CN109932644A - Circuit breaker failure diagnostic method based on integrated intelligent algorithm - Google Patents

Circuit breaker failure diagnostic method based on integrated intelligent algorithm Download PDF

Info

Publication number
CN109932644A
CN109932644A CN201910153404.7A CN201910153404A CN109932644A CN 109932644 A CN109932644 A CN 109932644A CN 201910153404 A CN201910153404 A CN 201910153404A CN 109932644 A CN109932644 A CN 109932644A
Authority
CN
China
Prior art keywords
value
standard state
weeds
seed
algorithm
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
Application number
CN201910153404.7A
Other languages
Chinese (zh)
Inventor
李永丽
吴玲玲
卢扬
孙广宇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University
Original Assignee
Tianjin University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tianjin University filed Critical Tianjin University
Priority to CN201910153404.7A priority Critical patent/CN109932644A/en
Publication of CN109932644A publication Critical patent/CN109932644A/en
Pending legal-status Critical Current

Links

Abstract

The present invention relates to a kind of circuit breaker failure diagnostic method based on integrated intelligent algorithm, including the following steps: monitoring breaker operation circuit coil current simultaneously extracts its current characteristic amount and time characteristic quantity, collects training sample and test sample;Standard state selection is carried out using invasive weed optimization algorithm;Data Discretization is carried out using the continuous variable discretization algorithm based on standard state probability assignments;Prior probability is calculated according to training sample;Calculate the posterior probability that test sample belongs to each fault type;Test sample is included into the maximum fault type of posterior probability, obtains fault diagnosis result.

