CN104376502A - Electric power customer credit comprehensive evaluation method based on grey relational degree - Google Patents

Abstract

The invention relates to an electric power customer credit comprehensive evaluation method based on the grey relational degree, and belongs to the field of electric power analysis. The new electric power customer credit evaluation indexes including big customer direct-purchase electricity contract breach electric quantity ratio and the new energy power generation grid connection quantity are taken into consideration, and an electric power customer credit rating evaluation index system is established in the seven aspects of electricity utilization payment credit, the electric power law and regulation credit, the electric power cooperation credit, the operation capacity credit, the social communication credit, the mortgage credit guarantee and development prospect credit of the electric power customers. Three evaluation methods with different attributes including a layer analysis method, an entropy weight method and a neural network method are integrated, combined weight is adopted, linear combination coefficients are determined by minimizing the deviation between single evaluation weights and combination weights, the potential information of an evaluated object is fully extracted, and the one-sidedness of a single evaluation method is avoided. Compared with the prior art, the electric power customer credit comprehensive evaluation method based on the grey relational degree has the advantages that a grey relational degree analysis method is adopted, operation is easy, efficiency is high, and the quantity of needed data is small.

Description

Based on the power customer credit integrated evaluating method of grey relational grade

Technical field

The invention belongs to electric power analysis field, particularly a kind of power customer credit integrated evaluating method.

Background technology

According to statistics, China's generated energy 5.32 trillion kilowatt hour in 2013, increase by about 7.5% on a year-on-year basis, electricity consumption elasticity is 0.974.Wherein: hydropower amount 8,700 hundred million kilowatt hour; Thermal power output 4.2 trillion kilowatt hour; Nuclear power generated energy 1,100 hundred million kilowatt hour; Wind-power electricity generation amount 1,340 hundred million kilowatt hour; Solar electrical energy generation amount 7,000,000,000 kilowatt hour; Biomass fermentation electricity 42,000,000,000 kilowatt hour.In order to these more and more huger generating supplys of reasonable distribution, supervise the service condition of electric power, Credit Evaluation of Power Consumers work seems more and more important simultaneously.

Meanwhile, new energy power generation grid-connection figureofmerit reflects responsiveness that user calls energy-saving and emission-reduction and the self-sustaining degree of electricity consumption oneself.The response of larger explanation user to energy-saving and emission-reduction is more positive, and have stronger environmental protecting commonweal image, social prestige adds, and promise breaking possibility is relatively little.Simultaneously generation of electricity by new energy not only can oneself from foot portions electricity, and according to " distributed power generation management tentative method " relevant regulations that country prints and distributes for 2013, user can also obtain corresponding generation of electricity by new energy subsidy, oneself self-sustaining ability of larger explanation user power utilization is stronger, the electricity charge pressure need paying power supply enterprise is less, even can also offset the electricity charge and obtain profit.

At present, the method for common Credit Evaluation of Power Consumers mainly contains cluster analysis, grey correlation analysis GRA, approaches ideal point distance etc.Cluster analysis affects very large by singular value, responsive to initial choosing value, need specify clusters number in advance, when, data inadequate at sample are inaccurate, efficiency is lower.Approach ideal point Furthest Neighbor to be easily understood, but the relevant information replication problem caused between index can not be solved.

So this area needs a kind of new technology badly to change such present situation.

Summary of the invention

Technical matters to be solved by this invention: for shortcomings and deficiencies of the prior art, the invention provides and a kind ofly adopt grey Relational Analysis Method, is easy to the power customer credit integrated evaluating method based on grey relational grade that operates, efficiency is high, desired data is few.

The present invention is design like this: based on the power customer credit integrated evaluating method of grey relational grade, it is characterized in that: comprise the following steps:

Step one: the data gathering the electricity consumption payment credit of power customer, power rule of law credit, electric power cooperation credit, management ability credit, social interaction credit, secured credit guarantee, development prospect credit, for power customer credit comprehensive evaluation;

Step 2: build the power customer credit comprehensive evaluation model based on grey relational grade, namely ideal indicator set is constructed, by the data discrete degree determination resolution ratio in analytical procedure one, thus build the degree of association matrix of coefficients of evaluation object, generate degree of association vector;

Step 3: comprehensive analytical hierarchy process, the entropy assessment of Objective Weight, the neural network of intelligence tax power using subjective weights, by minimizing each " single evaluation weight " and the deviation determination linear combination coefficient of " combining weights ", obtain object weight vectors;

Step 4: the degree of association calculating power customer based on step 2 and step 3, association angle value is larger, represents power customer credit better, association angle value is less, represent power customer credit poorer, according to the degree of association, each power customer is sorted, obtain the report of power customer credit comprehensive evaluation.

