is dropped rapidly to;
Y1、Y2、Y3、Y4、Y5、Y6The possible fault diagnosis type of primary cut-out is represented, wherein:Network output quantity Y1Represent
Breaker is normal, Y2Represent operation power too low, Y3Represent closing iron core initial time bite, Y4Represent operating mechanism bite, Y5
Represent iron core idle stroke too big, Y6Represent auxiliary switch action contact not good.
In above-mentioned steps, the characteristic quantity I of waveform is changed over time by analyzing closing coil current waveform1、I2、I3、t1、
t2、t3、t4、t5, obtain iron core start-stop time, coil electricity time, with reference to breaker relevant parameter, and then obtain breaker fortune
Row state, including the action of operating voltage voltage, iron core idle stroke, operating mechanism situation, auxiliary switch.
Fault Diagnosis for HV Circuit Breakers method based on FOA-RBF provided in an embodiment of the present invention, primary cut-out failure
It is as follows that diagnostic method will solve technical problem:
1st, partial high pressure breaker uses regular stoppage in transit inspection and repair system, this relatively conservative scheduled overhaul, with certain
Blindness.Therefore, operation conditions, the trouble point of grasp breaker promptly and accurately, reduces excessive large-area power-cuts maintenance, increases
The specific aim of strong maintenance, can reach the purpose of the reliability and economy that improve distribution network system.
2nd, using the method for diagnosing faults based on BP neural network, the method has rate of convergence to partial high pressure breaker
Slowly, network performance is poor, be easily absorbed in local extreme points, the shortcomings of fault-tolerant ability is not strong, therefore improve rate of convergence and diagnosis effect is aobvious
Obtain particularly important.
3rd, using using the method for diagnosing faults based on RBF neural, the method needs difference to partial high pressure breaker
The preset values different to spreading parameter spread are attempted, and are wasted time and energy, thus need one kind can in global scope from
The dynamic RBF neural method for diagnosing faults for finding spread optimal values, obtains network performance optimal.
4th, it is a kind of base during the artificial intelligence approach of FOA-RBF is applied to the fault diagnosis of primary cut-out by the present invention
In the Fault Diagnosis for HV Circuit Breakers method of FOA-RBF, quick effective detection can be carried out to failure.
Specific embodiment:
1 sets up Fault Diagnosis for HV Circuit Breakers characteristic quantity sample:
1.1 characteristic quantity sample extractions:
Waveform is changed over time as shown in figure 1, in t using mechanic character tester collection primary cut-out switching current0-
t1Stage, coil electricity, electric current rises to maximum I1;In t1-t2Stage, iron core starts action, and load increases electric current and declines,
Until iron core strikes buckle electric current and reaches minimum I2;In t2-t3Stage, iron core stopping action, coil current exponentially rises
Extremely close to maximum steady state value I3;In t3-t4Stage, continuity on last stage, electric current reaches maximum steady state value I3;In t4-t5Stage,
Auxiliary switch disjunction, it is zero that electric current is dropped rapidly to.The feature of waveform is changed over time by analyzing closing coil current waveform
Amount I1、I2、I3、t1、t2、t3、t4、t5Deng, the parameters such as iron core start-stop time, coil electricity time are obtained, join with reference to breaker correlation
Number, and then the High Voltage Circuit Breaker Condition is obtained, such as operating voltage voltage, iron core idle stroke, operating mechanism situation, auxiliary switch action
Deng.
Randomly select 24 groups of typical primary cut-out switching currents and change over time wave character amount sample, such as the institute of table 1
Show.Wherein, characteristic quality of sample I is chosen1、I2、I3、t1、t2、t3、t4、t5As the input quantity of FOA-RBF networks, Y1、Y2、Y3、Y4、
Y5、Y6As the output quantity of FOA-RBF neutral nets, the possible fault diagnosis type of primary cut-out is represented.Network output quantity Y1
Represent breaker normal, Y2Represent operation power too low, Y3Represent closing iron core initial time bite, Y4Represent operating mechanism card
It is puckery, Y5Represent iron core idle stroke too big, Y6Represent auxiliary switch action contact not good.Neutral net output quantity binary coding
It is indicated:Y1(1 0 0 0 0 0);Y2(0 2 0 0 0 0);Y3(0 0 1 0 0 0);Y4(0 0 0 1 0 0);Y5(0 0
0 0 1 0);Y6(0 0 0 0 0 1)。
The primary cut-out switching current of table 1 changes over time wave character amount sample
1.2 characteristic quantity sample preprocessings
Due to sample characteristics I1、I2、I3、t1、t2、t3、t4、t5With different magnitude and unit, it is necessary to carry out pretreatment it is right
Sample characteristics are normalized, to improve network training speed.Normalized method is to be converted into the input quantity of neutral net
Numerical value between [0,1], normalization formula is as follows:
Pik *=(Pik-PK-min)/(PK-max-PK-min)
Wherein, PikRepresent that primary cut-out switching current changes over time i-th data of kth kind characteristic quantity in waveform;
PK-minRepresent the minimum value of all data in kth kind characteristic quantity;PK-maxRepresent the maximum of all data in kth kind characteristic quantity;
Pik *Represent the result data after i-th data normalization of kth kind characteristic quantity.
