CN105260863A - Fault single influence factor analysis method based on power cable fault information - Google Patents
Fault single influence factor analysis method based on power cable fault information Download PDFInfo
- Publication number
- CN105260863A CN105260863A CN201510845900.0A CN201510845900A CN105260863A CN 105260863 A CN105260863 A CN 105260863A CN 201510845900 A CN201510845900 A CN 201510845900A CN 105260863 A CN105260863 A CN 105260863A
- Authority
- CN
- China
- Prior art keywords
- cable
- influence factor
- fault
- data
- sample
- 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
Abstract
The invention discloses a fault single influence factor analysis method based on power cable fault information, which mainly comprises two steps of sample size estimation and single influence factor analysis based on a Cox proportional hazard model. Firstly, cable fault data is preprocessed on the basis of a data preprocessing principle to obtain sample data, and sample size is estimated based on the sample data, so as to guarantee the applicability of the sample data. Then, the Cox proportional hazard model is used for carrying out single influence factor analysis on influence factors of cable faults. The fault single influence factor analysis method based on power cable fault information can determine the influence factors of cable faults and influence degree of each influence factor according to cable fault data, so as to guide procurement, construction and maintenance of the cable, and improve the reliability of cable lines as well as operational reliability and stability of a power grid.
Description
Technical field
The invention belongs to power equipment operation management technical field, particularly relate to a kind of Trouble ticket analysis of Influential Factors method based on power cable fault information.
Background technology
Up to now, less to the research of power cable fault analysis of Influential Factors both at home and abroad, mainly can consider that the method used has following several.
(1) principal component analysis (PCA).
Principal component analysis (PCA) studies the important statistical method of one how multi objective problem being converted into less overall target, the problem of higher dimensional space can be transformed into lower dimensional space and go process by it, problem is made to become fairly simple, directly perceived, and uncorrelated mutually between these less overall targets, most information of original index can be provided again.But applicability inspection to be carried out to raw data before carrying out principal component analysis (PCA), stronger correlativity should be there is between each variable of inspection raw data, there is the possibility of data structure simplification.
(2) factor analysis.
The basic thought of factor analysis is classified by observational variable, correlativity is higher, and namely contact and divide more closely in same class, the correlativity between inhomogeneity variable is then lower, so each class variable in fact just represents basic structure, i.e. a common factor.Its feature is to adopt the mode merged to reduce situational variables, and information loss achieves and minimizes, but computation process is complicated, such as needs to carry out factor rotation, factor score calculating etc.
(3) gray relative analysis method.
Grey Incidence Analysis is in systems development process, if the trend of two factors vary has consistance, namely synchronous intensity of variation is higher, is namely that the two correlation degree is higher; Otherwise, then lower.Therefore, gray relative analysis method is according to the similar of development trend between factor or different degree, that is " grey relational grade ", as correlation degree between measurement factor.The method is applicable to dynamic course analysis, but requires that data volume is large, calculation of complex.
Summary of the invention
The object of the present invention is to provide a kind of Trouble ticket analysis of Influential Factors method based on power cable fault information, income analysis result can be used for instructing the buying of cable, construction and maintenance.
For achieving the above object, the present invention adopts following technical scheme:
Based on a Trouble ticket analysis of Influential Factors method for power cable fault information, comprising:
Step 1, based on sample data sample estimates amount, this step comprises sub-step further:
1.1 pre-service cable fault data, obtain sample data, and before described sample data comprises the fault that 1. each cable is corresponding, the concrete of working time or truncated time, 2. malfunction, 3. influence factor and correspondence thereof affects element;
1.2 based on sample data, and for affecting element under each influence factor, influence factor i-th is affected element and is designated as A group, under this influence factor, other influences factor is designated as B group, according to N=(Z
1-α+ Z
1-γ)
2[dp (1-p) (ln Δ)
2]
-1calculate i-th and affect sample size N corresponding to element, the maximal value i.e. smallest sample amount of this influence factor in sample size; Wherein, d is the failure rate of cable, and the cable count namely broken down accounts for the ratio of cable sum; P affects the ratio that cable count corresponding to element accounts for cable sum in A group;
h
afor affecting the failure rate of cable corresponding to element in A group, namely in A group, affect the ratio that failure cable number corresponding to element accounts for the cable sum of its correspondence; h
bfor affecting the failure rate of cable corresponding to element in B group, namely in B group, affect the ratio that failure cable number corresponding to element accounts for the cable sum of its correspondence; Z
1-αfor inspection level; Z
1-γfor the power of test of expection; I=1,2 ... I, I are the picture elements number under influence factor;
Step 2, single analysis of Influential Factors, this step comprises further:
2.1 choose the sample data of current analysis of Influential Factors according to the smallest sample amount of current influence factor;
2.2 respectively affect element under current influence factor, adopt SPSS Software tool to carry out respectively:
A () is based on the sample data of current analysis of Influential Factors, the covariant of element for Cox proportional hazard model is affected with current, be 0 as null hypothesis using the current regression parameter affecting element, adopt association's variations per hour method to carry out PH test of hypothesis, judge that whether current to affect element relevant with cable fault according to the Wald value obtained; If relevant, perform sub-step (b); Otherwise, terminate;
B (), using the current element that affects as benchmark, generating dummy variable affects the relatively current relative risk affecting element of element, thus acquisition dummy variable affects element and the current size affecting element and affect cable fault.