The power customer credit comprehensive evaluation model based on grey relational grade built in described step 2 comprises new energy power generation grid-connection amount and Direct Purchase of Electric Energy by Large Users promise breaking electricity ratio two indices.

Described step 3 finally can obtain basic weight vectors, possibility weight vectors and be satisfied with weight vectors three kinds of weight vectors most, will be satisfied with weight vectors most as object weight vectors.

By above-mentioned design proposal, the present invention can bring following beneficial effect:

Namely the grey correlation analysis that the present invention adopts is the grey majorized model for evaluating things, by the mathematics manipulation to existed system data, by comparing the correlation degree between things, understanding the mutual relationship of internal system and variation tendency, being applicable to that sample is less, system in the not congruent situation of information.

The method Comprehensive analytic hierarchy process that the present invention proposes and genetic algorithm, and make use of the objective information that data provide, more science, reliably, synthetically power customer credit is evaluated, avoid the one-sidedness of single evaluation method, between different weight, search out the most satisfactory combination weight of balanced inconsistency.Meanwhile, grey relational grade comprehensive evaluation model has features such as being easy to operation, efficiency is high, desired data is few, and evaluation effect can be made better.Effectively can define the credit grade of client, carry out otherness power supply service for power supply enterprise and marketing strategy provides important reference frame, thus help power supply enterprise to improve tariff recovery rate, reduce business risk.

Accompanying drawing explanation

Illustrate that the invention will be further described with embodiment below in conjunction with accompanying drawing.

Fig. 1 is the kind of the image data of the power customer credit integrated evaluating method that the present invention is based on grey relational grade.

Fig. 2 is the process flow diagram of the power customer credit integrated evaluating method that the present invention is based on grey relational grade.

Embodiment

The power customer credit integrated evaluating method based on grey relational grade as shown in the figure, is characterized in that: comprise the following steps:

Step one: the data gathering the electricity consumption payment credit of power customer, power rule of law credit, electric power cooperation credit, management ability credit, social interaction credit, secured credit guarantee, development prospect credit, for power customer credit comprehensive evaluation;

Step 2: build the power customer credit comprehensive evaluation model based on grey relational grade, namely ideal indicator set is constructed, by the data discrete degree determination resolution ratio in analytical procedure one, thus build the degree of association matrix of coefficients of evaluation object, generate degree of association vector;

Step 3: comprehensive analytical hierarchy process, the entropy assessment of Objective Weight, the neural network of intelligence tax power using subjective weights, by minimizing each " single evaluation weight " and the deviation determination linear combination coefficient of " combining weights ", obtain object weight vectors;

Step 4: the degree of association calculating power customer based on step 2 and step 3, association angle value is larger, represents power customer credit better, association angle value is less, represent power customer credit poorer, according to the degree of association, each power customer is sorted, obtain the report of power customer credit comprehensive evaluation.

The power customer credit comprehensive evaluation model based on grey relational grade built in described step 2 comprises new energy power generation grid-connection amount and Direct Purchase of Electric Energy by Large Users promise breaking electricity ratio two indices.

Described step 3 finally can obtain basic weight vectors, possibility weight vectors and be satisfied with weight vectors three kinds of weight vectors most, will be satisfied with weight vectors most as object weight vectors.

