CN105303296A - Electric power equipment full-life state evaluation method - Google Patents

Abstract

The invention discloses an electric power equipment full-life state evaluation method. The electric power equipment full-life state evaluation method comprises steps that, a defective text of electric power equipment is pre-processed by utilizing the nature language processing NLP technology to acquire a characteristic vector of the detective text; similarity between the defective text and a preset training set is calculated according to the characteristic vector; a defect grade of the defective text is calculated according to the similarity, the defect grade is converted into a first health state index HItext, and the first health state index HItext is taken as a target health state index HI; state fusion processing on the target health state index HI is carried out to acquire a result which is taken as a unit health period of the electric power equipment at the present time; the unit health period of the electric power equipment at the present time and all previous unit health periods are integrated to acquired full-life state evaluation information of the electric power equipment. Through the electric power equipment full-life state evaluation method, mining for a lot of effective information contained in the defective text is realized, and accuracy and comprehensiveness of electric power equipment full-life state evaluation are improved.

Description

A kind of power equipment life-cycle method for evaluating state

Technical field

The present invention relates to technical field of power systems, particularly relate to a kind of power equipment life-cycle method for evaluating state.

Background technology

In power industry, in order to effectively grasp health status and the service life state of power equipment, staff can evaluate the health status of power equipment usually.

In electrical network, the vital electrical equipment of minority is had to be assembled with state monitoring apparatus, as the oil chromatogram analysis instrument of transformer, partial discharge monitoring instrument etc.The status monitoring amount that these state monitoring apparatus obtain can reflect some specific characteristic quantity of this power equipment in real time.But, on the one hand due to the reason such as financial cost and technical development, many power equipments also do not have confined state monitoring device, in the future for a long period of time, manual inspection and live detection will be much equipment topmost " monitoring " modes, realize the life-cycle state evaluation to power equipment with this, but there is the low shortcoming of accuracy in this evaluation method.On the other hand, even if be assembled with state monitoring apparatus, such as certain class monitoring sensor, also the key feature amount that this kind of monitoring sensor is corresponding can only be reflected, but for the detection of such monitoring sensor less than and the stronger inferior health defect of ambiguity, still need to rely on manual inspection, adopt text report defect record to be got off as much as possible, to realize the omnibearing life-cycle state evaluation of this power equipment, this evaluation method also exists the shortcoming of the low and one-sidedness of accuracy equally.

Therefore, the problem that the accuracy of power equipment life-cycle state evaluation and the comprehensive those skilled in the art of being need to solve at present how is effectively improved.

Summary of the invention

The object of this invention is to provide a kind of power equipment life-cycle method for evaluating state, achieve the excavation to a large amount of objective effective information contained in defect text, make full use of the life-cycle state of these effective informations to power equipment to evaluate, improve the accuracy of power equipment life-cycle state evaluation and comprehensive.

For solving the problems of the technologies described above, the invention provides a kind of power equipment life-cycle method for evaluating state, comprising:

Use the defect text of natural language processing NLP technology to power equipment to carry out pre-service, obtain the proper vector of described defect text;

The similarity between described defect text and default training set is calculated according to described proper vector;

According to the defect rank of defect text described in described Similarity Measure, and described defect rank is converted to the first health status index HI _text, by described first health status index HI _textas target health status index HI;

Carry out state fusion process to described target health status index HI, the result obtained is as the unit healthy cycle of power equipment described in this;

By described power equipment this unit healthy cycle and the units all before healthy cycle integrate, obtain the life-cycle state evaluation information of described power equipment.

Preferably, the method also comprises:

Obtain condition monitoring device and described power equipment is monitored to the Condition Monitoring Data obtained, feature extraction is carried out to described Condition Monitoring Data, obtains comprehensive characteristics index L _signal, and by described comprehensive characteristics index L _signalbe converted to the second health status index HI _signal;

By described first health status index HI _textwith described second health status index HI _signalbe fused into general health index HI _always, wherein, described HI _always={ HI _text, HI _signal, by described general health index HI _alwaysas described target health status index HI.

Preferably, described default training set is vector space W _all;

Wherein, described W _all=[w _ab] _a*B, wherein, w _ab=0 or 1, A be the quantity of the defect text of described power equipment, B is the dimension of the proper vector of the defect text of described power equipment.