Above-mentioned sub-step 1.1 is carried out for cable fault data, comprises further:
(a) rule of thumb with demand artificially selected CABLE MATERIALS fault influence factor and corresponding affect element;
B () carries out respectively for each influence factor: calculating respectively affects the ratio that cable count corresponding to element accounts for cable sum, and that deletes that this ratio is less than preset value affects cable fault data corresponding to element, and preset value sets according to actual requirement;
C (), to the cable broken down He do not break down, obtains working time and truncated time before fault respectively;
D the malfunction of () mark cable, cable fault state comprises the two states that do not break down and break down.
Above-mentioned influence factor is one or more in cable body manufacturer, cable accessory manufacturer, unit in charge of construction, cable length.In the malfunction of described mark cable, the state adopting " 0 " mark cable not break down, adopts the state that one token cable breaks down.
Tool of the present invention has the following advantages:
The inventive method according to cable fault data, can determine the influence factor of cable fault and the influence degree of each influence factor, thus instructs the buying of cable, construction and maintenance, with reliability, the stability of the reliability and operation of power networks that improve cable line.
Accompanying drawing explanation
Fig. 1 is PH test of hypothesis schematic diagram;
Fig. 2 is the particular flow sheet of the inventive method.
Embodiment
Below in conjunction with accompanying drawing to thinking of the present invention, based on correlation theory and embodiment be described in detail.
The present invention mainly comprises Estimation of Sample Size and single analysis of Influential Factors two steps based on Cox proportional hazard model.First, based on data prediction principle, pre-service is carried out to cable fault data and obtain sample data, based on sample data sample estimates amount, to ensure the applicability of sample data.Then, Cox proportional hazard model is used to carry out single analysis of Influential Factors to the influence factor of cable fault.Cox proportional hazard model is existing model, is proposed in 1972 by Britain statistician D.R.Cox.
Step 1, Estimation of Sample Size.
In view of influence factor variable is many classified variables and Cox proportional hazard model feature, before carrying out single analysis of Influential Factors, pre-service need be carried out to cable fault data and obtain sample data.Described cable fault data obtain mainly through carrying out observation to cable status, arrange truncated time to viewing test, before truncated time, cable once break down, record trouble time the viewing test terminated this cable.
(1) pre-service of cable fault data
Pre-service is carried out based on following data prediction principle:
What a () determined influence factor and correspondence thereof affects element.
Influence factor is rule of thumb artificially selected with demand, such as, investigates different cable body manufacturer and different cable length to the influence degree of cable fault, cable body manufacturer and cable length can be selected as influence factor if want.Each image factors all somely affects element to having.Such as, cable body manufacturer influence factor, the element that affects of its correspondence is each concrete cable body manufacturer; To cable length influence factor, preset variant length of interval, these length of interval and cable length influence factor corresponding affect element.
B () deletes small probability data.
With cable fault data for handling object, carry out respectively for each influence factor.If certain affects the ratio that cable count corresponding to element account for cable sum and is less than preset value in influence factor, then this affects cable fault data corresponding to element is small probability data.Small probability data are very little on the impact of overall data rule, delete small probability data and problem can be made to simplify.
In this concrete enforcement, preset value is set to 0.05, but is not limited to this value.Preset value is empirical value, can carry out setting or adjusting according to actual requirement.
Such as, cable fault data comprise the fault data of 100 cables, and with cable body manufacturer for influence factor, cable body manufacturer M1 is that of this influence factor affects element.Article 100, only have 2 cables to be that cable body manufacturer M1 produces in cable, then the cable count that M1 is corresponding is 2, and cable adds up to 100,2/100=0.02, then the fault data of the cable of M1 production should be deleted.
C () is according to working time or truncated time before cable fault data acquisition cable fault.
To the cable broken down, obtain working time before its fault, this working time and failure date deduct the date of putting into operation.To the cable do not broken down, obtain its truncated time, truncated time and truncation date deduct the date of putting into operation.
(d) mark cable malfunction.
Represent with " 0 " state that cable does not break down, represent with " 1 " state that cable breaks down.
This is concrete implements, and the sample data obtained to comprise before cable I D and fault corresponding to each cable working time or truncated time, malfunction, body manufacturer, annex manufacturer, unit in charge of construction, cable length.
(2) Estimation of Sample Size
The Cox Proportional hazards Return Law is applied widely in survival analysis, but, need the problem of how many sample sizes not to be well solved about application the method actually always.
Cox proportional hazard model is as follows:
h(t|x
1,x
2,…,x
p)＝h
0(t)exp(β
1x
1+β
2x
2+…+β
px
p)(1)
In formula (1), h (t|x
1, x
2..., x
p) represent risk function, h
0risk function based on (t), t represents working time or truncated time before fault; x
irepresent i-th may be relevant to cable fault covariant, i.e. influence factor; β
ifor x
iregression parameter.
