CN103473450A - Attribute interval identification method for intelligent power distribution network efficiency evaluation - Google Patents

Abstract

The invention relates to an attribute interval identification method for intelligent power distribution network efficiency evaluation. A three-layer attribute measuring matrix between intelligent power distribution network basic data and efficiency is built, intelligent power distribution network efficiency attribute intervals are identified automatically on the basis of index data standardization, and attribute measuring interval and comprehensive attribute measurement value of maximum entropy principle calculating performance index, standard index and comprehensive index. The method includes the steps of firstly, reading information required by the evaluation; secondly, building an attribute measuring matrix of comprehensive index, standard index and performance index; thirdly, standardizing index data; fourthly calculating attribute measuring intervals of comprehensive index, standard index and performance index; fifthly, calculating comprehensive attribute measurement of efficiency index, and identifying attribute intervals of intelligent power distribution network efficiency so as to realize intelligent power distribution network efficiency evaluation.

Description

Discrimination method between the attribute area of intelligent distribution network high efficiency assessment

Technical field

The present invention relates to discrimination method between the attribute area of intelligent distribution network high efficiency, belong to the theoretical interleaving techniques application with attribute mathematics of intelligent distribution network.

Background technology

Intelligent grid is the development trend of following electrical network, there is the features such as safety, self-healing, interaction, compatibility, clean, efficient, high-quality, in the reply climate change, ensure national energy security, promote the green economy development, promote the aspects such as power industry Energy restructuring and there is vital role, its core driving force be solve energy security with environmental issue, tackle climate change.The essential distinction of intelligent grid and traditional electrical network is that variation has occurred energy balance model.The tradition electrical network is regulated and controled according to user's electricity needs, thus Real-time Balancing.In intelligent grid, power supply and load wide area are distributed in electrical network, realize that Electric Power Real-time Balancing need to be regulated and controled power supply, load and electrical network simultaneously, have changed physical aspect and the scheduling controlling pattern of electrical network.

Intelligent distribution network is the important component part of intelligent grid, has the essential characteristic of intelligent grid.Applying of distributed power source and large capacity accumulator system, the access of micro-electrical network and electric automobile, the plurality of energy supplies modes such as hot and cold, Electricity Federation product coexist, the large capacity dynamic load such as central air conditioner is more and more, the development and application of smart machine, make power supply, load in intelligent distribution network present the feature that wide area distributes, the user participates in the Changing Pattern that interaction has changed load, energy, information and business two-way flow.

Energy environment pressure has proposed the demand of efficient operation to intelligent distribution network, intelligent distribution network can adopt multiple means to realize efficient operation.Make rational planning for and utilize distributed power source can dwindle fault incidence, improve the long distance that trend distributes, reduces energy and carry, realize energy-saving and emission-reduction.The correct guidance user participates in interaction and can improve the operation of power networks level, reduce peak-valley difference, reduce or delay the power grid construction investment.Distributed power source exert oneself stability and adjustable controllability poor, the risk of the higher power network safety operation of permeability is just larger.User's mutual-action behavior has spontaneity and randomness, makes the variation of part throttle characteristics be difficult to prediction, and in addition correct guidance will not affect the safe and stable operation of electrical network.It is to formulate the Scientific Construction scheme, take full advantage of distributed energy, promote the user to participate in the theoretical foundation of interaction, the development of guiding high efficiency that the high efficiency of intelligent distribution network is assessed, and has theoretical significance and practical value.

In the mankind's the understanding world, describe certain element with a little interval of length variations and whether there is the tolerance of certain attribute than more rationally credible with a definite number.The object that the information that intelligent distribution network high efficiency assessment relates to is numerous, estimate and concern complexity, carrying out intelligent power distribution web, high efficiency operation assessment based on the attribute mathematics theory can be estimated intelligent distribution network objectively, and appraisal procedure can be implemented, applicability is good.Identification between attribute area is the core of attribute mathematics method, therefore discrimination method between the attribute area of needs research intelligent distribution network high efficiency.