Preferably, the defect text of described utilization natural language processing NLP technology to power equipment carries out pre-service, and the process obtaining the proper vector of described defect text is:

Word segmentation processing is carried out to the defect text of described power equipment, obtains participle;

Word frequency statistics process is carried out to described participle, and accordingly described participle is sorted according to word frequency order from big to small, obtain the first word sequence;

Stop words process is gone to described first word sequence, obtains the second word sequence;

Text vector is carried out to described second word sequence, obtains the proper vector of described defect text.

Preferably, the process of the described similarity calculated between described defect text and default training set according to described proper vector is:

Employing k nearest neighbour classification algorithm kNN calculates the similarity S between described defect text and described default training set respectively _h, wherein, h=1,2 ... A;

Described $S_h=\frac{\underset{l=1}{\overset{B}{\Σ}}{w}_l\×w_{hl}}{\sqrt{\underset{l=1}{\overset{B}{\Σ}}{w}_l^2}\sqrt{\underset{l=1}{\overset{B}{\Σ}}{w}_{hl}^2}};$

S _hfor the similarity between h training set text in described defect text and described default training set;

W is the proper vector of described defect text; w _hfor the proper vector of h training set text in described default training set; w _lfor the l dimension value of w, w _hlfor w _hl dimension value.

Preferably, the described defect rank according to defect text described in described Similarity Measure, and described defect rank is converted to the first health status index HI _textprocess be:

According to described similarity S _hnumerical values recited to described similarity S _hsort, select front k and be worth maximum described similarity S _hand the training set text preset described in the most similar corresponding k bar in training set;

Maximum described similarity S is worth according to described front k _hand the training set text preset in training set calculates the defect rank L of described defect text described in the most similar corresponding k bar _text;

Wherein, described in $L_{text}=\frac{\underset{m=1}{\overset{k}{\Σ}}{S}_m{L}_m}{\underset{m=1}{\overset{k}{\Σ}}{S}_m};$

Wherein, described S _mfor presetting the similarity between the training set text in training set described in described defect text and m article; L _mfor the defect rank of m bar of training set text in k before given in advance described default training set the most similar;

By the first health status conversion relational expression, described defect rank is converted to the first health status index HI _text;

Wherein, described first health status conversion relational expression

Described L _tmaxfor the upper limit of the defect rank of described power equipment given in advance, described L _tminfor the lower limit of the defect rank of described power equipment given in advance.

Preferably, described acquisition condition monitoring device monitors the Condition Monitoring Data obtained to described power equipment, carry out feature extraction to described Condition Monitoring Data, obtains comprehensive characteristics index L _signal, and by described comprehensive characteristics index L _signalbe converted to the second health status index HI _signalprocess be:

Obtain condition monitoring device and described power equipment is monitored to the Condition Monitoring Data obtained, carry out feature extraction, obtain n characteristic signal index to described Condition Monitoring Data, wherein, n described characteristic signal index includes time-frequency index and normal index;

According to Minkowski distance relation formula, n described characteristic signal index is changed, obtain described comprehensive characteristics index L _signal;

Wherein, described comprehensive characteristics index $L_{signa1}=\sqrt[r]{\underset{q=1}{\overset{n}{\Σ}}|S_{abnormalq}-S_{normalq}{|}^r};$

R=1 or 2 or ∞, S _abnormalqbe the time-frequency index of q described characteristic signal, S _normalqit is the normal index of q described characteristic signal;

By the second health status conversion relational expression by comprehensive characteristics index L _signalbe converted to the second health status index HI _signal;

Wherein, described second health status conversion relational expression

Described L _smaxfor the upper limit of the comprehensive characteristics index of described electronic equipment given in advance, described L _sminfor the lower limit of the comprehensive characteristics index of described electronic equipment given in advance.