Two classification independent variable x
1with other covariant x
2, x
3... x
pseparate.When carrying out single analysis of Influential Factors, as only considered influence factor x
i, then the regression coefficient of other influence factor is 0.X
ivalue is designated as A group when being 1, x
ivalue is designated as B group when being 0.A group refers to influence factor x
ifor the component that certain is considered, B group refers to influence factor x
ifor the non-component that certain is considered.
For the body manufacturer influence factor of cable, x
ibe equivalent to affect element under body manufacturer influence factor, it is a variable, and this variable can be taken as body manufacturer M1, M2 etc.To body manufacturer M1, body manufacturer M1 is designated as A group, non-body manufacturer M1 is designated as B group, then adopts sample computing formula (2) to calculate Estimation of Sample Size value N1 corresponding to body manufacturer M1.To body manufacturer M2, body manufacturer M2 is designated as A group, non-body manufacturer M2 is designated as B group, then adopts sample computing formula (2) to calculate Estimation of Sample Size value N2 corresponding to body manufacturer M2.The like, calculate body manufacturer M3, M4 respectively ... corresponding Estimation of Sample Size value N3, N4 ...The cable sample amount that in the Estimation of Sample Size value that all body manufacturers are corresponding, maximal value is namely required.
The concrete computation process respectively affecting Estimation of Sample Size value corresponding to element under certain influence factor is as follows:
Test of hypothesis is: null hypothesis H0: β
i=0; Alternative hypothesis H1:
β
ifor regression parameter in Cox proportional hazard model.When H0 sets up, represent that failure rate when every bar cable is designated as A group and B two groups is respectively equal.When H1 sets up,
Nineteen eighty-three Schoenfeld utilizes the large sample theory based on likelihood function, uses the gradation inspection null hypothesis H0 of Score Test, therefore proposes the formula calculating Cox proportional hazard model sample size, as follows:
N＝(Z
1-α+Z
1-γ)
2[dp(1-p)(lnΔ)
2]
-1(2)
Adopt formula (2) to calculate respectively and respectively affect Estimation of Sample Size value corresponding to element, be specially: affect element by i-th and be designated as A group, other influences factor is designated as B group, adopt formula (2) to calculate i-th and affect sample size corresponding to element, now, in formula (2), N represents in A group affects sample size corresponding to element; D is the failure rate of cable, and the cable count namely broken down in sample data accounts for the ratio of cable sum; P affects the ratio that cable count corresponding to element accounts for cable sum in A group;
h
afor affecting the failure rate of cable corresponding to element in A group, namely in A group, affect the ratio that failure cable number corresponding to element accounts for the cable sum of its correspondence; h
bfor affecting the failure rate of cable corresponding to element in B group, namely in B group, affect the ratio that failure cable number corresponding to element accounts for the cable sum of its correspondence; Z
1-α=0.05, represent inspection level; Z
1-γ=80%, represent the power of test of expection.
Make i get 1 successively, 2 ... q, then allly under can obtaining influence factor affect Estimation of Sample Size value corresponding to element, and q affects first prime number under representing influence factor.
Two, Cox proportional hazard model is analyzed
Classical Cox proportional hazard model shown in formula (1) is expressed as follows:
In formula (3), h (t|X) represents risk function, and X is influence factor collection, and t represents working time or truncated time before fault; h
0risk function based on (t); x
irepresent i-th may be relevant to cable fault influence factor, β
ifor x
iregression parameter, p represent influence factor concentrate impact first prime number.Work as β
ifor timing, represent influence factor x
iwith cable fault positive correlation; Work as β
ifor time negative, represent influence factor x
iwith cable fault negative correlation; Work as β
iwhen being 0, represent influence factor x
iuncorrelated with cable fault.
In classical Cox proportional hazard model, do not consider basic risk function h
0t the distribution form of (), therefore classical Cox proportional hazard model is also called semi-parameter model.
(1) PH test of hypothesis correlation theory
The Hazard ratio RR i.e. ratio of two individual risk functions of classical Cox proportional hazard model, is shown in formula (4).Individuality of the present invention refers to cable.
In formula (4), h
i(t), h
jt () represents the risk function of cable i and j respectively.
Can find out according to formula (4), Hazard ratio RR and basic risk function h
0t () has nothing to do, and have nothing to do with time t, is called for short PH hypothesis.