24 groups of characteristic quality of sample are as neutral net input quantity by normalized (after accurate decimal point three), failure
Diagnostic-type carries out binary coding as neutral net output quantity, obtains pretreated data, as shown in table 2.
The pretreated characteristic quantity sample of table 2
The 2 FOA-RBF network models for setting up Fault Diagnosis for HV Circuit Breakers:
2.1RBF neural networks principles:
RBF neural is three layers of feed forward type neutral net being made up of input layer, hidden layer, output layer.In input layer,
One primary cut-out switching current waveform sample characteristic quantity of any one node on behalf, its input vector is X=[I1, I2, I3,
t1, t2, t3, t4, t5].In hidden layer, RBF realizes input layer to the Nonlinear Mapping of output layer, and its excitation function is more
Using Gaussian function:
Wherein:C and σ represents center and the variance of Gaussian function respectively, | | xp-ci| | it is European norm.
In output layer, network is output as:
Wherein:xpP-th sample characteristics input value of primary cut-out is represented, is connection weight of the hidden layer to output layer,
wijIt is the connection weight of hidden layer to output layer, yjRepresent that the reality of j-th output node of network corresponding with input sample is defeated
Go out.
2.2RBF network performance affecting parameters spread
Call format can create a radial base neural net in the Neural Network Toolbox of Matlab softwares:
Net=newrb (P, T, goal, spread)
Wherein:P is input vector, and T is target output vector, and goal divides equally error target for network, and spread is footpath
To the spreading parameter of basic function.
Spreading parameter spread has reacted response duration of the neuron output to input, the size diameter influence of spread values
The performance of RBF neural.The value of spread is bigger, and Function Fitting curve is more smooth, but approximate error value also increases;
The value of spread is smaller, and function approximation error amount reduces, but matched curve can be unsmooth, and approximate procedure occurred adaptation.
The selection of spread, should make neuron produce the input range of response to cover sufficiently large region, and spread mistakes are avoided again
Causing each neuron greatly has the input vector response region for overlapping.In the past during design RBF neural, generally choose respectively
Different spread preset values are attempted, so cumbersome and time consuming, do not reach preferable effect also sometimes.
2.3FOA-RBF models
Value for spreading parameter spread directly affects the performance of RBF neural, using fruit bat optimized algorithm
(FOA) advantage that search capability is strong, rate of convergence is fast, finds spread optimal values, to optimize RBF internetworkings in global scope
Energy.Because the ability of looking for food of fruit bat is strong, oxyopia, using spread optimal solutions as food source, nerve is made by iteration optimizing
Network output valve reaches minimum with actual value mean square deviation.Then calculate between fruit bat body position and the origin of coordinates distance and calculate
Inverse to obtain flavor concentration decision content, i.e. spreading parameter spread, during the value of Spread substituted into RBF network training sentences,
Network output valve is obtained, using the mean square error MSE of output valve and sample object output valve as flavor concentration decision function Smell
(q):
Wherein:N is RBF training sample numbers, and m is RBF network output layer nodes, xpqIt is network reality output, ypq
For network objectives are exported.