Preferably, describedly carry out state fusion process to described target health status index HI, the result obtained as the process in the unit healthy cycle of power equipment described in this is:

Suppose that the unit of power equipment described in this healthy cycle is i-th unit healthy cycle of described power equipment, by healthy for described i-th unit cycle use represent; Wherein, described in obtain respectively by Ratio-type state fusion relationship model formula;

Described Ratio-type state fusion relationship model formula is:

$\left\{\begin{array}{c}{\stackrel{*}{HI}}_i={\mathrm{HI}}_i*\exp \[\mathrm{\δ}\left({\mathrm{age}}_i\right)*({\stackrel{*}{HI}}_{i-1}-1)\]\\ {\stackrel{*}{HI}}_{i+1}={\stackrel{*}{HI}}_i*\exp \[\mathrm{\δ}\left({\mathrm{age}}_{i+1}\right)*({\stackrel{*}{HI}}_i-1)\]\end{array}\right.$

Wherein, described in for comprehensive state evaluation result when described defect text starts pre-service;

Described HI _ifor described first health status index or described second health status index;

Described δ (age _i) be the indicative function of the enlistment age of described power equipment, wherein,

Described AGE=100.

A kind of power equipment life-cycle method for evaluating state provided by the invention, carrying out pre-service by using the defect text of natural language processing NLP technology to power equipment, calculating similarity between defect text and default training set; According to the defect rank of Similarity Measure defect text, and defect rank is converted to the first health status index HI _text; By the first health status index HI _textas target health status index HI; State fusion process is carried out to target health status index HI, obtains the unit healthy cycle of this power equipment and then obtain the life-cycle state evaluation information of power equipment.

Visible, the method is based on the defect text of power equipment, by a series of process to defect text, try to achieve the unit healthy cycle of power equipment, finally obtain the life-cycle state evaluation information of power equipment, achieve the excavation to a large amount of objective effective information contained in defect text, make full use of the life-cycle state of these effective informations to power equipment and evaluate, improve the accuracy of power equipment life-cycle state evaluation and comprehensive.

Accompanying drawing explanation

In order to be illustrated more clearly in the technical scheme in the embodiment of the present invention, be briefly described to the accompanying drawing used required in prior art and embodiment below, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.

Fig. 1 is the process flow diagram of the process of a kind of power equipment life-cycle method for evaluating state provided by the invention;

Fig. 2 is the process flow diagram of the process of another kind of power equipment life-cycle method for evaluating state provided by the invention;

Fig. 3 is the process flow diagram that a kind of defect text to power equipment provided by the invention carries out pretreated process;

Fig. 4 is a kind of process flow diagram defect rank being converted to the process of the first health status index provided by the invention;

Fig. 5 is the process flow diagram of the process of a kind of acquisition provided by the invention second health status index;

Fig. 6 is the schematic diagram of the life-cycle state evaluation information of a kind of power equipment provided by the invention.

Embodiment

Core of the present invention is to provide a kind of power equipment life-cycle method for evaluating state, achieve the excavation to a large amount of objective effective information contained in defect text, make full use of the life-cycle state of these effective informations to power equipment to evaluate, improve the accuracy of power equipment life-cycle state evaluation and comprehensive.

For making the object of the embodiment of the present invention, technical scheme and advantage clearly, below in conjunction with the accompanying drawing in the embodiment of the present invention, technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.

Embodiment one

Please refer to Fig. 1, Fig. 1 is the process flow diagram of the process of a kind of power equipment life-cycle method for evaluating state provided by the invention; The method comprises:

Step s101: use the defect text of natural language processing NLP technology to power equipment to carry out pre-service, obtain the proper vector of defect text;

Step s102: calculate the similarity between defect text and default training set according to proper vector;

Step s103: according to the defect rank of Similarity Measure defect text, and defect rank is converted to the first health status index HI _text, by the first health status index HI _textas target health status index HI;

Step s104: state fusion process is carried out to target health status index HI, the result obtained is as the unit healthy cycle of this power equipment;

Step s105: by power equipment this unit healthy cycle and the units all before healthy cycle integrate, obtain the life-cycle state evaluation information of power equipment.

A kind of power equipment life-cycle method for evaluating state provided by the invention, carrying out pre-service by using the defect text of natural language processing NLP technology to power equipment, calculating similarity between defect text and default training set; According to the defect rank of Similarity Measure defect text, and defect rank is converted to the first health status index HI _text; By the first health status index HI _textas target health status index HI; State fusion process is carried out to target health status index HI, obtains the unit healthy cycle of this power equipment and then obtain the life-cycle state evaluation information of power equipment.

Visible, the method is based on the defect text of power equipment, by a series of process to defect text, try to achieve the unit healthy cycle of power equipment, finally obtain the life-cycle state evaluation information of power equipment, achieve the excavation to a large amount of objective effective information contained in defect text, make full use of the life-cycle state of these effective informations to power equipment and evaluate, improve the accuracy of power equipment life-cycle state evaluation and comprehensive.