Summary of the invention

Technical matters: the purpose of this invention is to provide discrimination method between the attribute area of intelligent distribution network high efficiency assessment, by gathering structured data and the service data of intelligent distribution network, set up step by step the Attribute Measure interval matrix of each index, based on principle of maximum entropy computation attribute measurement index, between the attribute area of identification intelligent distribution network high efficiency, thereby realize the high efficiency assessment to intelligent distribution network.

Technical scheme: the assessment of intelligent distribution network high efficiency is by automated intelligent distribution net work structure data and service data, set up the Attribute Measure interval of each evaluation index according to the level of high efficiency, and carry out step by step Attribute Measure calculating, form discrimination method between intelligent distribution network high efficiency attribute area.

Between the attribute area of intelligent distribution network high efficiency assessment of the present invention, discrimination method is by setting up the intelligent distribution network basic data to three layers of attribute measure matrix between high efficiency, on achievement data standardization basis, Attribute Measure interval and synthesized attribute measure value based on principle of maximum entropy calculation of performance indicators, criterion index and overall target, between the attribute area of automatic Identification intelligent distribution network high efficiency.Concrete steps are as follows:

The first step: read the assessment information needed, comprise power grid construction and structural information, operation of power networks information, user dependability demand information, distributed power source capacity and operation information, electromagnetism-noise monitoring information, data of information system, power distribution network Workflow messages, user interaction information;

Second step: set up the attribute measure matrix of overall target, criterion index and performance index, specific as follows:

The attribute measure space of evaluation index divides cut set C in order _k, the attribute measure matrix of m evaluation index is as follows:

Wherein, C ₁～C _kthe element of attribute measure space; I _jfor evaluation index, j=1,2 ... m; [a _jk, b _jk] be between k the cut section of j index on F between attribute area, meet a _jk≤ b _jk, k=1,2 ..., K, note A=[a _jk] _{m * K}for lower bound canonical matrix, B=[b _jk] _{m * K}for upper bound canonical matrix;

The 3rd step: the achievement data standardization, specific as follows:

Regulation C ₁the lower bound standard a of evaluation index j in the level evaluation criterion _j1be normalized to relative degree of membership s _j1=0, C _kevaluation index j lower bound standard a in the level evaluation criterion _jkbe normalized to relative degree of membership s _jk=1, pass through formula

determine C _kthe lower bound standard a of index j in the level evaluation criterion _jkrelative degree of membership,

Therefore, lower bound canonical matrix A=[a _jk] _{m * K}be transformed to the relative degree of membership matrix S of lower bound standard=[s _jk] _{m * K}, wherein: the index that larger its larger grade of desired value is direct index, on the contrary the index that larger its less grade of desired value is negative index, adopts following formula will estimate sample matrix X=[x simultaneously _ij] _{n * m}be transformed to the relative degree of membership matrix to A $\underset{\‾}{F}={\left[{\underset{\‾}{f}}_{\mathrm{ij}}\right]}_{n\×m};$

Direct index: ${\underset{\&OverBar;}{f}}_{\mathrm{ij}}=\left\{\begin{array}{c}0,x_{\mathrm{ij}}<a_{j1}\\ \frac{x_{\mathrm{ij}}-a_{j1}}{a_{\mathrm{jk}}-a_{j1}},a_{j1}\&le;x_{\mathrm{ij}}\&le;a_{\mathrm{jk}}\\ 1,x_{\mathrm{ij}}>a_{\mathrm{jk}}\end{array}\right.;$ Or negative index: ${\underset{\&OverBar;}{f}}_{\mathrm{ij}}=\left\{\begin{array}{c}0,x_{\mathrm{ij}}>a_{j1}\\ \frac{x_{\mathrm{ij}}-a_{j1}}{a_{\mathrm{jk}}-a_{j1}},a_{\mathrm{jk}}\&le;x_{\mathrm{ij}}\&le;a_{j1}\\ 1,x_{\mathrm{ij}}<a_{\mathrm{jk}}\end{array}\right.$