The flow chart of FOA optimization RBF neurals as shown in Figure 2, is comprised the following steps that:
(1) it is Sizepop to give fruit bat population size, and maximum iteration is Maxgen, random initializtion fruit bat colony
Position coordinates is:(X_axis,Y_axis);
(2) assign fruit bat individual random direction and distance using smell search of food, Random is detection range, then
Xi=X_axis+Random
Yi=Y_axis+Random
(3) due to food position cannot be learnt in advance, therefore first estimate with origin apart from Disti, then to calculate taste dense
Degree decision content Si, i.e. spread function spread, this value is the inverse of distance:
Si=1/Disti
(4) value of spread function spread is brought into BBF Neural Network Toolbox training function statement
In net=newrb (P, T, goal, spread), by network training, nerve net is obtained using the emulation of sim functions
Network is exported
Value, using the mean square error MSE of output valve and sample output valve as flavor concentration decision function Smelli, that is,
Error sum of squares.
Smelli=function (si)
(5) the minimum fruit bat as optimum individual of flavor concentration in the fruit bat colony is found out so that the value of error sum of squares
Reach minimum;
[bestSmll bestindex]=min (Smelli)
(6) record and retain best flavors concentration value bestSmell with it in X, the coordinate of Y-axis, at this time fruit bat colony
Flown to the position using vision:
Smellbest=bestSmell
X_axis=X (bestindex)
Y_axis=Y (bestindex)
(7) step (2)~(5) and then iteration optimizing successively, are repeated, and judges best flavors concentration whether better than preceding
One iteration best flavors concentration, and current iteration number of times is less than greatest iteration number, if then performing step (6).
2.2.3FOA-RBF network fault diagnosis
Training sample of 18 groups of characteristic quantity samples as RBF neural is chosen in table 2, being left 6 groups of samples is used for net
The inspection of network diagnosis effect, sets up the Fault Diagnosis for HV Circuit Breakers structure based on FOA-RBF as shown in Figure 3.
In simulation calculation software Matlab R2010b running environment, program is write in global scope inner iteration optimizing, protect
The value of optimal spread function spread is stayed, the primary cut-out failure based on FOA-RBF is set up using Neural Network Toolbox and is examined
It is disconnected.Fruit bat colony initial position as [0,1] is set, it is [- 10,10], population scale that fruit bat random flight direction is interval with distance
It is 20, maximum iteration is 50.Neural Network Toolbox call format net=newrb (P, T, goal, spread), is set
It is 0.01 that network divides equally error target goal.The simulation run in Matlab softwares, the parameter that is expanded spread optimum values are
0.1623, the performance error curve of network training is as shown in Figure 4.
For the performance error of comparing cell training, BP neural network is designed.It is fixed to be existed according to multilayer neural network mapping
Reason, proves that an arbitrary continuous function can set up mapping relations with a three-layer neural network, therefore use herein in theory
Three layers of BP neural network, wherein input layer number are 8, represent that primary cut-out switching current changes over time waveform respectively
Characteristic quantity I1、I2、I3、t1、t2、t3、t4、t5;Output layer nodes are 6, and the possible fault diagnosis of primary cut-out is represented respectively
Type Y1、Y2、Y3、Y4、Y5、Y6;Choose hidden layer node number m=10.Because the degree of membership of fuzzy logic is between [0,1]
Value, so the activation primitive of each neuron is respectively tansig and purelin, training function uses traingda.Using god
It is three layers of BP networks of 8-10-6 through one structure of newff function creations in network tool case, maximum frequency of training is 2000, instruction
It is 0.01 to practice target error, and learning rate is 0.1.BP network training error curves are as shown in Figure 5.
Network training error curve Fig. 4 and Fig. 5, the network for setting at the same time respectively error target goal be 0.01 situation
Under, BP neural network needs 716 steps to can be only achieved training objective error, and the change in oscillation of network training error performance is brighter
It is aobvious, and FOA-RBF networks only need 5 steps to can reach training objective error, rate of convergence is significantly improved, and network training effect is obvious
Improve.
By this six groups of test sample input quantities of 19-24 in table 2, the above-mentioned trained FOA-RBF networks for finishing are input to
In, operation obtains the reality output of Fault Diagnosis for HV Circuit Breakers, as shown in table 3.According to maximum membership grade principle, contrast is actual
Output and target output, the reality output that can obtain trained FOA-RBF networks match with target output, can be accurate
The identification various possible fault types of primary cut-out, the Fault Diagnosis for HV Circuit Breakers method based on FOA-RBF is effectively reliable.