Embodiment two

Please refer to Fig. 2, Fig. 2 is the process flow diagram of the process of another kind of power equipment life-cycle method for evaluating state provided by the invention; The method comprises:

Step s201: use the defect text of natural language processing NLP technology to power equipment to carry out pre-service, obtain the proper vector of defect text;

As preferably, please refer to Fig. 3, Fig. 3 is the process flow diagram that a kind of defect text to power equipment provided by the invention carries out pretreated process; This step is specially:

Step s2011: word segmentation processing is carried out to the defect text of power equipment, obtains participle;

Be understandable that, the defect text of natural language processing NLP (NaturalLanguageProcessing) technology to power equipment is used to carry out pre-service, first based on " electrical equipment fault and the basic dictionary of defect " that writing is in advance good, set up Markov model (HiddenMarkovModel, HMM) word segmentation processing is carried out to defect text, obtain participle.

Step s2012: word frequency statistics process is carried out to participle, and accordingly participle is sorted according to word frequency order from big to small, obtain the first word sequence;

Step s2013: go stop words process to the first word sequence, obtains the second word sequence;

Be understandable that, after obtaining the first word sequence, according to " dictionary of stopping using " established in advance with reference to expertise, stop words process gone to the first word sequence, obtain the second word sequence.

Step s2014: carry out text vector to the second word sequence, obtains the proper vector of defect text.

Step s202: calculate the similarity between defect text and default training set according to proper vector;

As preferably, default training set is vector space W _all;

Wherein, W _all=[w _ab] _a*B, wherein, w _ab=0 or 1, A be the quantity of the defect text of power equipment, B is the dimension of the proper vector of the defect text of power equipment.

Be understandable that, default training set here is also vector space W _allit is the defect text collection sorted out according to expertise.

As preferably, the process of the similarity calculated between defect text and default training set according to proper vector is:

Employing k nearest neighbour classification algorithm kNN (k-NearestNeighbor) calculates the similarity S between defect text and default training set respectively _h, wherein, h=1,2 ... A;

$S_h=\frac{\underset{l=1}{\overset{B}{\Σ}}{w}_l\×w_{hl}}{\sqrt{\underset{l=1}{\overset{B}{\Σ}}{w}_l^2}\sqrt{\underset{l=1}{\overset{B}{\Σ}}{w}_{hl}^2}};$

S _hfor the similarity between h training set text in defect text and default training set;

W is the proper vector of defect text; w _hfor the proper vector of h training set text in default training set; w _lfor the l dimension value of w, w _hlfor w _hl dimension value.

Step s203: according to the defect rank of Similarity Measure defect text, and defect rank is converted to the first health status index HI _text;

As preferably, please refer to Fig. 4, Fig. 4 is a kind of process flow diagram defect rank being converted to the process of the first health status index provided by the invention;

According to the defect rank of Similarity Measure defect text, and defect rank is converted to the first health status index HI _textprocess be:

Step s2031: according to similarity S _hnumerical values recited to similarity S _hsort, select front k and be worth maximum similarity S _hand the most similar corresponding k bar presets the training set text in training set;

Be understandable that, k is here the optimum k value chosen in advance, is described below to optimum choosing of k value:

Select the test set sorted out according to expertise and training set respectively, adopt kNN algorithm to calculate in test set similarity between test set text x and training set text y;

$S_{xv}=\frac{\underset{j=1}{\overset{M}{\Σ}}{w}_{xj}\×w_{yj}}{\sqrt{\underset{i=1}{\overset{M}{\Σ}}{w}_{xj}^2}\sqrt{\underset{j=1}{\overset{M}{\Σ}}{w}_{yj}^2}};$

Wherein, described S _xyfor test set text x in test set and similarity between training set text y;

W _xfor the proper vector of test set text x in test set; w _yfor the proper vector of the training set text y in default training set; w _xjfor w _xjth dimension value, w _yjfor w _yjth dimension value.