Classical Cox proportional hazard model must meet PH hypothesis, and when covariant does not meet PH hypothesis, model lost efficacy, and can reduce the test effect of model.The covariant simultaneously increased in time for Hazard ratio, can over-evaluate its Relative hazard, so time Cox proportional hazard model can not use, and other method need be sought.When covariant meets PH hypothesis, then carry out next step analysis.The method of inspection of PH hypothesis generally has two large classes, graphic interpretation and time covariant method.Graphic interpretation is fairly simple, but adds the difficulty of range estimation judgement, covariant method when therefore adopting.Relate to P value in test of hypothesis, P value refers to the probability that sample observed result occurs or the result more extreme than sample occurs when null hypothesis obtains for true time.When P value is very little, showing that the probability that null hypothesis occurs is very little, is a small probability event, so just has reason to refuse null hypothesis.According to P value size, there are following two kinds of situations:
A () P < 0.05, judges by force, refusal null hypothesis;
B () P > 0.05, weak judgement, accepts null hypothesis.
Figure 1 shows that PH one-sided hypothesis test schematic diagram, when P value is less than the level of significance 0.05 preset, observation data has dropped on region of rejection, now should refuse null hypothesis.
Time covariant method by introducing the time covariant of a structure on the basis of original classical model, see formula (5).Only have a covariant x in hypothesized model, the time covariant of structure is xlg (t), then tests.The null hypothesis H of test of hypothesis
0: regression parameter is 0.Judged by the P value of inspection, as P > 0.05, show that null hypothesis is set up, model parameter is 0, i.e. δ=0, shows that the reciprocation of covariant x and time is not obvious.
h(t,x)＝h
0(t)exp(β·x+δ·xlg(t))(5)
(2) single analysis of Influential Factors
Such as, according to the 10kV cable data that certain power supply administration provides, the influence factor of substantially known 10kV cable fault may be cable body manufacturer, cable accessory manufacturer, unit in charge of construction and cable length.This four classes influence factor will be considered in this embodiment.Due to great majority in the cable data that Utilities Electric Co. provides about the data record of cable body manufacturer, cable accessory manufacturer, unit in charge of construction, cable length etc. occurs incomplete, therefore this four classes influence factor can not be brought in Cox proportional hazard model simultaneously carry out multiplicity, but can, according to the data cases of influence factor, select different cables to carry out single factor analysis.
The test of hypothesis principle that what single analysis of Influential Factors adopted is in statistics.When the regression coefficient of covariant is 0 in Cox scale model, illustrate that this covariant and cable fault have nothing to do; When the regression coefficient of covariant is non-zero, illustrate that this covariant is relevant with cable fault.In SPSS (StatisticalProductandServiceSolutions, statistical product and service solution) instrument, solve the regression parameter of Cox proportional hazard model, null hypothesis β is set
i=0, namely suppose that covariant and cable fault have nothing to do, checked by Wald, inspection null hypothesis, receive null hypothesis when P value is greater than 0.05, namely think that covariant and cable fault have nothing to do, otherwise covariant is relevant with cable fault.
Respectively affect the influence degree of element to cable fault under can obtaining influence factor according to single analysis of Influential Factors, thus provide guidance to the buying of cable, construction and maintenance.
(3) logistic regression model (Logic Regression Models) will be adopted below to determine cable fault influence factor kind and the extent of injury to cable fault, to verify the correctness of the inventive method.
Logistic regression model belongs to probabilistic type non-linear regression, a kind of multivariable technique of relation between main research two classification observations and some influence factors.Whether fault belongs to two classified variables to cable, the influence factor of cable fault belongs to many classified variables, therefore Logistic regression model is applicable to determining fault effects factor kind, but can not differentiate the influence degree size of influence factor, can not do cable buying, construction guidance.
Logistic regression model primary expression form is as follows:
In formula (6): Q by the failure rate of research cable colony, x
1, x
2... x
krepresent the influence factor of cable fault, b
ifor x
icoefficient, i=1,2 ... k.Work as b
iwhen=0, influence factor x
ito cable, whether fault does not make significant difference; Work as b
iwhen ≠ 0, influence factor x
ito cable, whether fault does not make significant difference.
Influence factor kind can be determined based on Logistic analysis of regression model result.If desired instruct cable to purchase and construction, then need specifically to determine the extent of injury of each influence factor to cable fault.Such as cable body manufacturer Mi has been defined as cable fault influence factor, then need the failure rate PM continuing to compare the corresponding cable of cable body manufacturer Mi
i, PM
ibe worth larger, Mi is larger to cable fault influence degree, should select cable fault influence degree minimum cable body manufacturer during enterprise procurement, should pay close attention to the maximum cable body manufacturer of cable fault influence degree in operation maintenance.
Embodiment
The 10kV cable fault data that the cable fault data of the present embodiment provide for certain power supply administration, select cable body manufacturer, cable accessory manufacturer, unit in charge of construction and cable length to be influence factor, adopt the inventive method to analyze this four classes influence factor respectively to the influence degree of cable fault.
(1) Estimation of Sample Size.
According to data prediction principle, process is carried out to cable fault data and obtains sample data, in table 1.
Table 1 sample data
In table 1, the cable length scope affecting element L1, L2 corresponding is respectively 0 ~ 5km, 5 ~ 10km.
If influence factor to be analyzed is " cable body manufacturer ", to each concrete cable body manufacturer sample estimates amount successively.
