CN104112204A - Evaluation method for efficient operation of power supply quality - Google Patents

Abstract

The invention relates to an evaluation method for efficient operation of power supply quality. The evaluation method includes the steps of: (1) defining a mutual degree of importance between indexes of the same level, adopting an analytic hierarchy process to calculate weights (2) of indexes of the same level in a superior index and converting an original data result of an internal evaluation index to a quantized index value of 0 to 100; (3) correcting a customer satisfaction calculation result; (4) establishing a variation trend function using evaluation time as a cross axis and an evaluation index as a longitudinal axis by an average value of multiple movements of a historical evaluation results of the evaluation index, and predicting an evaluation index of the next period; and (5) redefining the indexes of the same level. The evaluation method for efficient operation of power supply quality in the invention can provides a basis for revision of service quality standards, thereby realizing contrastive analysis of internal service quality evaluation and external customer evaluation, realizing power supply service quality monitoring, evaluation and auxiliary analysis functions based on a service process, and providing a basis for revising the quality standards.

Description

The evaluation method of the efficient operation of power supply quality

One, technical field

The invention belongs to electrical power services operation technical field, the evaluation method of the efficient operation of especially a kind of power supply quality.

Two, background technology

From 2007, State Grid Corporation of China has issued power supply service quality evaluation method (trying) the > > of < < State Grid Corporation of China, each province (city) company organizes one after another for service quality evaluation activity, but because power supply service quality lacks unified evaluation criterion and evaluation means, cause evaluation result objective not and comprehensive, also do not set up power supply service quality evaluation of tissue mechanism, evaluation criterion and evaluation flow process, power supply service quality evaluation system fails to realize normality operation.

Three, summary of the invention

The object of the invention is for the deficiencies in the prior art, and propose the efficient evaluation method of moving of a kind of power supply quality.

The present invention solves its technical matters and takes following technical scheme to realize:

An evaluation method for the efficient operation of power supply quality, comprises that step is as follows:

(1) analytic hierarchy process; Define the mutual significance level between index at the same level, adopt analytical hierarchy process to calculate the weight of index at the same level in higher level's index, the differentiation that solves index at the same level significance level in higher level's index;

Wherein, described index is divided into internal indicator and external indicator,

Internal indicator includes: service, bill service, the service of client's calibration and 95598 services are accepted in new clothes and increase-volume and change electricity consumption service, troublshooting service, consulting inquiry service, complaint, report and suggestion;

External indicator includes: outside overall target and outside special index,

Wherein, outside overall target further comprises: satisfaction, quality perception, value perception, image appraisal, user expectation, client's complaint, customer loyalty;

Wherein, outside special index comprises: power supply reliability, the quality of power supply, new clothes, increase-volume and change electricity consumption service, troublshooting service, consulting inquiry service, complaint, report and suggestion are accepted have a power failure informing service, Very Important Person of service, paying service, bill service, client's arrearage and stopped rationing the power supply and inform and serve and electricity consumption guide service;

(2) build multistage linear function, convert the raw data results of desk evaluation index to desired value that 0-100 quantizes;

(3) by the relation that affects of the structure variable in customer satisfaction computation model, calculate, then by factor of influence substitution customer satisfaction computation model, customer satisfaction result of calculation is revised;

(4), by the trend method of moving average, it is the variation tendency function that transverse axis, evaluation number are the longitudinal axis that the history evaluation result of evaluation index is built to evaluation time by moving average repeatedly, predicts the evaluation number in next cycle;

(5) redefining of index at the same level, while having desired value to be sky in index at the same level, redefines the new system into higher level's index by the index at the same level of non-NULL, carries out weight reallocation, guarantees that the weight sum of index at the same level equals 1 all the time.

And the weight in described step (2) adopts definitely important, very important, important, important a little and five ranks of no less important.

And the calculating in described step (3) and the specifically employing PLS method of revising are calculated and are revised.