According to similarity S _xynumerical values recited to similarity S _xysort, select front k and be worth maximum similarity S _xyand the training set text in the most similar corresponding k bar training set;

Then maximum similarity S is worth based on front k _xyand the training set text in the most similar corresponding k bar training set calculates the defect rank L of test set text x _{text_x};

Wherein, $L_{text-x}=\frac{\underset{c=1}{\overset{k}{\Σ}}{S}_c{L}_c}{\underset{c=1}{\overset{k}{\Σ}}{S}_c};$

Wherein, S _cfor the similarity between the training set text in test set text x and c article of training set; L _cfor the defect rank of c bar of training set text in k before given in advance training set the most similar; In the application, the higher description defect of defect rank is more serious.

Same, the default defect rank that test set text x also has expert group rule of thumb given in advance accordingly, by the L calculated _text-xcompare with default defect rank given in advance, when the difference when is between the two less than 0.5 (supposing that grade pitch is 1), then classification is correct.

Due to the training set of different capabilities, k value possibility when test set text x classification accuracy is the highest is also inconsistent.In tradition kNN method, k value is by artificial subjective selected, and the application proposes to adopt sensitivity algorithm to carry out the optimizing of k value, by setting different k values, repeats above-mentioned calculating L successively to each k value _text-xprocess, and calculate the classification accuracy rate of test set text x under different value of K, the k value (L also now calculated when selection sort accuracy is the highest _text-xcompare with default defect rank given in advance, difference is between the two minimum), now using this k value of correspondence as optimum k value.

Step s2032: be worth maximum similarity S according to front k _hand the most similar corresponding k bar training set text preset in training set calculates the defect rank L of defect text _text;

Wherein, $L_{text}=\frac{\underset{m=1}{\overset{k}{\Σ}}{S}_m{L}_m}{\underset{m=1}{\overset{k}{\Σ}}{S}_m};$

Wherein, S _mthe similarity between the training set text in training set is preset for defect text and m article; L _mfor the defect rank of m bar of training set text in k before given in advance default training set the most similar;

Be understandable that, L here _textfor the evaluation rank of the defect of power equipment.

Step s2033: defect rank is converted to the first health status index HI by the first health status conversion relational expression _text;

Wherein, the first health status conversion relational expression

L _tmaxfor the upper limit of the defect rank of power equipment given in advance, L _tminfor the lower limit of the defect rank of power equipment given in advance.

Here, we define standardized health status index HI _standard, its scope is [0,1], wherein, and HI _standard=0 represents electrical equipment fault, HI _standard=1 to represent power equipment completely healthy.

Step s204: obtain condition monitoring device and power equipment is monitored to the Condition Monitoring Data obtained, feature extraction is carried out to Condition Monitoring Data, obtains comprehensive characteristics index L _signal, and by comprehensive characteristics index L _signalbe converted to the second health status index HI _signal;

As preferably, please refer to Fig. 5, Fig. 5 is the process flow diagram of the process of a kind of acquisition provided by the invention second health status index;

Obtain condition monitoring device and power equipment is monitored to the Condition Monitoring Data obtained, feature extraction is carried out to Condition Monitoring Data, obtains comprehensive characteristics index L _signal, and by comprehensive characteristics index L _signalbe converted to the second health status index HI _signalprocess be:

Step s2041: obtain condition monitoring device and power equipment is monitored to the Condition Monitoring Data obtained, carry out feature extraction to Condition Monitoring Data, obtain n characteristic signal index, wherein, n characteristic signal index includes time-frequency index and normal index;

Step s2042: according to Minkowski distance relation formula, n characteristic signal index is changed, obtain comprehensive characteristics index L _signal;

Wherein, comprehensive characteristics index $L_{signa1}=\sqrt[r]{\underset{q=1}{\overset{n}{\Σ}}|S_{abnormalq}-S_{normalq}{|}^r};$

R=1 or 2 or ∞, S _abnormalqbe the time-frequency index of q characteristic signal, S _normalqit is the normal index of q characteristic signal;

Be understandable that, r here in different situations, chooses different values.

Such as, when being simple linear relationship between Condition Monitoring Data, r=1, now L _signalcorresponding to CityBlock distance.

When Condition Monitoring Data is the data such as frequency, voltage, r=2, now L _signalcorresponding to Euclidean distance.

When Condition Monitoring Data is very discrete, r=∞, now L _signalcorresponding to Chebychev distance.