Description

Circuit breaker failure diagnostic method based on integrated intelligent algorithm
Technical field
The invention belongs to electrical equipment and electrical engineering technical fields.
Background technique
The safe and stable operation of electric system has the fast development of national economy and the normal operation of society great Meaning.The main function of breaker is: the excision faulty line when electric system is broken down rapidly and accurately, to cut off event Hinder electric current, prevents from causing the more serious consequences such as large-area power-cuts, guarantee the safe and stable operation of electric system.Therefore, if The normal work of breaker tripping or malfunction, electric system or in which a part will be destroyed, and cause few to user Power transmission or power quality degenerate to unacceptable stage, or even cause the damage of personal injury and electrical equipment.Therefore, how Fast and accurately carrying out fault diagnosis to breaker has important research significance.
Existing research shows that the failure cause of breaker is caused to have very much, wherein mechanical breakdown (including operating mechanism and control Circuit processed) account for the 70%~80% of whole failures.The aging of breaker secondary circuit and operating mechanism in During Process of Long-term Operation It can cause failure, such as operation brownout, combined floodgate incipient stage bite, operating mechanism bite etc..Since operation circuit belongs to Secondary circuit, voltage class is lower and signal acquisition is relatively easy to, by monitoring that breaker operator circuit coil electric current can obtain To the information of circuit breaker internal, therefore, research is based on the diagnosis of breaker operator circuit coil failure of the current to diagnosis breaker Fault category guarantees that safe and stable operation of power system is of great significance.
Summary of the invention
The purpose of the present invention is be difficult to realize for accuracy rate is not high, algorithm is complicated present in artificial intelligence diagnosis' algorithm The problems such as, by studying the feature of breaker operator circuit coil electric current, proposes one kind and broken by Bayes classifier The method of road device fault diagnosis.Based on mentioned bayes classification method, the continuous change based on standard state probability assignments has been used Discretization algorithm is measured, solves the problems, such as that Bayes classifier is only applicable to discretization variable, meanwhile, optimized using invasive weed Algorithm carries out standard state selection, has obtained optimality criterion state Choice, has quick and precisely carried out circuit breaker failure to reach The purpose of diagnosis.Technical solution is as follows:
A kind of circuit breaker failure diagnostic method based on integrated intelligent algorithm, including the following steps:
1) monitoring breaker operation circuit coil current and its current characteristic amount and time characteristic quantity are extracted, collects training sample Sheet and test sample;
2) standard state selection is carried out using invasive weed optimization algorithm
2.1) initialization of population
Determine initialization population scale N0With maximum population scale Nmax, maximum number of iterations itermax, required problem dimension D can generate the upper limit seed of seed numbermaxWith lower limit seedmin, nonlinear exponent n, standard deviation initial value σinitAnd end value σfinal, initial ranging space minimum value xminWith maximum value xmax
2.2) fitness function
According to the extracted current characteristic amount of step 1) and time characteristic quantity, definition variance is fitness function:
In formula, e (x) indicates standard state belonging when dividing according to the standard state currently chosen, and x indicates data root Numerical value after being divided according to the standard state currently chosen, θ are a multidimensional data, respectively indicate each characteristic quantity and are characterized Attribute standard state choosing method;
2.3) it breeds
Weeds generated seed amount in reproductive process is related with the fitness value of weeds, and fitness value is better, numerous The seed grown is more, calculates seed number m according to following equation:
In formula, m indicates seed number;F indicates the fitness value of current weeds;fminAnd fmaxIt respectively indicates miscellaneous in current population The minimum fitness value and maximum adaptation angle value of grass
2.4) space diffusion profile
The seed that weeds generate is distributed in parent weeds by σ according to being mean value with 0 in the way of the normal distribution of standard deviation Around, weeds of new generation are grown into, the standard deviation of every generation calculates according to the following formula:
In formula, iter indicates current iteration number;δcurIndicate current standard deviation;N indicates the non-linear reconciliation factor, usually It is set as 3;
2.5) competitive exclusion rule
When the sum of weeds and the number of offspring reach preset maximum population scale NmaxWhen, algorithm will execute competitive row Reprimand rule is ranked up the fitness value of all individuals in the population of weeds and offspring's composition by size, retains fitness value High individual eliminates remaining individual;
2.6) termination condition
Maximum number of iterations itermaxOr the fitness function value of adjacent generations does not terminate iteration in variation twice, most Standard state choosing method θ is obtained eventually;
3) continuous variable discretization algorithm of the application based on standard state probability assignments carries out Data Discretization;
3.1) standard state is reasonably selected using invasive weed optimization algorithm, data is formed to the discretization with discrimination As a result;
3.2) after selecting optimality criterion state, according to selected standard state, the standard state probability of the attribute is calculated Allocation matrix Px
4) prior probability is calculated according to training sample.
5) posterior probability that test sample belongs to each fault type is calculated.
6) test sample is included into the maximum fault type of posterior probability, obtains fault diagnosis result.
Detailed description of the invention
Fig. 1 invasive weed optimization algorithm flow chart
Fig. 2 Bayes's classification flow chart
Fig. 3 fault diagnosis result figure
Fig. 4 divide-shut brake coil current signature waveform figure
Specific embodiment
Bayesian algorithm is a kind of probabilistic approach using statistics as Fundamentals of Mathematics, and the failure of small-scale data is examined Disconnected performance is good, is capable of handling more classification tasks
Naive Bayes Classifier is earliest bayes classification method, and core concept is that selection highest posterior probability is made For the index for determining classification.This method is assumed to be independent from each other between each conditional attribute, i.e. any one conditional attribute value Probability do not influenced completely by other conditions attribute.