Four, advantage of the present invention and good effect

1, the present invention realizes the robotization of Data Collection and analysis, has avoided human intervention;

2, the present invention improves the acquisition channel that external client evaluates feedback data, comprise channels such as utilizing 95598 voice platforms, 95598 intelligent interaction websites, SMS platform, business hall evaluation system, third party's investigation, customer evaluation feedback to comprehensive satisfaction and special service satisfactory situation gathers, and realizes the collection by all kinds of means of customer evaluation;

3, the present invention solve in the past with manual type carry out that the statistical computation workload existing in service quality appraisal is large, the lookup service short slab problem such as not in time;

4, the inventive method can be carried out multidimensional analysis and Continual Improvement to service quality, can find the service short slab of different services channels, service item; Can find different customer types and different power supply area service short slab;

5, the present invention can provide foundation for the revision of service quality standard, realize internal services quality evaluation and external client and evaluate comparative analysis, power supply service quality monitoring, evaluation and the assistant analysis function of realization based on service procedure, for revision quality standard provides foundation;

6, the present invention is that different levels represent the relevant information that power supply service quality is evaluated.

Accompanying drawing explanation

Fig. 1 is the functional digraph that in the present invention, piecewise function is derived;

Fig. 2 is structural model and the measurement model schematic diagram in PLS algorithm of the present invention.

Five, embodiment

Below in conjunction with accompanying drawing, the invention process is further described, following examples are descriptive, are not determinate, can not limit protection scope of the present invention with this.

An evaluation method for the efficient operation of power supply quality, it is as follows that the method comprising the steps of:

(1) analytic hierarchy process; Define the mutual significance level between index at the same level, adopt analytical hierarchy process to calculate the weight of index at the same level in higher level's index, the differentiation that solves index at the same level significance level in higher level's index;

1. index comprises internal indicator and external indicator,

Internal indicator includes: service, bill service, the service of client's calibration and 95598 services are accepted in new clothes and increase-volume and change electricity consumption service, troublshooting service, consulting inquiry service, complaint, report and suggestion;

Wherein, new clothes and increase-volume and change electricity consumption service further comprises: power supply plan time for replying encashment ratio, be subject to electrical engineering design audit time encashment ratio, intermediate inspection time encashment ratio, completion check time encashment ratio, to client submit to contract for the supply and use of electricity text time encashment ratio, the dress table of intending signing to connect electricity time encashment ratio, client's industry expands the quantity of applying to install service time limit compliance rate, violation " three do not specify ";

Wherein, complaint, report and suggestion are accepted service and are further comprised: customer complaint contact time encashment ratio, complain time for replying encashment ratio, report time for replying encashment ratio, 1,000,000 clients to check and verify the rate of complaints and every 1,000,000 user quality event generation numbers;

Wherein, 95598 services further comprise: 95598 bell ring three sound (12s) are answered rate, 95598 manual service and answered that rate, 95598 business consultations inquiries rate of reply, 95598 regulation service handling work orders distribute accuracy rate, 95598 regulation business receipt confirm that promptness rates, 95598 regulation business repay time encashment ratios and 95598 regulation service handling work orders distribute promptness rate;

External indicator includes: outside overall target and outside special index,

Wherein, outside overall target further comprises: satisfaction, quality perception, value perception, image appraisal, user expectation, client's complaint, customer loyalty;

Wherein, outside special index comprises: power supply reliability, the quality of power supply, new clothes, increase-volume and change electricity consumption service, troublshooting service, consulting inquiry service, complaint, report and suggestion are accepted have a power failure informing service, Very Important Person of service, paying service, bill service, client's arrearage and stopped rationing the power supply and inform and serve and electricity consumption guide service;

2. the weight of above-mentioned these indexs is adopted to definitely important, very important, important, important a little, five ranks of no less important, as the foundation of determining judgement quantitative values, in this foundation, set A _iwith A _jwhen two factors are carried out importance degree comparison, compare yardstick a _ijimplication as shown in table 1; For n factor x ₁, x ₂..., x _n, utilize tournament method to carry out the comparative result of significance level between factor as shown in table 2; Obtain comparator matrix A:

$A=\left[\begin{array}{cccc}{a}_{11}& a_{12}& ...& a_{1n}\\ a_{21}& a_{22}& ...& a_{2n}\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ \&CenterDot;& \&CenterDot;& \&CenterDot;& \&CenterDot;\\ a_{n1}& a_{n2}& ...& a_{\mathrm{nn}}\end{array}\right]$

Wherein: a _ii=1, a _ij=1/a _ji

Table 1 is yardstick a relatively _ijimplication

Table 2 is comparative result between two

Suppose to do in matrix A while comparing between two, make w _ibe the significance level of i index, w _jbe the significance level of j index, a _ijbe that i index is with respect to the significance level fiducial value of j index, that is:

$a_{\mathrm{ij}}=\frac{w_i}{w_j}---\left(a\right)$

According to this matrix, can by certain method, obtain the value of weight vector, conventionally have and method, root method, eigenvalue method and least square method etc., here article eigenvalue method.

Eigenvalue method:

Make each component to clarification of objective vector be

W＝(w ₁,w ₂,…,w _n) ^T (b)

If had and matrix A meets

$a_{\mathrm{ij}}=\frac{a_{\mathrm{ik}}}{a_{\mathrm{jk}}},i,j,k=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;,n---\left(c\right)$

A becomes consistency matrix, is called for short consistent battle array, and n rank consistency matrix A has following character:

I, unique non-zero characteristics root that the order of A is 1, A is n;

II, arbitrary row (OK) vector of A is all to the proper vector with characteristic root n;

If the paired relatively judgment matrix obtaining is consistent battle array, the proper vector corresponding to characteristic root n normalizing represents the weight of each factor to target (or upper strata factor), and this vector is called weight vector;

If be not consistent battle array than the judgment matrix A of gained in pairs, but in inconsistent allowed band, corresponding to the maximum characteristic root λ of A _macproper vector (after normalization) as weight vector W, W meets

AW＝λ _maxW (d)

Component (the w of W wherein ₁, w ₂..., w _n) be exactly the weight coefficient corresponding to n factor;

(2) build multistage linear function, convert the raw data results of desk evaluation index to desired value that 0-100 quantizes;

The threshold value that touches the mark B obtains b and divides, and surpasses in the unit of threshold value B index mxm. x1 and obtains a and divide, and obtains c and divides, if index x between B and x1, must be divided into the * (A-b) of b+ (x-B)/(x1-B) lower than index minimum x2 in the unit of threshold value B; If index x between x2 and B, must be divided into the * (c-b) of b+ (x-B)/(X2-B).Index parameter comprises: a, and b, c, x1, x2,

The breakdown repair of take is shown up time limit compliance rate as example, first determine that the show up evaluation criterion segmentation separation of time limit compliance rate of breakdown repair is 100 minutes, 80 minutes, 60 minutes, 0 minute, if the breakdown repair time limit compliance rate 100% of showing up wherein, must be divided into 100,90% must be divided into 80,85% must be divided into 60,80% must be divided into 0.Setting up the breakdown repair time limit compliance rate piecewise function of showing up is:

Piecewise function, derives functional digraph as shown in Figure 1 thus;

As shown in Figure 1, it is 94% that assumed fault is rushed to repair the time limit compliance rate of showing up, and by this piecewise function of x=94% substitution, the score value that draws mapping is F (x)=88, be that breakdown repair the actual of time limit compliance rate of showing up must be divided into 88 minutes, the follow-up ranking operation that carries out aggregative index with this score.