Direct Purchase of Electric Energy by Large Users promise breaking electricity ratio index reflects the credit of user in social interaction and stability.Direct Purchase of Electric Energy by Large Users contract as a kind of special commercial contract, be to evaluate in power customer credit index system in social interaction Credit Factors the most closely, the most direct one, be also one that the most easily obtains accurate information.On the other hand, straight power purchase promise breaking electricity needs user to pay corresponding penalty for nonperformance of contract, can increase the grid electricity fee cost of enterprise.Therefore, the social interaction credit of less explanation user is better, and stability is stronger, and self overall planning ability is stronger, and grid electricity fee cost is less.This index can be expressed as that user causes the moon, Contract generation to contract the ratio of moon power consumption lower than the contract skew component electricity of electricity 97% and straight purchase risk of straight purchase risk.

The present invention is based on the grey relational grade analysis method establishment evaluation and decision model of gray system theory:

One, the determination of correlation coefficient

Desirable (benchmark) index set of structure, the correlation coefficient between the index of power customer i and ideal indicator is:

${\mathrm{\ξ}}_{\mathrm{ik}}=\frac{\underset{1\≤i\≤m,1\≤j\≤n}{\min }|x_{0j}-x_{\mathrm{ij}}|+\mathrm{\ρ}\underset{1\≤i\≤m,1\≤j\≤n}{\max }|x_{0j}-x_{\mathrm{ij}}|}{|x_{0k}-x_{\mathrm{ik}}|+\mathrm{\ρ}\underset{1\≤i\≤m,1\≤j\≤n}{\max }|x_{0j}-x_{\mathrm{ij}}|};$

Above-mentioned formula (2) is utilized to calculate incidence coefficient matrix:

$E=\left[\begin{array}{cccc}{\mathrm{\ξ}}_{11}& {\mathrm{\ξ}}_{12}& ...& {\mathrm{\ξ}}_{1n}\\ {\mathrm{\ξ}}_{21}& {\mathrm{\ξ}}_{22}& ...& {\mathrm{\ξ}}_{2n}\\ .& .& & .\\ .& .& & .\\ .& .& & .\\ {\mathrm{\ξ}}_{m1}& {\mathrm{\ξ}}_{m2}& ...& {\mathrm{\ξ}}_{\mathrm{mn}}\end{array}\right].$

Two, the determination of resolution ratio

Directly affecting net result from the size of the known ρ of above formula, in order to fully demonstrate the globality of association, adopting mean method, namely to calculate and the relation curve of ρ and the area S of coordinate axis determine the span of ρ.Due to the separation and war degree of index sample various degrees, quantize the average that this index can be designated as all sample absolute difference:

${\mathrm{\Δ}}_v=\frac{1}{\mathrm{nm}}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}|x_{\mathrm{ij}}-x_{0j}|;$

And remember, then the span of ρ is 2.

Three, the determination of the degree of association

Grey incidence coefficient matrix computations is utilized to obtain the weighted association degree vector of each power customer index set and ideal indicator set: R ^t=ξ ω=[R ₁r ₂r _m], in formula

$R_i=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\ξ}}_{\mathrm{ik}}{\mathrm{\ω}}_k,$

Be wherein the weight of index k, found the method determination weight of optimum combination weight by formula below.According to the degree of association, each power consumer is sorted, be worth larger, represent power customer credit better, otherwise value is less, represents power customer credit poorer.

Four, combining weights defining method

Single evaluation method has " Preference ", this can cause evaluation result to occur drift, in order to avoid the one-sidedness that single method determination weight exists, adopt the minimum deviation method for uncertain based on each single evaluation weight and combining weights, determine combining weights, the evaluation method of different attribute is merged, equilibrium be unified into one more objective, more scientific, the more index weights of system, namely respectively by the AHP method of subjective weights, the neural network based on genetic algorithm that the entropy assessment of Objective Weight and intelligence compose power obtains basic weight vectors, possible weight vectors and the most satisfied weight vectors.

Basic weight sets and possibility weight sets:

In order to avoid the one-sidedness of single evaluation method determination weight, consider the information that the multiple different attribute evaluation method of comprehensive utilization is transmitted, namely determine basic weight sets by different evaluation method, and build possibility weight sets by linear combination.Suppose to use L kind method to compose power respectively to index, namely obtain the weight vectors of L different attribute, and then a structure basic weight vectors collection, then forming linear combination by this L weight vectors is: in formula, it is linear combination coefficient; For the one based on basic weight may weight vectors.All possible weight vectors collection can be expressed as

$\left\{\mathrm{\ω}\right|\mathrm{\ω}=\underset{h=1}{\overset{L}{\mathrm{\Σ}}}{\mathrm{\λ}}_h{\mathrm{\ω}}_h^T,{\mathrm{\λ}}_h>0\},$

This vector set be basic weight vectors is intersected, the result of collectionization.After constructing possibility weight vectors collection, key how may select the most satisfied weight vectors by weight sets from numerous.