Equipped with n breaker monitoring state information, X is usediIndicate (status information is 8 characteristic values), XiValue be xi, Wherein 1≤i≤8 and i ∈ Z indicate that (fault type of division is divided into six kinds to fault category variable, respectively normally, operates with C Brownout, close a floodgate the incipient stage have bite, operating mechanism have bite, iron core idle stroke excessive and auxiliary switch movement contact not It is good), the value of C is ci, ci∈{1,2,3,4,5,6}.T represents training sample, T={ x1,x2,…,xn,ciIndicate training sample This.At this point, ciConditional probability can rewrite are as follows:
Therefore, according to Bayesian formula, posterior probability it is writeable are as follows:
Test sample is calculated for the posterior probability of whole fault types according to formula (4), and highest posterior probability is selected to make For the index for determining classification.
In order to reduce loss of data in discretization process, discretization probability matrix is calculated using formula (6).
Troubleshooting step is as follows:
1) monitoring breaker operation circuit coil current and three of them current characteristic amount and five temporal characteristics amounts are extracted, such as Shown in Fig. 4, training sample is collected, format is as shown in table 1.
1 sample data of table
Table 1Sampledata
2) standard state selection is carried out using invasive weed optimization algorithm.
2.1) initialization of population
Determine initialization population scale N0With maximum population scale Nmax, maximum number of iterations itermax, required problem dimension D can generate the upper limit seed of seed numbermaxWith lower limit seedmin, nonlinear exponent n, standard deviation initial value σinitAnd end value σfinal, initial ranging space minimum value xminWith maximum value xmax
2.2) it breeds
Weeds generated seed amount in reproductive process is related with the fitness value of weeds, and fitness value is better, numerous The seed grown is more.Therefore, the seed number of different weeds generations is not identical, can calculate seed number with formula (9).
2.3) space diffusion profile
The seed that weeds generate is distributed in parent weeds by σ according to being mean value with 0 in the way of the normal distribution of standard deviation Around, grow into weeds of new generation.The standard deviation of every generation is calculated with formula (10).
2.4) competitive exclusion rule
When the sum of weeds and the number of offspring reach preset maximum population scale NmaxWhen, algorithm will execute competitive row Reprimand rule is ranked up the fitness value of all individuals in the population of weeds and offspring's composition by size, retains fitness value High individual (for optimizing maximum value) eliminates remaining individual.Weeds are first made to breed and capture rapidly the field of adaptation, Then the stronger weeds of competitiveness under opposite stable environment are remained and continue searching space.
3) continuous variable discretization algorithm of the application based on standard state probability assignments carries out Data Discretization.
3.1) standard state e is reasonably selected using invasive weed optimization algorithmx=[ex1,ex2,…,exn], data are formed Discretization results with discrimination.
3.2) after selecting optimality criterion state, according to selected standard state, the standard state probability of the attribute is calculated Allocation matrix Px, PxFor the matrix of m × n rank, Px(m, n) indicates the x of m-th of sampleiBelong to n-th of standard state exnProbability. Due to a data only can in probability assignments to two continuous standard state, PxThe calculation formula of middle element is formula (11).
4) prior probability is calculated according to following formula
P (y=eyj, z=ezl...) and=Py(k,j)·Pz(k,l)…
5) posterior probability that test sample belongs to each fault type is calculated according to the following formula.
In formula, PxIndicate that fault category standard state probability assignments matrix (is assigned to the probability in each standard state It can be 1 or 0);PyAnd PzIndicate the standard state probability assignments matrix of each conditional attribute.
6) test sample is included into the maximum fault type of posterior probability.
Conventional method is the Bayes's classification that discretization is carried out based on interval division, and method of the invention is invasive weed Optimization and Bayes's integrated intelligent algorithm.In order to analyze the validity that this method diagnoses circuit breaker failure, with temporal characteristics amount Parameter is inputted as sample with 8 parameters of current characteristic amount, randomly selects 20 groups of data as training sample, then is taken 10 groups remaining Data are as test sample, respectively using the bayes method of conventional discrete and invasive weed of the present invention optimization and shellfish This hybrid algorithm of leaf carries out fault diagnosis to breaker, and fault diagnosis result is as shown in table 2, theoretical value, the method in table Value, the numerical value expression breaker mechanical state of conventional method value.
The result shows that the use of the accuracy rate that conventional method carries out fault diagnosis being 50%, and the method is used to carry out The predicted value of fault diagnosis and theoretical value coincidence factor are higher, and the accuracy rate of fault diagnosis reaches 100%, significantly larger than conventional side Method.
Although simultaneously this method failure diagnosis time be 2.29ms compared to conventional method fault diagnosis speed under Drop, but still there is greater advantage compared to the diagnosis speed (66ms) of neural network algorithm.
2 fault diagnosis result of table
Table 5The result offault diagnosis
The accuracy of fault diagnosis, training are carried out to breaker in order to analyze this method in the case where different sample sizes Sample size takes 10,12,14,16,18,20 respectively, and corresponding test sample quantity all takes 10 groups, uses the mixing intelligence Energy algorithm and conventional method carry out fault diagnosis to breaker, and every group of repetition is diagnosed 30 times and be averaged, fault diagnosis Accuracy rate is as shown in Figure 3.
It can be seen from figure 3 that obviously high using invasive weed optimization and Bayes's integrated intelligent algorithm, fault diagnosis accuracy rate In the fault diagnosis accuracy rate of conventional method.Because conventional method will cause data information and largely lose, and of the present invention Invasive weed optimization and Bayes's integrated intelligent algorithm by apply invasive weed optimization algorithm selection standard state, will be discrete As a result it is divided into two parts, a part is discrete standard state, and a part is continuous probability, is believed to reduce in discretization process Breath is lost, to improve the accuracy rate of fault diagnosis.
Simultaneously it can also be seen that being carried out using invasive weed of the present invention optimization and Bayes's integrated intelligent algorithm When fault diagnosis, fault diagnosis accuracy rate will not with the reduction of sample number sharp fall, therefore, for sample This data carry out fault diagnosis, still there is good diagnosis effect.