Be described for cable body manufacturer M1:
Cable body manufacturer M1 is designated as A group, and non-cable body manufacturer M1 is designated as B group.According to one-sided Z
1-αthe inspection level of=0.05 and expection Z
1-βthe power of test of=80% estimates the sample size of M1.
(1) known according to table 1, the cable that cable body manufacturer M1 produces has 2, therefore cable count corresponding to M1 accounts for the ratio P=2/30=0.067 of cable sum.There is 1 fault occurred in 2 cables that M1 is corresponding, therefore the failure rate h of cable corresponding to M1
a=1/2=0.5.There are 5 fault occurred in 28 cables that non-M1 produces, therefore the failure rate h of cable corresponding to non-M1
b=5/28=0.179.Article 30, the cable count broken down in cable is 6, therefore cable fault rate d=6/30=0.2.
(2) by parameter h
a, h
bwith d for people's formula (2), obtain the sample size N=174 that M1 is corresponding.
Said method is adopted to calculate the sample size of other cable body manufacturers, the smallest sample amount of sample size maximal value and cable body manufacturer influence factor.In the present embodiment, the smallest sample amount N of cable body manufacturer influence factor
min=1008.Be influence factor to be analyzed with cable accessory manufacturer, unit in charge of construction, cable length respectively, calculate the smallest sample amount of each influence factor, in table 2.
The smallest sample amount of each influence factor of table 2
(2) Cox proportional hazard model is analyzed
1. PH test of hypothesis
For " cable body manufacturer " analysis of Influential Factors, when meeting table 2 sample size demand, selecting 1403 sample datas, in table 3, carrying out PH test of hypothesis based on this sample data.
The sample data of table 310kV cable body manufacturer
Covariant when constructing in SPSS, i.e. cable body manufacturer * lg (t), then in Cox is than risk model, analyze covariant " cable body manufacturer " and " cable body manufacturer * lg (t) ", SPSS directly can export PH test of hypothesis result, in table 4.
The PH test of hypothesis result of table 4 " cable body manufacturer "
In table 4, the estimated value of regression parameter β in the corresponding Cox proportional hazard model of B, SE represents the standard error of B, Wald represents the Wald statistic of regression parameter β, df represents degree of freedom, and Sig represents the P value that Wald checks, and Exp (B) represents Relative hazard.According to the PH method of inspection of Cox proportional hazard model, time covariant P value be 0.377>0.05, show covariant " cable body manufacturer " meet PH hypothesis.
In like manner respectively PH test of hypothesis is carried out to the cable length of the cable accessory manufacturer of 1113 cables, the unit in charge of construction of 4063 cables and 6443 cables, the time covariant P value of cable accessory manufacturer is 0.076>0.05, meet PH hypothesis, the time covariant P value of unit in charge of construction is 0.064>0.05, meet PH hypothesis, the time covariant P value of cable length is 0.085>0.05, meets PH hypothesis.
2. single analysis of Influential Factors
The test of hypothesis principle that what single analysis of Influential Factors adopted is in statistics.When the recurrence of the covariant in Cox scale model be parameter is 0, illustrate that this covariant and cable fault have nothing to do; When the regression parameter of covariant is non-zero, illustrate that this covariant is relevant with cable fault.In SPSS, solve the regression parameter of Cox proportional hazard model, null hypothesis is set: β
i=0, namely suppose that covariant and cable fault have nothing to do, checked by Wald, inspection null hypothesis, receive null hypothesis when P value is greater than 0.05, think that covariant and cable fault have nothing to do, otherwise covariant is relevant with cable fault.
The P value of covariant cable body manufacturer M5 is 0.019<0.05 as can be seen from Table 5, shows that cable body manufacturer M5 has a significant impact 10kV cable fault.When analyzing cable body manufacturer, select cable body manufacturer M5 as benchmark in SPSS, survived 4 dummy variable M1, M2, M3, M4.The P value of dummy variable cable body manufacturer M1 is less than 0.05, cable body manufacturer M1 and cable body manufacturer M5 is affecting significant difference to cable fault, and dummy variable cable body manufacturer M2, M3, M4 relative to cable body manufacturer M5 on cable fault impact on difference not remarkable.Consider relative risk, cable body manufacturer M5 is minimum on cable fault impact, should recommend to adopt, the relative risk Exp (B)=892.475 of cable body manufacturer M1, cable fault is had the greatest impact, should pay close attention in operation maintenance.
The results of univariate logistic analysis of table 5 " cable body manufacturer "
Known with reason table 6 ~ 8: 1) cable accessory manufacturer N5 is minimum on cable fault impact, should recommend to adopt, cable accessory manufacturer N1 has the greatest impact to cable fault, should pay close attention in operation maintenance.2) unit in charge of construction I3 has the greatest impact to cable fault, should pay close attention in operation maintenance, and unit in charge of construction I1 is minimum on cable fault impact, should recommend to adopt.3) cable length L3 has the greatest impact to cable fault, should pay close attention in operation maintenance, and cable length L2 is minimum on cable fault impact, should recommend to adopt.