Most satisfactory combination weight

Adopt " addition " Integration Method to combine analytical hierarchy process and Information Entropy determination weight, but to how finding most satisfactory combination weight do not set off a discussion.Adopt the minimum deviation method for uncertain based on each single evaluation weight and combining weights, determine combining weights, can be summed up as and the linear combination coefficient in formula is optimized, make with deviation minimize.Can derive symmetry model is thus:

$\min {\left|\right|\underset{h=1}{\overset{L}{\mathrm{\Σ}}}{\mathrm{\λ}}_h{\mathrm{\ω}}_h^T-{\mathrm{\ω}}_i^T\left|\right|}^2(i=\mathrm{1,2},...,L);\underset{h}{\overset{L}{\mathrm{\Σ}}}{\mathrm{\λ}}_h=1.$

By solve above-mentioned model can obtain one coordinate mutually with multiple weight assignment method, mutually balanced consistent combining weights result.Optimization first order derivative condition can be obtained according to differentiation of a matrix character, and then try to achieve linear combination coefficient:

$\underset{h=1}{\overset{L}{\mathrm{\Σ}}}{\mathrm{\λ}}_h{\mathrm{\ω}}_i{\mathrm{\ω}}_h^T={\mathrm{\ω}}_i{\mathrm{\ω}}_h^T(i=\mathrm{1,2},...,L).$

Below with instantiation application operating process:

Choose electric company of Jilin Province 10 power customer raw data of 2010 to 2012 years and carry out Electric Power Customer Credit Rank Appraisal, form dimension matrix.

Adopt corresponding normalization method for dissimilar index, realize nondimensionalization.The neural network based on genetic algorithm respectively by the analytical hierarchy process (AHP) of subjective weights, the entropy assessment of Objective Weight and intelligence tax power obtains 3 basic weight vectors, and it is as shown in the table.

Each index weights form

Table can obtain grey incidence coefficient matrix thus, and obtains degree of association vector and the rank of each power customer index set and ideal indicator set.

Claims (3)

1., based on the power customer credit integrated evaluating method of grey relational grade, it is characterized in that: comprise the following steps:

Step one: the data gathering the electricity consumption payment credit of power customer, power rule of law credit, electric power cooperation credit, management ability credit, social interaction credit, secured credit guarantee, development prospect credit, for power customer credit comprehensive evaluation;

Step 2: build the power customer credit comprehensive evaluation model based on grey relational grade, namely ideal indicator set is constructed, by the data discrete degree determination resolution ratio in analytical procedure one, thus build the degree of association matrix of coefficients of evaluation object, generate degree of association vector;

Step 3: comprehensive analytical hierarchy process, the entropy assessment of Objective Weight, the neural network of intelligence tax power using subjective weights, by minimizing each " single evaluation weight " and the deviation determination linear combination coefficient of " combining weights ", obtain object weight vectors;

Step 4: the degree of association calculating power customer based on step 2 and step 3, association angle value is larger, represents power customer credit better, association angle value is less, represent power customer credit poorer, according to the degree of association, each power customer is sorted, obtain the report of power customer credit comprehensive evaluation.

2. the power customer credit integrated evaluating method based on grey relational grade according to claim 1, is characterized in that: the power customer credit comprehensive evaluation model based on grey relational grade built in described step 2 comprises new energy power generation grid-connection amount and Direct Purchase of Electric Energy by Large Users promise breaking electricity ratio two indices.

3. the power customer credit integrated evaluating method based on grey relational grade according to claim 1, it is characterized in that: described step 3 finally can obtain basic weight vectors, possibility weight vectors and be satisfied with weight vectors three kinds of weight vectors most, will be satisfied with weight vectors most as object weight vectors.