Claims (1)

1. a kind of circuit breaker failure diagnostic method based on integrated intelligent algorithm, including the following steps:
1) monitoring breaker operation circuit coil current and its current characteristic amount and time characteristic quantity are extracted, collect training sample and Test sample;
2) standard state selection is carried out using invasive weed optimization algorithm
2.1) initialization of population
Determine initialization population scale N0With maximum population scale Nmax, maximum number of iterations itermax, required problem dimension D, energy Generate the upper limit seed of seed numbermaxWith lower limit seedmin, nonlinear exponent n, standard deviation initial value σinitWith end value σfinal, Initial ranging space minimum value xminWith maximum value xmax
2.2) fitness function
According to the extracted current characteristic amount of step 1) and time characteristic quantity, definition variance is fitness function:
In formula, e (x) indicates standard state belonging when dividing according to the standard state currently chosen, and x indicates data according to working as The standard state of preceding selection divided after numerical value, θ be a multidimensional data, respectively indicate the category that each characteristic quantity is characterized The standard state choosing method of property;
2.3) it breeds
Weeds generated seed amount in reproductive process is related with the fitness value of weeds, and fitness value is better, breeding Seed is more, calculates seed number m. according to following equation
In formula, m indicates seed number;F indicates the fitness value of current weeds;fminAnd fmaxRespectively indicate weeds in current population Minimum fitness value and maximum adaptation angle value
2.4) space diffusion profile
The seed that weeds generate is distributed in the week of parent weeds by σ according to being mean value with 0 in the way of the normal distribution of standard deviation It encloses, grows into weeds of new generation, the standard deviation of every generation calculates according to the following formula:
In formula, iter indicates current iteration number;δcurIndicate current standard deviation;N indicates the non-linear reconciliation factor, is usually arranged It is 3;
2.5) competitive exclusion rule
When the sum of weeds and the number of offspring reach preset maximum population scale NmaxWhen, algorithm will execute competitive exclusion rule Then, the fitness value of all individuals is ranked up by size in the population formed to weeds and offspring, and it is high to retain fitness value Individual eliminates remaining individual;
2.6) termination condition
Maximum number of iterations itermaxOr the fitness function value of adjacent generations does not terminate iteration in variation twice, it is final to obtain To standard state choosing method θ;
3) continuous variable discretization algorithm of the application based on standard state probability assignments carries out Data Discretization;
3.1) standard state is reasonably selected using invasive weed optimization algorithm, data is formed to the discretization knot with discrimination Fruit;
3.2) after selecting optimality criterion state, according to selected standard state, the standard state probability assignments of the attribute are calculated Matrix Px
4) prior probability is calculated according to training sample;
5) posterior probability that test sample belongs to each fault type is calculated;
6) test sample is included into the maximum fault type of posterior probability, obtains fault diagnosis result.
CN201910153404.7A 2019-02-28 2019-02-28 Circuit breaker failure diagnostic method based on integrated intelligent algorithm Pending CN109932644A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910153404.7A CN109932644A (en) 2019-02-28 2019-02-28 Circuit breaker failure diagnostic method based on integrated intelligent algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910153404.7A CN109932644A (en) 2019-02-28 2019-02-28 Circuit breaker failure diagnostic method based on integrated intelligent algorithm