The results of univariate logistic analysis of table 6 " cable accessory manufacturer "
The single factor analysis of table 7 " unit in charge of construction "
The single factor analysis of table 8 " cable length "
(3) cable fault list analysis of Influential Factors result verification
1. the influence factor of cable fault is confirmed based on Logistic regression model
Respectively cable body manufacturer, cable accessory manufacturer, unit in charge of construction, this four classes influence factor of cable length are included in separately in Logistic regression model, through SPSS software analysis, the results are shown in Table 9 ~ 12.
As shown in Table 9, cable body manufacturer is the variable in logistic regression model equation, its coefficient is-17.474, the saliency value sig=0.013<0.05 of wald statistic, illustrate that estimates of parameters is not 0 significantly, therefore cable body manufacturer is cable fault influence factor.Known with reason table 10 ~ 12, cable accessory manufacturer, unit in charge of construction, cable length are the influence factor of cable fault.
The Logistic analysis of regression model result of table 9 cable body manufacturer
Table 10 cable accessory manufacturer Logistic analysis of regression model result
Table 11 unit in charge of construction Logistic analysis of regression model result
Table 12 cable length Logistic analysis of regression model result
2. the cable fault rate that each influence factor is corresponding is calculated
Respectively cable fault rate corresponding to element is affected, the cable body manufacturer information of table 13 for collecting under having determined influence factor according to the calculating of cable fault data.
Table 1310kV cable body manufacturer tables of data
As shown in Table 13, the cable that cable body manufacturer M1 produces is 244, and the cable wherein broken down has 61, therefore cable fault rate corresponding to M1 is PM1=61/244=0.25, in like manner can calculate the corresponding cable fault rate of other cable body manufacturers, in table 14.
As shown in Table 14: the cable fault rate of M1 is maximum, the cable fault rate of M2, M3, M4, M5 is minimum, therefore recommends to select cable body manufacturer M2, M3, M4, M5, should pay close attention to cable body manufacturer M1 in operation maintenance.This result is identical with analysis result of the present invention.
In like manner can calculate each cable accessory manufacturer, unit in charge of construction, failure rate that cable length is corresponding respectively, in table 15 ~ 17.From table 15 ~ 17: the cable fault rate that cable accessory manufacturer N1 is corresponding is maximum, the cable fault rate that cable accessory manufacturer N5 is corresponding is minimum, therefore recommend to select cable accessory manufacturer N5, cable accessory manufacturer N1 should be paid close attention in operation maintenance.The cable fault rate that unit in charge of construction I3 is corresponding is maximum, and the cable fault rate that unit in charge of construction I1 is corresponding is minimum, therefore recommends to select unit in charge of construction I1, should pay close attention to unit in charge of construction I3 in operation maintenance.The cable fault rate that cable length L3 is corresponding is maximum, and the cable fault rate that cable length L2 is corresponding is minimum, therefore recommends to select cable length L2, should pay close attention to cable length L3 in operation maintenance.The analysis result of cable accessory manufacturer, unit in charge of construction, cable length is all identical with analysis result of the present invention.
The corresponding cable fault rate of table 14 cable body manufacturer
The corresponding cable fault rate of table 15 cable accessory manufacturer
The corresponding cable fault rate of table 16 cable construction manufacturer
The corresponding cable fault rate of table 17 cable length manufacturer
The present invention is successfully authenticated the correctness of the inventive method analysis result by logistic analysis of regression model.Notice, although integrated use logistic regression model and cable fault rate corresponding to statistical computation influence factor can analyze cable fault influence factor equally, but comparatively the present invention, analytical mathematics is more complicated, analytical procedure is various, and calculated amount is larger, particularly when the failure factor of Water demand increases, the increase of this calculated amount is more obvious, so can analyze cable fault influence factor by prioritizing selection Cox proportional hazard model during practical application.
Claims (3)
1., based on a Trouble ticket analysis of Influential Factors method for power cable fault information, it is characterized in that, comprise step:
Step 1, based on sample data sample estimates amount, this step comprises sub-step further:
1.1 pre-service cable fault data, obtain sample data, and before described sample data comprises the fault that 1. each cable is corresponding, the concrete of working time or truncated time, 2. malfunction, 3. influence factor and correspondence thereof affects element;
1.2 based on sample data, and for affecting element under each influence factor, influence factor i-th is affected element and is designated as A group, under this influence factor, other influences factor is designated as B group, according to N=(Z
1-α+ Z
1-γ)
2[dp (1-p) (ln Δ)
2]
-1calculate i-th and affect sample size N corresponding to element, the maximal value i.e. smallest sample amount of this influence factor in sample size; Wherein, d is the failure rate of cable, and the cable count namely broken down accounts for the ratio of cable sum; P affects the ratio that cable count corresponding to element accounts for cable sum in A group;
h
afor affecting the failure rate of cable corresponding to element in A group, namely in A group, affect the ratio that failure cable number corresponding to element accounts for the cable sum of its correspondence; h
bfor affecting the failure rate of cable corresponding to element in B group, namely in B group, affect the ratio that failure cable number corresponding to element accounts for the cable sum of its correspondence; Z
1-αfor inspection level; Z
1-γfor the power of test of expection; I=1,2 ... I, I are the picture elements number under influence factor;
Step 2, single analysis of Influential Factors, this step comprises further:
2.1 choose the sample data of current analysis of Influential Factors according to the smallest sample amount of current influence factor;
2.2 respectively affect element under current influence factor, adopt SPSS Software tool to carry out respectively:
A () is based on the sample data of current analysis of Influential Factors, the covariant of element for Cox proportional hazard model is affected with current, be 0 as null hypothesis using the current regression parameter affecting element, adopt association's variations per hour method to carry out PH test of hypothesis, judge that whether current to affect element relevant with cable fault according to the Wald value obtained; If relevant, perform sub-step (b); Otherwise, terminate;
B (), using the current element that affects as benchmark, generating dummy variable affects the relatively current relative risk affecting element of element, thus acquisition dummy variable affects element and the current size affecting element and affect cable fault.