Step s2043: by the second health status conversion relational expression by comprehensive characteristics index L _signalbe converted to the second health status index HI _signal;

Wherein, the second health status conversion relational expression

L _smaxfor the upper limit of the comprehensive characteristics index of electronic equipment given in advance, L _sminfor the lower limit of the comprehensive characteristics index of electronic equipment given in advance.

Step s205: by the first health status index HI _textwith the second health status index HI _signalbe fused into general health index HI _always, wherein, HI _always={ HI _text, HI _signal, by general health index HI _alwaysas target health status index HI;

Step s206: state fusion process is carried out to target health status index HI, the result obtained is as the unit healthy cycle of this power equipment;

Be understandable that, here the unit healthy cycle refer to this power equipment from coming into operation (comprise initially put into operation, overhaul/defect elimination or transformation and upgrade complete after put into operation), occur (disfigurement discovery or status monitoring feature extraction find) to event, then running due to band defect causes event degree to be upgraded, finally to taking measures, the time of whole process that event terminates (measure comprise maintenance, defect elimination, replacing, retired etc.) this whole process.

As preferably, carry out state fusion process to target health status index HI, the result obtained as the process in the unit healthy cycle of this power equipment is:

Suppose that the unit of this power equipment healthy cycle is i-th unit healthy cycle of power equipment, healthy for i-th unit cycle is used represent; Wherein, obtain respectively by Ratio-type state fusion relationship model formula;

Ratio-type state fusion relationship model formula is:

$\left\{\begin{array}{c}{\stackrel{*}{HI}}_i={\mathrm{HI}}_i*\exp \[\mathrm{\δ}\left({\mathrm{age}}_i\right)*({\stackrel{*}{HI}}_{i-1}-1)\]\\ {\stackrel{*}{HI}}_{i+1}={\stackrel{*}{HI}}_i*\exp \[\mathrm{\δ}\left({\mathrm{age}}_{i+1}\right)*({\stackrel{*}{HI}}_i-1)\]\end{array}\right.$

Wherein, for comprehensive state evaluation result when defect text starts pre-service;

HI _ibe the first health status index or the second health status index;

Be understandable that, here HI _ispecifically get the first health status index or the second health status index is determined according to concrete actual conditions, some power equipment such as, do not have state monitoring apparatus, then now there is no the second health status index, now HI _ithe first health status index can only be got; When power equipment is provided with state monitoring apparatus, then now have the second health status index, now staff can decide HI according to actual conditions _iget the first health status index or the second health status index.

δ (age _i) be the indicative function of the enlistment age of power equipment, wherein,

AGE＝100。

Step s207: by power equipment this unit healthy cycle and the units all before healthy cycle integrate, obtain the life-cycle state evaluation information of power equipment.

Be understandable that, the life-cycle state evaluation information of power equipment here by power equipment this unit healthy cycle and the units all before healthy cycle form.

Please refer to Fig. 6, Fig. 6 is the schematic diagram of the life-cycle state evaluation information of a kind of power equipment provided by the invention.

I-th healthy cycle SHC of unit _iarrangement set is:

${\mathrm{SHC}}_i=\{({\mathrm{age}}_{(i-1)+},{\mathrm{HI}}_{(i-1)+}^{*});TTE;({\mathrm{age}}_i,{\stackrel{*}{HI}}_i);Delay;({\mathrm{age}}_{(i+1)},\stackrel{*}{{\mathrm{HI}}_{i+1}})\};$

Namely this power equipment is in enlistment age age _ishi Fasheng defect event, defect text through feature extraction through kNN classification or Monitoring Data, then obtains health index and is after merging after band defect runs the Delay time, in age _i+1moment takes defect elimination or maintenance measure M _i+1.

The present embodiment is on the basis of embodiment one, by with adopting NLP technology that defect text is carried out to the first health status index that a series of process obtains and Condition Monitoring Data carried out to the second health status index that a series of process obtains and merge, obtain general health index, according to Ratio-type state fusion relationship model formula, general health index is processed again, obtain this unit healthy cycle of power equipment, and then finally obtain the life-cycle state evaluation information of power equipment.

Visible, the equipment health status evaluation result that the present embodiment will be extract by the text mining of electrical equipment fault defect, merge with the evaluation result based on Condition Monitoring Data, realize the health status historical process of power equipment and complete the presenting of life-cycle state evaluation information, further increase the accuracy of power equipment life-cycle state evaluation and comprehensive.