$b=\frac{y_1{X}_2-y_2{X}_1}{X_2-X_1};$

(3) by the relation that affects of the structure variable in customer satisfaction computation model, calculate, then by factor of influence substitution customer satisfaction computation model, customer satisfaction result of calculation is revised;

1. PLS side's ratio juris

The method that partial least square method adopts composition to extract, is provided with q dependent variable { y ₁..., y _qand p independent variable { x ₁..., x _p, in order to study the statistical relationship of dependent variable and independent variable, we have observed n sample point, have formed thus the tables of data X=[x of independent variable and dependent variable ₁..., x _p] _{n * p}and Y=[y ₁..., y _q] _{n * q}, partial least squares regression is extracted respectively composition t in X and Y ₁and u ₁(that is to say t ₁x ₁..., x _plinear combination, u ₁y ₁..., y _qlinear combination), when extracting this two compositions, for the needs of regretional analysis, have following two requirements:

A:t ₁and u ₁should carry as wide as possible their variation information in tables of data separately;

B:t ₁and u ₁degree of correlation can reach maximum,

These two requirements show, t ₁and u ₁representative data Table X and Y as well as possible, simultaneously the composition t of independent variable ₁ingredient u to dependent variable ₁there is again the strongest interpretability.

At first composition t ₁and u ₁after being extracted, partial least squares regression is implemented respectively X to t ₁, recurrence and Y to u ₁recurrence.If regression equation has reached satisfied precision, algorithm stops; Otherwise, will utilize X by t ₁, the residual, information after explanation and Y are by u ₁residual, information after explanation is carried out the second composition of taking turns and is extracted.If finally X has been extracted to m composition t altogether ₁..., t _m, partial least squares regression will be by carrying out y _kto t ₁..., t _mrecurrence, and then be expressed as y _kabout former variable x ₁..., x _pregression equation, k=1,2 ..., q.

2. the algorithm steps of PLS

First data are done to standardization.The data matrix of X after standardization is designated as E ₀=(E ₀₁..., E _0p) _{n * p}, the data matrix of Y after standardization is designated as F ₀=(F ₀₁..., F _0q) _{n * q},

First step note t ₁e ₀first composition, t ₁=E ₀w ₁, w ₁e ₀first axle, it is a vector of unit length, || w ₁||=1,

Note u ₁f ₀first composition, u ₁=F ₀c ₁, c ₁f ₀first axle, it is a vector of unit length, and || c ₁||=1,

If t ₁, u ₁can represent well respectively the data variation information in X and Y, according to principal component analysis (PCA) principle, should have

Var(t ₁)→max (1)

Var(u ₁)→max (2)

On the other hand, due to the needs of regression modeling, require again t ₁to u ₁there is maximum interpretability, by the thinking of canonical correlation analysis, t ₁with u ₁the degree of correlation should reach maximal value,

r(t ₁,u ₁)→max (3)

Therefore, integrate, in partial least squares regression, we require t ₁with u ₁covariance reach maximum,

$\mathrm{Cov}(t_1,u_1)=\sqrt{\mathrm{Var}\left(t_1\right)\mathrm{var}\left(u_1\right)}r(t_1,u_1)\&RightArrow;\max ---\left(4\right)$

Solve above-mentioned optimization problem, adopt Lagrangian Arithmetic.

Try to achieve axle w, and c, after, composition can be obtained

t ₁＝E ₀w ₁ (5)

u ₁＝F ₀c ₁ (6)

W ₁and c ₁respectively matrix E ' ₀f ₀f ' ₀e ₀and F ' ₀e ₀e ' ₀f ₀the corresponding proper vector of eigenvalue of maximum, then, ask respectively E ₀and F ₀to t ₁, u ₁three regression equations

E ₀＝t ₁p′ ₁+E ₁ (7)

F ₀＝u ₁q′ ₁+F′ ₁ (8)

F ₀＝t ₁r′ ₁+F ₁ (9)

In formula, p ₁, q ₁, r ₁it is respectively regression coefficient vector

$p_1=\frac{E_0^{\&prime;}{t}_1}{{\left|\right|t_1\left|\right|}^2}---\left(10\right)$

$q_1=\frac{F_0^{\&prime;}{u}_1}{{\left|\right|u_1\left|\right|}^2}---\left(11\right)$

$r_1=\frac{F_0^{\&prime;}{t}_1}{{\left|\right|t_1\left|\right|}^2}---\left(12\right)$

And E ₁, F ' ₁, F ₁it is respectively the residual matrix of three regression equations.