2., as claimed in claim 1 based on the Trouble ticket analysis of Influential Factors method of power cable fault information, it is characterized in that:
Sub-step 1.1 is carried out for cable fault data, comprises further:
(a) rule of thumb with demand artificially selected CABLE MATERIALS fault influence factor and corresponding affect element;
B () carries out respectively for each influence factor: calculating respectively affects the ratio that cable count corresponding to element accounts for cable sum, and that deletes that this ratio is less than preset value affects cable fault data corresponding to element, and preset value sets according to actual requirement;
C (), to the cable broken down He do not break down, obtains working time and truncated time before fault respectively;
D the malfunction of () mark cable, cable fault state comprises the two states that do not break down and break down.
3., as claimed in claim 2 based on the Trouble ticket analysis of Influential Factors method of power cable fault information, it is characterized in that:
Described influence factor is one or more in cable body manufacturer, cable accessory manufacturer, unit in charge of construction, cable length.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510845900.0A CN105260863A (en) | 2015-11-26 | 2015-11-26 | Fault single influence factor analysis method based on power cable fault information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510845900.0A CN105260863A (en) | 2015-11-26 | 2015-11-26 | Fault single influence factor analysis method based on power cable fault information |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105260863A true CN105260863A (en) | 2016-01-20 |
Family
ID=55100542
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510845900.0A Pending CN105260863A (en) | 2015-11-26 | 2015-11-26 | Fault single influence factor analysis method based on power cable fault information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105260863A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106294982A (en) * | 2016-08-08 | 2017-01-04 | 国网江苏省电力公司徐州供电公司 | Cable fault virtual city 3D modelling system |
| WO2018036402A1 (en) * | 2016-08-26 | 2018-03-01 | 阿里巴巴集团控股有限公司 | Method and device for determining key variable in model |
| CN110503301A (en) * | 2019-07-17 | 2019-11-26 | 江苏大学 | A kind of electric car conduction charging Safety Influence Factors evaluation method based on Cox proportional hazard model |
| CN111199782A (en) * | 2019-12-30 | 2020-05-26 | 东软集团股份有限公司 | Etiology analysis method, etiology analysis device, storage medium, and electronic apparatus |
| CN111274546A (en) * | 2020-02-26 | 2020-06-12 | 广东电网有限责任公司 | Cable system fault analysis method |
| CN111611545A (en) * | 2020-05-18 | 2020-09-01 | 国网江苏省电力有限公司电力科学研究院 | Cable aging state evaluation method and device based on principal component analysis and logistic regression |
| CN114036450A (en) * | 2021-10-25 | 2022-02-11 | 中国电子科技集团公司第二十九研究所 | Method for quickly assessing testability indexes of electronic information equipment and storage medium |
| CN115249075A (en) * | 2022-09-22 | 2022-10-28 | 国网山西省电力公司太原供电公司 | Safe operation and maintenance management method and system for cable tunnel |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102207533A (en) * | 2011-04-12 | 2011-10-05 | 陕西化建工程有限责任公司 | Method for searching fault of power cable |
| CN103383769A (en) * | 2012-05-04 | 2013-11-06 | 山西省电力公司阳泉供电公司 | Grid fault hidden danger management analytical method and system |
| CN104655985A (en) * | 2015-01-30 | 2015-05-27 | 国家电网公司 | Power grid fault diagnosis method based on and-or tree model |
| CN104899690A (en) * | 2015-06-08 | 2015-09-09 | 江苏省电力公司苏州供电公司 | Power cable operation maintenance method based on historical fault information and operating state |
-
2015
- 2015-11-26 CN CN201510845900.0A patent/CN105260863A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102207533A (en) * | 2011-04-12 | 2011-10-05 | 陕西化建工程有限责任公司 | Method for searching fault of power cable |
| CN103383769A (en) * | 2012-05-04 | 2013-11-06 | 山西省电力公司阳泉供电公司 | Grid fault hidden danger management analytical method and system |
| CN104655985A (en) * | 2015-01-30 | 2015-05-27 | 国家电网公司 | Power grid fault diagnosis method based on and-or tree model |
| CN104899690A (en) * | 2015-06-08 | 2015-09-09 | 江苏省电力公司苏州供电公司 | Power cable operation maintenance method based on historical fault information and operating state |
Non-Patent Citations (3)
| Title |
|---|