In like manner, B is transformed to the relative degree of membership matrix of upper bound standard

x is transformed to the relative degree of membership matrix to B $\stackrel{\&OverBar;}{F}={\left[{\stackrel{\&OverBar;}{f}}_{\mathrm{ij}}\right]}_{n\&times;m},$

The 4th step: the Attribute Measure interval of calculation of performance indicators, criterion index, overall target, specific as follows:

N data sample is as follows to the Attribute Measure interval matrix of k opinion rating:

Wherein:

be i data sample x _iwith respect to C _kthe Attribute Measure interval of opinion rating; μ _ikfor the lower bound Attribute Measure, by sample x _icalculate with A, meet

for upper bound Attribute Measure, by sample x _icalculate with B, meet

Definition lower bound broad sense Weighted distance d _ikmean i " the lower bound difference " of estimating sample and k level,

${\underset{\&OverBar;}{D}}_{\mathrm{ik}}={\underset{\&OverBar;}{\mathrm{\&mu;}}}_{\mathrm{ik}}{d}_{\mathrm{ik}}={\underset{\&OverBar;}{\mathrm{\&mu;}}}_{\mathrm{ik}}\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\underset{\&OverBar;}{f}}_{\mathrm{ij}}-{\underset{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)$

Wherein

the generalized distance of estimating sample, μ _ikneed to make to add generalized weighed distance sum minimum between all samples to be evaluated and evaluation criterion,

simultaneously, will μ _ikbe considered as i evaluation sample and belong to C _k" the lower bound probability " of grade; According to principle of maximum entropy, obtain the lower bound Attribute Measure μ _ikcalculating formula as follows:

${\underset{\&OverBar;}{\mathrm{\&mu;}}}_{\mathrm{ik}}=\frac{\exp [-B\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\underset{\&OverBar;}{f}}_{\mathrm{ij}}-{\underset{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)]}{\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\exp [-B\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\underset{\&OverBar;}{f}}_{\mathrm{ij}}-{\underset{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)]}$

Wherein: ω _jfor the weight of every evaluation index, and meet

b is positive constant;

In like manner, upper bound Attribute Measure

calculating formula be ${\stackrel{\&OverBar;}{\mathrm{\&mu;}}}_{\mathrm{jk}}=\frac{\exp [-B\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\stackrel{\&OverBar;}{f}}_{\mathrm{ij}}-{\stackrel{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)]}{\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\exp [-B\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\stackrel{\&OverBar;}{f}}_{\mathrm{ij}}-{\stackrel{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)]},$ Thereby obtain the Attribute Measure interval of evaluation index

The 5th step: the synthesized attribute that calculates the high efficiency index is estimated, and picks out between the attribute area of intelligent distribution network high efficiency, specific as follows:

At first adopt

calculate the homogenizing synthesized attribute of high efficiency index and estimate, obtain the synthesized attribute that sample xi belongs to the k class and estimate μ _ik, then adopt the reinforcement attribute Recognition Model to estimate between attribute area and identify the synthesized attribute of intelligent distribution network high efficiency, computing formula is as follows,

$q_{\mathrm{xi}}=\underset{m=1}{\overset{k-1}{\mathrm{\&Sigma;}}}(k-1)({\mathrm{\&mu;}}_{\mathrm{im}}+\frac{a}{k-1})$

Wherein, a=min (μ _1k, μ _2k, μ _mk), obtain thus i sample x _imark q _xi.

Beneficial effect: have the following advantages for discrimination method between the attribute area of intelligent distribution network high efficiency assessment described in the present invention:

(1) qualitative index and quantitative target in the assessment of intelligent distribution network high efficiency are organically combined, naturally realize the processing of different types of data.

(2) intelligent distribution network high efficiency evaluation process is directly exported between quantitative high efficiency evaluation index value and affiliated attribute area, and easily is converted into human brain and holds intelligible mark, makes high efficiency assess this challenge and becomes very directly perceived.

(3) by the method, the high efficiency of intelligent distribution network is assessed, the principal element that affects high efficiency can be provided, for the construction of intelligent distribution network provides foundation.