Publications (1)

Publication Number Publication Date
CN109932644A true CN109932644A (en) 2019-06-25

Family

ID=66986067

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910153404.7A Pending CN109932644A (en) 2019-02-28 2019-02-28 Circuit breaker failure diagnostic method based on integrated intelligent algorithm

Country Status (1)

Country Link
CN (1) CN109932644A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111272222A (en) * 2020-02-28 2020-06-12 西南交通大学 Transformer fault diagnosis method based on characteristic quantity set

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103135032A (en) * 2013-01-30 2013-06-05 福建省电力有限公司 External force factor diagnostic method causing single phase ground connection breakdown of electric transmission line
CN103245911A (en) * 2013-05-03 2013-08-14 云南电力试验研究院(集团)有限公司电力研究院 Breaker fault diagnosis method based on Bayesian network
CN103245861A (en) * 2013-05-03 2013-08-14 云南电力试验研究院(集团)有限公司电力研究院 Transformer fault diagnosis method based on Bayesian network
CN103606005A (en) * 2013-09-24 2014-02-26 西安电子科技大学 Conformal antenna array directional diagram comprehensive method based on invasive weed optimization
CN103679199A (en) * 2013-12-11 2014-03-26 中国石油大学(华东) Noisy independent component analysis method based on invasiveness weed algorithm
CN105606914A (en) * 2015-09-06 2016-05-25 南京航空航天大学 IWO-ELM-based Aviation power converter fault diagnosis method
CN107942312A (en) * 2017-11-13 2018-04-20 浙江大学 A kind of Intelligent radar sea target detection system and method based on differential evolution invasive weed optimization algorithm
CN108810914A (en) * 2018-05-08 2018-11-13 苏州工业职业技术学院 Based on the WSN Node distribution optimization methods for improving weeds algorithm
CN108918111A (en) * 2018-05-16 2018-11-30 国网山东省电力公司莱芜供电公司 Breaker mechanical method for diagnosing faults based on the classification of k- neighbour's Bayes's multi-tag