| ZEYANG TANG,CHENGKE ZHOU: "Analysis of significant factors on cable failure using the Cox proportional hazard model", 《IEEE TRANSACTIONS ON POWER DELIVERY》 * |
| 徐英，骆福添: "COX 回归模型的样本含量的计算方法及软件实现", 《数理医药学杂志》 * |
| 方琼，栗薇: "基于SPSS 多元回归分析的中压配电网中性点接地方式研究", 《中国西部科技》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106294982A (en) * | 2016-08-08 | 2017-01-04 | 国网江苏省电力公司徐州供电公司 | Cable fault virtual city 3D modelling system |
| WO2018036402A1 (en) * | 2016-08-26 | 2018-03-01 | 阿里巴巴集团控股有限公司 | Method and device for determining key variable in model |
| TWI677830B (en) * | 2016-08-26 | 2019-11-21 | 香港商阿里巴巴集團服務有限公司 | Method and device for detecting key variables in a model |
| CN110503301A (en) * | 2019-07-17 | 2019-11-26 | 江苏大学 | A kind of electric car conduction charging Safety Influence Factors evaluation method based on Cox proportional hazard model |
| CN110503301B (en) * | 2019-07-17 | 2023-01-17 | 江苏大学 | Cox proportional risk model-based electric vehicle conduction charging safety influence factor evaluation method |
| CN111199782A (en) * | 2019-12-30 | 2020-05-26 | 东软集团股份有限公司 | Etiology analysis method, etiology analysis device, storage medium, and electronic apparatus |
| CN111199782B (en) * | 2019-12-30 | 2023-09-29 | 东软集团股份有限公司 | Etiology analysis method, device, storage medium and electronic equipment |
| CN111274546A (en) * | 2020-02-26 | 2020-06-12 | 广东电网有限责任公司 | Cable system fault analysis method |
| CN111611545A (en) * | 2020-05-18 | 2020-09-01 | 国网江苏省电力有限公司电力科学研究院 | Cable aging state evaluation method and device based on principal component analysis and logistic regression |
| CN114036450A (en) * | 2021-10-25 | 2022-02-11 | 中国电子科技集团公司第二十九研究所 | Method for quickly assessing testability indexes of electronic information equipment and storage medium |
| CN115249075A (en) * | 2022-09-22 | 2022-10-28 | 国网山西省电力公司太原供电公司 | Safe operation and maintenance management method and system for cable tunnel |
| CN115249075B (en) * | 2022-09-22 | 2022-12-06 | 国网山西省电力公司太原供电公司 | Safe operation and maintenance management method and system for cable tunnel |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105260863A (en) | Fault single influence factor analysis method based on power cable fault information | |
| Todorov et al. | Volatility jumps | |
| Kleibergen et al. | Non‐stationarity in garch models: A bayesian analysis | |
| Cavaliere | Limited time series with a unit root | |
| CN110333414A (en) | The multi-level state evaluating method of power transformer | |
| CN108596678A (en) | A kind of airline passenger value calculation method | |
| CN105046402A (en) | State evaluating method applied to secondary equipment of intelligent transformer station | |
| CN110137947B (en) | Grid voltage sag severity assessment method based on ITIC curve | |
| Kosapattarapim et al. | Evaluating the volatility forecasting performance of best fitting GARCH models in emerging Asian stock markets | |
| Figini et al. | Bayesian operational risk models | |
| Wong et al. | Money, interest rate, and stock prices: new evidence from Singapore and The United States | |
| Emenike Kalu | Weak-form efficiency after global financial crisis: Emerging stock market evidence | |
| Pham et al. | Foreign direct investment, exports and real exchange rate linkages in Vietnam: Evidence from a co-integration approach | |
| Gupta et al. | US monetary policy and BRICS stock market bubbles | |
| Mattera et al. | Option pricing under multifractional process and long-range dependence | |
| Hussain et al. | Dynamic linkages of exchange rate and stock return volatility evidence from Pakistan, India and China (PIC) | |
| CN107862476A (en) | A kind of metering table demand computational methods based on data analysis | |
| CN102982231A (en) | Quantitative calculation method for software confidence level | |
| Platen et al. | Dynamics of a Well-Diversified Equity Index | |
| CN103412814A (en) | Mobile terminal system safety test and intelligent repair system and method | |
| Lerch et al. | Verification of probabilistic forecasts for rare and extreme events | |
| Steland | Random walks with drift–a sequential approach | |
| DE102010044724A1 (en) | Method for improved loss energy forecasting in a transmission network | |
| Witlox et al. | The impact of risk model quality on portfolio Expected Shortfall | |
| Goldobina et al. | Construction of Adapted Three-Level Model of Control Processes of Information Security System of Critical Information Infrastructure Objects |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160120 |