Embodiment

Between the attribute area of intelligent distribution network high efficiency assessment, the identification scheme is to set up the intelligent distribution network basic data to three layers of attribute measure matrix between high efficiency, according to the intelligent distribution network basic data gathered, on the data normalization basis, Attribute Measure interval and synthesized attribute measure value based on principle of maximum entropy calculation of performance indicators, criterion index and overall target, between the attribute area of automatic Identification intelligent distribution network high efficiency.Concrete evaluation process is as follows:

The first step: read the assessment information needed, comprise power grid construction and structural information, operation of power networks information, user dependability demand information, distributed power source capacity and operation information, electromagnetism/noise monitoring information, data of information system, power distribution network Workflow messages, user interaction information etc., such as the partial data of Hexi Area, Nanjing power distribution network is as shown in the table;

The average frequency of power cut R11(times/year of user)	0.786	The average power off time R12(hour/year of user)	7.089
				Power failure electric weight accounting R13(%)	0.5	The average attaching capacity of circuit R14(MW)	7
N-1 qualification rate R15(%)	96.6	Circuit availability R21(%)	99
				Equipment availability R22(%)	98.5	Uninterrupted operation accounting R31(%)	11
Repair based on condition of component coverage rate R32(%)	95	Ten thousand yuan of Unit Assets year delivery E11(hundred million kWh/)	22600
				Ten thousand yuan of maximum supply load E12(MW/ of Unit Assets year)	7	Capacity-load ratio E13	2.1
Economical operation drift rate E21(%)	20	Equipment underloading rate E22(%)	19
				Comprehensive line loss per unit E23(%)	6.2	Unit Assets year operation expense E24(unit/ten thousand yuan)	589
Ten thousand yuan of loss of outage E25()	142.93	?	?

The bound of indices is as shown in the table:

Evaluation index	Lower limit	The upper limit	Evaluation index	Lower limit	The upper limit
						R11(times/year)	0.786	0.789	During R12(/ year)	7.089	7.093

R13（%）R15（%）	0.596.6	0.5396.8	R14（MW）R21（%）	799	7.599
						R22（%）	98.5	99	R31（%）	11	11.2
R32（%）	95	95.7	Ten thousand yuan of E11(hundred million kWh/)	22600	226500
						Ten thousand yuan of E12(MW/)	7	7.2	E13	2.1	2.3
E21（%）	20	20.3	E22（%）	19	19.3
						E23（%）	6.2	6.25	E24(unit/ten thousand yuan)	589	590
Ten thousand yuan of E25()	142.93	143	?	?	?

Second step: set up the attribute measure matrix of overall target, criterion index and performance index, specific as follows:

The attribute measure space of evaluation index divides cut set C in order _kmean, such as high efficiency is divided into " poor efficiency, lower, generally, higher, efficient " five grades, the attribute measure matrix of m evaluation index is as follows:

Wherein, C ₁～C _kthe element of attribute measure space; I _jfor evaluation index; J=1,2 ... m; [a _jk, b _jk] be between k the cut section of j index on F between attribute area, meet a _jk≤ b _jk, k=1,2 ..., K, note A=[a _jk] _{m * K}and B=[b _jk] _{m * K}be respectively that lower bound is estimated matrix and matrix is estimated in the upper bound.Such as the Hexi Area, Nanjing power distribution network next time relevant to reliability and economy, to estimate matrix as follows.

The 3rd step: the achievement data standardization, specific as follows:

Regulation C ₁the lower bound standard a of evaluation index j in the level evaluation criterion _j1be normalized to relative degree of membership s _j1=0, C _kevaluation index j lower bound standard a in the level evaluation criterion _jkbe normalized to relative degree of membership s _jk=1, pass through formula

determine C _kthe lower bound standard a of index j in the level evaluation criterion _jkrelative degree of membership.

Therefore, lower bound is estimated matrix A=[a _jk] _{m * K}be transformed to the relative degree of membership matrix S of lower bound standard=[s _jk] _{m * K}, wherein: the index that larger its larger grade of desired value is direct index, on the contrary the index that larger its less grade of desired value is negative index, simultaneously, adopts following formula will estimate sample matrix X=[x _ij] _{n * m}be transformed to the relative degree of membership matrix to A f=[ f _ij] _{n * m}.