The reality output of the Fault Diagnosis for HV Circuit Breakers of table 3 is exported with target
What technical solution of the present invention was brought has the beneficial effect that:
1 in primary cut-out switching current changes over time curve, by the sample characteristics I of different magnitude and unit1、
I2、I3、t1、t2、t3、t4、t5Pre-processed, rate of convergence is improved by normalizing;
2 realize the simple, advantage that search capability is strong, low optimization accuracy is high using FOA optimized algorithms, are found in global scope
RBF neural spreading parameter spread optimal values, choosing different spread preset values relative to traditional RBF networks is carried out instead
The method that retrial is gathered, FOA-RBF methods avoid human intervention, program automatic optimal and the optimal Spread values of reservation, time saving and energy saving, together
When network performance optimized;;
3 based on FOA-RBF Fault Diagnosis for HV Circuit Breakers compared with the Fault Diagnosis for HV Circuit Breakers based on BP networks,
Rate of convergence is faster, inferential capability is stronger, network training error performance is more excellent;
4 methods based on FOA-RBF carry out fault diagnosis to primary cut-out, and reality output is consistent with target output, energy
Enough accurately identification various possible fault types of primary cut-out, diagnosis effect is obvious.
The above, the only present invention preferably specific embodiment, but protection scope of the present invention is not limited thereto,
Any one skilled in the art in the technical scope of present disclosure, the change or replacement that can be readily occurred in,
Should all be included within the scope of the present invention.Therefore, protection scope of the present invention should be with the protection model of claims
Enclose and be defined.
Claims (2)
1. a kind of Fault Diagnosis for HV Circuit Breakers method based on FOA-RBF, it is characterised in that comprise the following steps:
A, waveform is changed over time according to primary cut-out switching current, extract characteristic quantity I1、I2、I3、t1、t2、t3、t4、t5, choosing
24 groups of typical datas are taken as characteristic quantity sample;
B, characteristic quantity sample is pre-processed, wherein characteristic quantity I1、I2、I3、t1、t2、t3、t4、t5It is normalized, examines
Disconnected type Y1、Y2、Y3、Y4、Y5、Y6Carry out binary coding;
C, using FOA algorithm search abilities are strong, the advantage that low optimization accuracy is high, to the spreading parameter of RBF networks in global scope
Spread is iterated optimizing, retains optimal spread values, sets up the fault diagnosis model based on FOA-RBF;
D, selection characteristic quality of sample I1、I2、I3、t1、t2、t3、t4、t5As the input quantity of FOA-RBF networks, Y1、Y2、Y3、Y4、Y5、
Y6As the output quantity of FOA-RBF neutral nets, training sample of 18 groups of characteristic quantity samples as FOA-RBF networks is chosen, remained
Lower 6 groups, as test samples, carry out Fault Diagnosis for HV Circuit Breakers;
In above-mentioned steps, each parameter is represented:
In t0-t1Stage, coil electricity, electric current rises to maximum I1;
In t1-t2Stage, iron core starts action, and load increases electric current and declines, until iron core strikes buckle electric current and reaches minimum
I2;
In t2-t3Stage, iron core stopping action, coil current is exponentially risen to close to maximum steady state value I3;
In t3-t4Stage, continuity on last stage, electric current reaches maximum steady state value I3;
In t4-t5Stage, auxiliary switch disjunction, it is zero that electric current is dropped rapidly to;
Y1、Y2、Y3、Y4、Y5、Y6The possible fault diagnosis type of primary cut-out is represented, wherein:Network output quantity Y1Represent open circuit
Device is normal, Y2Represent operation power too low, Y3Represent closing iron core initial time bite, Y4Represent operating mechanism bite, Y5Represent
Iron core idle stroke is too big, Y6Represent auxiliary switch action contact not good.