In this instructions, each embodiment adopts the mode of going forward one by one to describe, and what each embodiment stressed is the difference with other embodiments, between each embodiment identical similar portion mutually see.

Also it should be noted that, in this manual, the such as relational terms of first and second grades and so on is only used for an entity or operation to separate with another entity or operational zone, and not necessarily requires or imply the relation that there is any this reality between these entities or operation or sequentially.And, term " comprises ", " comprising " or its any other variant are intended to contain comprising of nonexcludability, thus make to comprise the process of a series of key element, method, article or equipment and not only comprise those key elements, but also comprise other key elements clearly do not listed, or also comprise by the intrinsic key element of this process, method, article or equipment.When not more restrictions, the key element limited by statement " comprising ... ", and be not precluded within process, method, article or the equipment comprising described key element and also there is other identical element.

Professional can also recognize further, in conjunction with unit and the algorithm steps of each example of embodiment disclosed herein description, can realize with electronic hardware, computer software or the combination of the two, in order to the interchangeability of hardware and software is clearly described, generally describe composition and the step of each example in the above description according to function.These functions perform with hardware or software mode actually, depend on application-specific and the design constraint of technical scheme.Professional and technical personnel can use distinct methods to realize described function to each specifically should being used for, but this realization should not thought and exceeds scope of the present invention.

To the above-mentioned explanation of the disclosed embodiments, professional and technical personnel in the field are realized or uses the present invention.To be apparent for those skilled in the art to the multiple amendment of these embodiments, General Principle as defined herein can without departing from the spirit or scope of the present invention, realize in other embodiments.Therefore, the present invention can not be restricted to these embodiments shown in this article, but will meet the widest scope consistent with principle disclosed herein and features of novelty.

Claims (8)

1. a power equipment life-cycle method for evaluating state, is characterized in that, comprising:

Use the defect text of natural language processing NLP technology to power equipment to carry out pre-service, obtain the proper vector of described defect text;

The similarity between described defect text and default training set is calculated according to described proper vector;

According to the defect rank of defect text described in described Similarity Measure, and described defect rank is converted to the first health status index HI _text, by described first health status index HI _textas target health status index HI;

Carry out state fusion process to described target health status index HI, the result obtained is as the unit healthy cycle of power equipment described in this;

By described power equipment this unit healthy cycle and the units all before healthy cycle integrate, obtain the life-cycle state evaluation information of described power equipment.

2. power equipment life-cycle method for evaluating state as claimed in claim 1, it is characterized in that, the method also comprises:

Obtain condition monitoring device and described power equipment is monitored to the Condition Monitoring Data obtained, feature extraction is carried out to described Condition Monitoring Data, obtains comprehensive characteristics index L _signal, and by described comprehensive characteristics index L _signalbe converted to the second health status index HI _signal;

By described first health status index HI _textwith described second health status index HI _signalbe fused into general health index HI _always, wherein, described HI _always={ HI _text, HI _signal, by described general health index HI _alwaysas described target health status index HI.

3. power equipment life-cycle method for evaluating state as claimed in claim 2, it is characterized in that, described default training set is vector space W _all;

Wherein, described W _all=[w _ab] _a*B, wherein, w _ab=0 or 1, A be the quantity of the defect text of described power equipment, B is the dimension of the proper vector of the defect text of described power equipment.

4. power equipment life-cycle method for evaluating state as claimed in claim 3, it is characterized in that, the defect text of described utilization natural language processing NLP technology to power equipment carries out pre-service, and the process obtaining the proper vector of described defect text is:

Word segmentation processing is carried out to the defect text of described power equipment, obtains participle;

Word frequency statistics process is carried out to described participle, and accordingly described participle is sorted according to word frequency order from big to small, obtain the first word sequence;

Stop words process is gone to described first word sequence, obtains the second word sequence;

Text vector is carried out to described second word sequence, obtains the proper vector of described defect text.

5. power equipment life-cycle method for evaluating state as claimed in claim 4, is characterized in that, the process of the described similarity calculated between described defect text and default training set according to described proper vector is:

Employing k nearest neighbour classification algorithm kNN calculates the similarity S between described defect text and described default training set respectively _h, wherein, h=1,2 ... A;

Described $S_h=\frac{\underset{l=1}{\overset{B}{\Σ}}{w}_l\×w_{hl}}{\sqrt{\underset{l=1}{\overset{B}{\Σ}}{w}_l^2}\sqrt{\underset{l=1}{\overset{B}{\Σ}}{w}_{hl}^2}};$

S _hfor the similarity between h training set text in described defect text and described default training set;

W is the proper vector of described defect text; w _hfor the proper vector of h training set text in described default training set; w _lfor the l dimension value of w, w _hlfor w _hl dimension value.