Direct index: ${\underset{\&OverBar;}{f}}_{\mathrm{ij}}=\left\{\begin{array}{c}0,x_{\mathrm{ij}}<a_{j1}\\ \frac{x_{\mathrm{ij}}-a_{j1}}{a_{\mathrm{jk}}-a_{j1}},a_{j1}\&le;x_{\mathrm{ij}}\&le;a_{\mathrm{jk}}\\ 1,x_{\mathrm{ij}}>a_{\mathrm{jk}}\end{array}\right.;$ Or negative index: ${\underset{\&OverBar;}{f}}_{\mathrm{ij}}=\left\{\begin{array}{c}0,x_{\mathrm{ij}}>a_{j1}\\ \frac{x_{\mathrm{ij}}-a_{j1}}{a_{\mathrm{jk}}-a_{j1}},a_{\mathrm{jk}}\&le;x_{\mathrm{ij}}\&le;a_{j1}\\ 1,x_{\mathrm{ij}}<a_{\mathrm{jk}}\end{array}\right.$

In like manner, B is transformed to the relative degree of membership matrix of upper bound standard

x is transformed to the relative degree of membership matrix to B $\stackrel{\&OverBar;}{F}={\left[{\stackrel{\&OverBar;}{f}}_{\mathrm{ij}}\right]}_{n\&times;m}.$

Such as the result that the Hexi Area, Nanjing power distribution network next time relevant to reliability and economy estimates after the matrix standardization is as follows.

The 4th step: the Attribute Measure interval of calculation of performance indicators, criterion index, overall target, specific as follows:

N data sample is as follows to the Attribute Measure interval matrix of k opinion rating:

Wherein:

be i data sample x _iwith respect to C _kthe Attribute Measure interval of opinion rating; μ _ikfor the lower bound Attribute Measure, by sample x _icalculate with A, meet

for upper bound Attribute Measure, by sample x _icalculate with B, meet

Definition lower bound broad sense Weighted distance d _ikmean i " the lower bound difference " of estimating sample and k level,

${\underset{\&OverBar;}{D}}_{\mathrm{ik}}={\underset{\&OverBar;}{\mathrm{\&mu;}}}_{\mathrm{ik}}{d}_{\mathrm{ik}}={\underset{\&OverBar;}{\mathrm{\&mu;}}}_{\mathrm{ik}}\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\underset{\&OverBar;}{f}}_{\mathrm{ij}}-{\underset{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)$

Wherein

the generalized distance of estimating sample, μ _ikneed to make to add generalized weighed distance sum minimum between all samples to be evaluated and evaluation criterion, simultaneously, will μ _ikbe considered as i evaluation sample and belong to C _k" the lower bound probability " of grade.According to principle of maximum entropy, obtain the lower bound Attribute Measure μ _ikcalculating formula as follows:

${\underset{\&OverBar;}{\mathrm{\&mu;}}}_{\mathrm{ik}}=\frac{\exp [-B\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\underset{\&OverBar;}{f}}_{\mathrm{ij}}-{\underset{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)]}{\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\exp [-B\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\underset{\&OverBar;}{f}}_{\mathrm{ij}}-{\underset{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)]}$

Wherein: ω _jfor the weight of every evaluation index, and meet

b is positive constant.

In like manner, upper bound Attribute Measure

calculating formula be ${\stackrel{\&OverBar;}{\mathrm{\&mu;}}}_{\mathrm{jk}}=\frac{\exp [-B\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\stackrel{\&OverBar;}{f}}_{\mathrm{ij}}-{\stackrel{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)]}{\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\exp [-B\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\stackrel{\&OverBar;}{f}}_{\mathrm{ij}}-{\stackrel{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)]},$ Thereby obtain the Attribute Measure interval of evaluation index

Such as the Hexi Area, Nanjing power distribution network Criterion Attribute relevant to reliability and economy estimated interval as follows.

The performance index Attribute Measure interval calculated by These parameters Attribute Measure interval is as follows.