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103135032A (en) * 2013-01-30 2013-06-05 福建省电力有限公司 External force factor diagnostic method causing single phase ground connection breakdown of electric transmission line
CN103245911A (en) * 2013-05-03 2013-08-14 云南电力试验研究院(集团)有限公司电力研究院 Breaker fault diagnosis method based on Bayesian network
CN103245861A (en) * 2013-05-03 2013-08-14 云南电力试验研究院(集团)有限公司电力研究院 Transformer fault diagnosis method based on Bayesian network
CN103606005A (en) * 2013-09-24 2014-02-26 西安电子科技大学 Conformal antenna array directional diagram comprehensive method based on invasive weed optimization
CN103679199A (en) * 2013-12-11 2014-03-26 中国石油大学(华东) Noisy independent component analysis method based on invasiveness weed algorithm
CN105606914A (en) * 2015-09-06 2016-05-25 南京航空航天大学 IWO-ELM-based Aviation power converter fault diagnosis method
CN107942312A (en) * 2017-11-13 2018-04-20 浙江大学 A kind of Intelligent radar sea target detection system and method based on differential evolution invasive weed optimization algorithm
CN108810914A (en) * 2018-05-08 2018-11-13 苏州工业职业技术学院 Based on the WSN Node distribution optimization methods for improving weeds algorithm
CN108918111A (en) * 2018-05-16 2018-11-30 国网山东省电力公司莱芜供电公司 Breaker mechanical method for diagnosing faults based on the classification of k- neighbour's Bayes's multi-tag

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
WANG YAN等: "Bayesian Network Based Fault Section Estimation in Power Systems", 《 TENCON 2006 - 2006 IEEE REGION 10 CONFERENCE》 *
杨睿等: "基于粗糙集和贝叶斯网络的电力信息通信系统快速故障诊断方法", 《科技通报》 *
荣亚君等: "用粗糙集理论和贝叶斯网络诊断SF6断路器故障", 《高电压技术》 *
谢文靖等: "基于概率神经网络的高压断路器故障诊断模型", 《江南大学学报(自然科学版)》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111272222A (en) * 2020-02-28 2020-06-12 西南交通大学 Transformer fault diagnosis method based on characteristic quantity set

Similar Documents

Publication Publication Date Title
Xie et al. Dimensionality reduction of synchrophasor data for early event detection: Linearized analysis
CN102638100B (en) District power network equipment abnormal alarm signal association analysis and diagnosis method
US8200372B2 (en) Methods and processes for managing distributed resources in electricity power generation and distribution networks
Tang et al. Condition monitoring and assessment of power transformers using computational intelligence
Kamwa et al. Development of rule-based classifiers for rapid stability assessment of wide-area post-disturbance records
Morison et al. Power system security assessment
Ni et al. Software implementation of online risk-based security assessment
Musirin et al. On-line voltage stability based contingency ranking using fast voltage stability index (FVSI)
Thang et al. Analysis of power transformer dissolved gas data using the self-organizing map
CN106291351B (en) High-voltage circuitbreaker fault detection method based on convolutional neural networks algorithm
Jain et al. Fast voltage contingency screening using radial basis function neural network
CN104392390B (en) A kind of secondary equipment of intelligent converting station appraisal procedure based on TOPSIS models
CN102707256B (en) Fault diagnosis method based on BP-Ada Boost nerve network for electric energy meter
CN103744850B (en) A kind of electrical network disaster real-time monitoring device and method based on intuitionistic fuzzy-rough sets
CN104700321B (en) A kind of power transmission and transformation equipment state operation trend analysis method
US20150154504A1 (en) Support vector machine enhanced models for short-term wind farm generation forecasting
CN106251059B (en) Cable state evaluation method based on probabilistic neural network algorithm
Xu et al. A classification approach for power distribution systems fault cause identification
US5625751A (en) Neural network for contingency ranking dynamic security indices for use under fault conditions in a power distribution system
da Silva et al. Composite reliability assessment based on Monte Carlo simulation and artificial neural networks
CN102779230B (en) State analysis and maintenance decision judging method of power transformer system
CN106154209A (en) Electrical energy meter fault Forecasting Methodology based on decision Tree algorithms
CN103630869A (en) Clustering algorithm-based exceptional event analysis method for evaluating whole state of electric meter
CN101464964B (en) Pattern recognition method capable of holding vectorial machine for equipment fault diagnosis
Jiang et al. Spatial-temporal synchrophasor data characterization and analytics in smart grid fault detection, identification, and impact causal analysis

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20190625

RJ01 Rejection of invention patent application after publication