6. power equipment life-cycle method for evaluating state as claimed in claim 5, is characterized in that, the described defect rank according to defect text described in described Similarity Measure, and described defect rank is converted to the first health status index HI _textprocess be:

According to described similarity S _hnumerical values recited to described similarity S _hsort, select front k and be worth maximum described similarity S _hand the training set text preset described in the most similar corresponding k bar in training set;

Maximum described similarity S is worth according to described front k _hand the training set text preset in training set calculates the defect rank L of described defect text described in the most similar corresponding k bar _text;

Wherein, described in $L_{text}=\frac{\underset{m=1}{\overset{k}{\Σ}}{S}_m{L}_m}{\underset{m=1}{\overset{k}{\Σ}}{S}_m};$

Wherein, described S _mfor presetting the similarity between the training set text in training set described in described defect text and m article; L _mfor the defect rank of m bar of training set text in k before given in advance described default training set the most similar;

By the first health status conversion relational expression, described defect rank is converted to the first health status index HI _text;

Wherein, described first health status conversion relational expression

Described L _tmaxfor the upper limit of the defect rank of described power equipment given in advance, described L _tminfor the lower limit of the defect rank of described power equipment given in advance.

7. power equipment life-cycle method for evaluating state as claimed in claim 6, it is characterized in that, described acquisition condition monitoring device monitors the Condition Monitoring Data obtained to described power equipment, carry out feature extraction to described Condition Monitoring Data, obtains comprehensive characteristics index L _signal, and by described comprehensive characteristics index L _signalbe converted to the second health status index HI _signalprocess be:

Obtain condition monitoring device and described power equipment is monitored to the Condition Monitoring Data obtained, carry out feature extraction, obtain n characteristic signal index to described Condition Monitoring Data, wherein, n described characteristic signal index includes time-frequency index and normal index;

According to Minkowski distance relation formula, n described characteristic signal index is changed, obtain described comprehensive characteristics index L _signal;

Wherein, described comprehensive characteristics index $L_{signal}=\sqrt[r]{\underset{q=1}{\overset{n}{\Σ}}|S_{abnormalq}-S_{normalq}{|}^r};$

R=1 or 2 or ∞, S _abnormalqbe the time-frequency index of q described characteristic signal, S _normalqit is the normal index of q described characteristic signal;

By the second health status conversion relational expression by comprehensive characteristics index L _signalbe converted to the second health status index HI _signal;

Wherein, described second health status conversion relational expression

Described L _smaxfor the upper limit of the comprehensive characteristics index of described electronic equipment given in advance, described L _sminfor the lower limit of the comprehensive characteristics index of described electronic equipment given in advance.

8. the power equipment life-cycle method for evaluating state as described in any one of claim 1-7, it is characterized in that, describedly carry out state fusion process to described target health status index HI, the result obtained as the process in the unit healthy cycle of power equipment described in this is:

Suppose that the unit of power equipment described in this healthy cycle is i-th unit healthy cycle of described power equipment, by healthy for described i-th unit cycle use represent; Wherein, described in obtain respectively by Ratio-type state fusion relationship model formula;

Described Ratio-type state fusion relationship model formula is:

$\left\{\begin{array}{c}{\stackrel{*}{HI}}_i=H{I}_i*\exp \[\mathrm{\δ}\left({\mathrm{age}}_i\right)*({\stackrel{*}{HI}}_{i-1}-1)\]\\ {\stackrel{*}{HI}}_{i+1}={\stackrel{*}{HI}}_i*\exp \[\mathrm{\δ}\left({\mathrm{age}}_{i+1}\right)*({\stackrel{*}{HI}}_i-1)\]\end{array}\right.$

Wherein, described in for comprehensive state evaluation result when described defect text starts pre-service;

Described HI _ifor described first health status index or described second health status index;

Described δ (age _i) be the indicative function of the enlistment age of described power equipment, wherein,

Described AGE=100.