The criterion Criterion Attribute calculated by above-mentioned performance index Attribute Measure interval is estimated interval as follows.

Estimate by above-mentioned criterion Criterion Attribute the high efficiency overall target Attribute Measure interval that interval calculates as follows.

The 5th step: the synthesized attribute that calculates the high efficiency index is estimated, and picks out between the attribute area of intelligent distribution network high efficiency, specific as follows:

At first adopt

calculate the homogenizing synthesized attribute of high efficiency index and estimate, obtain sample x _ithe synthesized attribute that belongs to the k class is estimated μ _ik, such as the high efficiency Synthetic Measurement of the south of the River, Nanjing eight district's power distribution networks is 0.20165.Then adopt the reinforcement attribute Recognition Model to estimate between attribute area and identify the synthesized attribute of intelligent distribution network high efficiency, computing formula is as follows,

$q_{\mathrm{xi}}=\underset{m=1}{\overset{k-1}{\mathrm{\&Sigma;}}}(k-1)({\mathrm{\&mu;}}_{\mathrm{im}}+\frac{a}{k-1})$

Wherein, a=min (μ _1k, μ _2k, μ _mk).Obtain thus i sample x _imark q _xi, such as the high efficiency overall target of the south of the River, Nanjing eight district's power distribution networks is 4.0064, corresponding mark is 80.128 minutes, therefore for the level of more efficient.

Claims (1)

1. discrimination method between the attribute area of intelligent distribution network high efficiency assessment, it is characterized in that setting up the intelligent distribution network basic data to three layers of attribute measure matrix between high efficiency, on achievement data standardization basis, Attribute Measure interval and synthesized attribute measure value based on principle of maximum entropy calculation of performance indicators, criterion index and overall target, between the attribute area of automatic Identification intelligent distribution network high efficiency, concrete steps are as follows:

The first step: read the assessment information needed, comprise power grid construction and structural information, operation of power networks information, user dependability demand information, distributed power source capacity and operation information, electromagnetism-noise monitoring information, data of information system, power distribution network Workflow messages, user interaction information;

Second step: set up the attribute measure matrix of overall target, criterion index and performance index, specific as follows:

The attribute measure space of evaluation index divides cut set C in order _k, the attribute measure matrix of m evaluation index is as follows:

Wherein, C ₁～C _kthe element of attribute measure space; I _jfor evaluation index, j=1,2 ... m; [a _jk, b _jk] be between k the cut section of j index on F between attribute area, meet a _jk≤ b _jk, k=1,2 ..., K, note A=[a _jk] _{m * K}for lower bound canonical matrix, B=[b _jk] _{m * K}for upper bound canonical matrix;

The 3rd step: the achievement data standardization, specific as follows:

Regulation C ₁the lower bound standard a of evaluation index j in the level evaluation criterion _j1be normalized to relative degree of membership s _j1=0, C _kevaluation index j lower bound standard a in the level evaluation criterion _jkbe normalized to relative degree of membership s _jk=1, pass through formula

determine C _kthe lower bound standard a of index j in the level evaluation criterion _jkrelative degree of membership,

Therefore, lower bound canonical matrix A=[a _jk] _{m * K}be transformed to the relative degree of membership matrix S of lower bound standard=[s _jk] _{m * K}, wherein: the index that larger its larger grade of desired value is direct index, on the contrary the index that larger its less grade of desired value is negative index, adopts following formula will estimate sample matrix X=[x simultaneously _ij] _{n * m}be transformed to the relative degree of membership matrix to A f=[ f _ij] _{n * m};