2. the Fault Diagnosis for HV Circuit Breakers method based on FOA-RBF according to claim 1, it is characterised in that above-mentioned
In step, the characteristic quantity I of waveform is changed over time by analyzing closing coil current waveform1、I2、I3、t1、t2、t3、t4、t5, obtain
To iron core start-stop time, coil electricity time, with reference to breaker relevant parameter, and then the High Voltage Circuit Breaker Condition is obtained, including behaviour
Make voltage, iron core idle stroke, operating mechanism situation, auxiliary switch action.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611227922.1A CN106707153A (en) | 2016-12-27 | 2016-12-27 | FOA-RBF based high-voltage circuit breaker fault diagnosis method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611227922.1A CN106707153A (en) | 2016-12-27 | 2016-12-27 | FOA-RBF based high-voltage circuit breaker fault diagnosis method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106707153A true CN106707153A (en) | 2017-05-24 |
Family
ID=58895497
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611227922.1A Pending CN106707153A (en) | 2016-12-27 | 2016-12-27 | FOA-RBF based high-voltage circuit breaker fault diagnosis method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106707153A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107992665A (en) * | 2017-11-27 | 2018-05-04 | 国家电网公司 | A kind of ultra-high voltage converter station alternating current filter on-line fault diagnosis analysis method |
CN108734202A (en) * | 2018-04-27 | 2018-11-02 | 西安工程大学 | A kind of Fault Diagnosis for HV Circuit Breakers method based on improved BP |
CN108828441A (en) * | 2018-06-12 | 2018-11-16 | 江苏镇安电力设备有限公司 | Fault Diagnosis for HV Circuit Breakers method |
CN108896906A (en) * | 2018-05-04 | 2018-11-27 | 宁波新胜中压电器有限公司 | One or two fusion kV switch actuation characteristics testers and test method |
CN109447236A (en) * | 2018-09-28 | 2019-03-08 | 重庆邮电大学 | A kind of method for diagnosing faults of hybrid vehicle heat management system |
CN110333440A (en) * | 2019-07-04 | 2019-10-15 | 深圳供电局有限公司 | Detection method and detection device for circuit breaker, computer equipment and storage medium |
CN110542851A (en) * | 2019-08-29 | 2019-12-06 | 广州供电局有限公司 | Fault diagnosis method and device for circuit breaker operating mechanism, computer and storage medium |
CN112272856A (en) * | 2018-07-05 | 2021-01-26 | 菲尼克斯电气公司 | Circuit breaker with intelligent limit value determination and method thereof |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0641446B1 (en) * | 1992-05-22 | 1997-01-08 | Siemens Aktiengesellschaft | Testing process for the quality control of electromagnetically actuated switching devices |
CN104793134A (en) * | 2015-04-29 | 2015-07-22 | 中国电力科学研究院 | Breaker operating mechanism fault diagnosis method based on least square support vector machine |
CN105259495A (en) * | 2015-07-03 | 2016-01-20 | 四川大学 | High-voltage circuit breaker operation mechanism state evaluation method based on opening-closing coil current characteristic quantity optimization |
CN105445657A (en) * | 2015-11-26 | 2016-03-30 | 国家电网公司 | Breaker operating mechanism state diagnosis method based on grey relational analysis |
CN105606997A (en) * | 2016-02-26 | 2016-05-25 | 国家电网公司 | Mechanical fault diagnosis method of high voltage breaker operation mechanism for electric power system |
CN105866669A (en) * | 2016-05-19 | 2016-08-17 | 华南理工大学 | Circuit breaker breaking-closing control loop fault diagnosis method |
CN106019131A (en) * | 2016-05-18 | 2016-10-12 | 四川大学 | State comprehensive evaluation method for high-voltage circuit breaker operating mechanism based on switching-on and switching-off coil currents |
-
2016
- 2016-12-27 CN CN201611227922.1A patent/CN106707153A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0641446B1 (en) * | 1992-05-22 | 1997-01-08 | Siemens Aktiengesellschaft | Testing process for the quality control of electromagnetically actuated switching devices |
CN104793134A (en) * | 2015-04-29 | 2015-07-22 | 中国电力科学研究院 | Breaker operating mechanism fault diagnosis method based on least square support vector machine |
CN105259495A (en) * | 2015-07-03 | 2016-01-20 | 四川大学 | High-voltage circuit breaker operation mechanism state evaluation method based on opening-closing coil current characteristic quantity optimization |
CN105445657A (en) * | 2015-11-26 | 2016-03-30 | 国家电网公司 | Breaker operating mechanism state diagnosis method based on grey relational analysis |
CN105606997A (en) * | 2016-02-26 | 2016-05-25 | 国家电网公司 | Mechanical fault diagnosis method of high voltage breaker operation mechanism for electric power system |