Direct index: ${\underset{\&OverBar;}{f}}_{\mathrm{ij}}=\left\{\begin{array}{c}0,x_{\mathrm{ij}}<a_{j1}\\ \frac{x_{\mathrm{ij}}-a_{j1}}{a_{\mathrm{jk}}-a_{j1}},a_{j1}\&le;x_{\mathrm{ij}}\&le;a_{\mathrm{jk}}\\ 1,x_{\mathrm{ij}}>a_{\mathrm{jk}}\end{array}\right.;$ Or negative index: ${\underset{\&OverBar;}{f}}_{\mathrm{ij}}=\left\{\begin{array}{c}0,x_{\mathrm{ij}}>a_{j1}\\ \frac{x_{\mathrm{ij}}-a_{j1}}{a_{\mathrm{jk}}-a_{j1}},a_{\mathrm{jk}}\&le;x_{\mathrm{ij}}\&le;a_{j1}\\ 1,x_{\mathrm{ij}}<a_{\mathrm{jk}}\end{array}\right.$

In like manner, B is transformed to the relative degree of membership matrix of upper bound standard

x is transformed to the relative degree of membership matrix to B $\stackrel{\&OverBar;}{F}={\left[{\stackrel{\&OverBar;}{f}}_{\mathrm{ij}}\right]}_{n\&times;m},$

The 4th step: the Attribute Measure interval of calculation of performance indicators, criterion index, overall target, specific as follows:

N data sample is as follows to the Attribute Measure interval matrix of k opinion rating:

Wherein: be i data sample x _iwith respect to C _kthe Attribute Measure interval of opinion rating; μ _ikfor the lower bound Attribute Measure, by sample x _icalculate with A, meet

for upper bound Attribute Measure, by sample x _icalculate with B, meet

Definition lower bound broad sense Weighted distance d _ikmean i " the lower bound difference " of estimating sample and k level,

${\underset{\&OverBar;}{D}}_{\mathrm{ik}}={\underset{\&OverBar;}{\mathrm{\&mu;}}}_{\mathrm{ik}}{d}_{\mathrm{ik}}={\underset{\&OverBar;}{\mathrm{\&mu;}}}_{\mathrm{ik}}\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\underset{\&OverBar;}{f}}_{\mathrm{ij}}-{\underset{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)$

Wherein

the generalized distance of estimating sample, μ _ikneed to make to add generalized weighed distance sum minimum between all samples to be evaluated and evaluation criterion,

simultaneously, will μ _ikbe considered as i evaluation sample and belong to C _k" the lower bound probability " of grade; According to principle of maximum entropy, obtain the lower bound Attribute Measure μ _ikcalculating formula as follows:

${\underset{\&OverBar;}{\mathrm{\&mu;}}}_{\mathrm{ik}}=\frac{\exp [-B\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\underset{\&OverBar;}{f}}_{\mathrm{ij}}-{\underset{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)]}{\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\exp [-B\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\underset{\&OverBar;}{f}}_{\mathrm{ij}}-{\underset{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)]}$

Wherein: ω _jfor the weight of every evaluation index, and meet

; B is positive constant;

In like manner, upper bound Attribute Measure

calculating formula be ${\stackrel{\&OverBar;}{\mathrm{\&mu;}}}_{\mathrm{jk}}=\frac{\exp [-B\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\stackrel{\&OverBar;}{f}}_{\mathrm{ij}}-{\stackrel{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)]}{\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\exp [-B\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\left({\mathrm{\&omega;}}_j\right|{\stackrel{\&OverBar;}{f}}_{\mathrm{ij}}-{\stackrel{\&OverBar;}{s}}_{\mathrm{jk}}\left|\right)]},$ Thereby obtain the Attribute Measure interval of evaluation index

The 5th step: the synthesized attribute that calculates the high efficiency index is estimated, and picks out between the attribute area of intelligent distribution network high efficiency, specific as follows:

At first adopt

calculate the homogenizing synthesized attribute of high efficiency index and estimate, obtain sample x _ithe synthesized attribute that belongs to the k class is estimated μ _ik, then adopt the reinforcement attribute Recognition Model to estimate between attribute area and identify the synthesized attribute of intelligent distribution network high efficiency, computing formula is as follows,

$q_{\mathrm{xi}}=\underset{m=1}{\overset{k-1}{\mathrm{\&Sigma;}}}(k-1)({\mathrm{\&mu;}}_{\mathrm{im}}+\frac{a}{k-1})$

Wherein, a=min (μ _1k, μ _2k, μ _mk), obtain thus i sample x _imark q _xi.