CN106019131A (en) * | 2016-05-18 | 2016-10-12 | 四川大学 | State comprehensive evaluation method for high-voltage circuit breaker operating mechanism based on switching-on and switching-off coil currents |
CN105866669A (en) * | 2016-05-19 | 2016-08-17 | 华南理工大学 | Circuit breaker breaking-closing control loop fault diagnosis method |
Non-Patent Citations (3)
Title |
---|
李宗辉等: "基于果蝇优化算法的GRNN电网故障诊断", 《电气开关》 * |
李晓豁等: "基于果蝇算法优化模糊RBF网络的液压破碎锤故障诊断", 《工程设计学报》 * |
程勇等: "基于正交算法的 RBF网络在高压断路器故障诊断中的应用研究", 《电气开关》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107992665A (en) * | 2017-11-27 | 2018-05-04 | 国家电网公司 | A kind of ultra-high voltage converter station alternating current filter on-line fault diagnosis analysis method |
CN108734202A (en) * | 2018-04-27 | 2018-11-02 | 西安工程大学 | A kind of Fault Diagnosis for HV Circuit Breakers method based on improved BP |
CN108896906A (en) * | 2018-05-04 | 2018-11-27 | 宁波新胜中压电器有限公司 | One or two fusion kV switch actuation characteristics testers and test method |
CN108896906B (en) * | 2018-05-04 | 2020-09-11 | 宁波新胜中压电器有限公司 | Primary and secondary fusion switch action characteristic tester and testing method |
CN108828441A (en) * | 2018-06-12 | 2018-11-16 | 江苏镇安电力设备有限公司 | Fault Diagnosis for HV Circuit Breakers method |
CN112272856A (en) * | 2018-07-05 | 2021-01-26 | 菲尼克斯电气公司 | Circuit breaker with intelligent limit value determination and method thereof |
CN112272856B (en) * | 2018-07-05 | 2024-04-02 | 菲尼克斯电气公司 | Device circuit breaker with intelligent determination of limit value and method thereof |
CN109447236A (en) * | 2018-09-28 | 2019-03-08 | 重庆邮电大学 | A kind of method for diagnosing faults of hybrid vehicle heat management system |
CN110333440A (en) * | 2019-07-04 | 2019-10-15 | 深圳供电局有限公司 | Detection method and detection device for circuit breaker, computer equipment and storage medium |
CN110542851A (en) * | 2019-08-29 | 2019-12-06 | 广州供电局有限公司 | Fault diagnosis method and device for circuit breaker operating mechanism, computer and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106707153A (en) | FOA-RBF based high-voltage circuit breaker fault diagnosis method | |
CN109141847B (en) | Aircraft system fault diagnosis method based on MSCNN deep learning | |
CN105930901B (en) | A kind of Diagnosis Method of Transformer Faults based on RBPNN | |
CN102707256B (en) | Fault diagnosis method based on BP-Ada Boost nerve network for electric energy meter | |
CN110879917A (en) | Electric power system transient stability self-adaptive evaluation method based on transfer learning | |
CN105675802A (en) | Transformer fault diagnosis method | |
CN112557826A (en) | Ship electric power system fault diagnosis method | |
CN117421571A (en) | Topology real-time identification method and system based on power distribution network | |
CN111130053B (en) | Power distribution network overcurrent protection method based on deep reinforcement learning | |
CN118194222A (en) | SCADA data-based space-time fusion wind turbine generator fault prediction method | |
CN114169395A (en) | Construction method and application of dominant instability mode recognition model of power system | |
CN104218571A (en) | Running state evaluation method for wind power generation equipment | |
CN104182613B (en) | The ship electric power plant fault diagnosis model construction method of the petri net based on rough set | |
CN110033126A (en) | Shot and long term memory network prediction technique based on attention mechanism and logistic regression | |
CN109214500A (en) | A kind of transformer fault recognition methods based on integrated intelligent algorithm | |
CN114707613B (en) | Layered depth strategy gradient network-based power grid regulation and control method | |
Zhang et al. | Research on DBN-based evaluation of distribution network reliability | |
Li et al. | Visualization of CNN transient voltage classification based on feature recognition and enhancement | |
CN114861792A (en) | Complex power grid key node identification method based on deep reinforcement learning | |
Lin et al. | A method of satellite network fault synthetic diagnosis based on C4. 5 algorithm and expert knowledge database | |
Zhao et al. | Application of deep neural networks for fault diagnosis in a hybrid AC/DC power grid | |
Wu | Application of intelligent system based on deep reinforcement learning in electrical engineering automation control | |
CN113054653A (en) | Power system transient stability evaluation method based on VGGNet-SVM | |
Chen et al. | Research on Pattern Recognition Performance of Control Chart Based on Deep Learning | |
Ning | Comparison of several common intelligent fault diagnosis knowledge-based method under nonlinear small sample conditions |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170524 |
|
RJ01 | Rejection of invention patent application after publication |