Second step residual matrix E ₁and F ₁replace E ₀and F ₀, then, ask second axle w ₂and c ₂and second composition t ₂, u ₂, have

t ₂＝E ₁w ₂ (13)

W ₂and c ₂respectively matrix E ' ₁f ₁f ' ₁e ₁and F ' ₁e ₁e ' ₁f ₁the corresponding proper vector of eigenvalue of maximum.

Calculate again regression coefficient vector

$p_2=\frac{E_1^{\&prime;}{t}_2}{{\left|\right|t_2\left|\right|}^2}---\left(14\right)$

$r_2=\frac{F_1^{\&prime;}{t}_2}{{\left|\right|t_2\left|\right|}^2}---\left(15\right)$

Thereby obtain two regression equations

E ₁＝t ₂p′ ₂+E ₂ (16)

F ₁＝t ₂r′ ₂+F ₂ (17)

So calculate, according to intersecting the definite number m that extracts composition of validity, just can obtain final regression equation

F ₀＝t ₁r′ ₁+…+t _kr′ _k+F _k,k＝1,2,…,m (18)

As shown in Figure 2, two models, consist of: one is structural model, and another is measurement model, the relation in model between each variable can represent by three matrix equation formulas below,

η＝Bη+Γξ+ζ

y＝Λ _yη+ε

x＝Λ _xξ+δ

Structural model specifically can be expressed as following system of equations:

$\left[\begin{array}{c}{\mathrm{\&eta;}}_1\\ {\mathrm{\&eta;}}_2\\ {\mathrm{\&eta;}}_3\\ {\mathrm{\&eta;}}_4\\ {\mathrm{\&eta;}}_5\\ {\mathrm{\&eta;}}_6\end{array}\right]=\left[\begin{array}{cccccc}0& 0& 0& 0& 0& 0\\ {\mathrm{\&beta;}}_{21}& 0& 0& 0& 0& 0\\ {\mathrm{\&beta;}}_{31}& {\mathrm{\&beta;}}_{32}& 0& 0& 0& 0\\ {\mathrm{\&beta;}}_{41}& {\mathrm{\&beta;}}_{42}& {\mathrm{\&beta;}}_{43}& 0& 0& 0\\ 0& 0& 0& {\mathrm{\&beta;}}_{54}& 0& 0\\ 0& 0& 0& {\mathrm{\&beta;}}_{64}& {\mathrm{\&beta;}}_{65}& 0\end{array}\right]\&CenterDot;\left[\begin{array}{c}{\mathrm{\&eta;}}_1\\ {\mathrm{\&eta;}}_2\\ {\mathrm{\&eta;}}_3\\ {\mathrm{\&eta;}}_4\\ {\mathrm{\&eta;}}_5\\ {\mathrm{\&eta;}}_6\end{array}\right]+\left[\begin{array}{c}{\mathrm{\&gamma;}}_1\\ 0\\ 0\\ {\mathrm{\&gamma;}}_4\\ 0\\ 0\end{array}\right]\&CenterDot;\mathrm{\&xi;}+\left[\begin{array}{c}{\mathrm{\&zeta;}}_1\\ {\mathrm{\&zeta;}}_2\\ {\mathrm{\&zeta;}}_3\\ {\mathrm{\&zeta;}}_4\\ {\mathrm{\&zeta;}}_5\\ {\mathrm{\&zeta;}}_6\end{array}\right]$

Measurement model specifically can be expressed as following system of equations:

$\left[\begin{array}{c}{x}_1\\ x_2\\ x_3\\ x_4\end{array}\right]=\left[\begin{array}{c}{\mathrm{\&lambda;}}_1\\ {\mathrm{\&lambda;}}_2\\ {\mathrm{\&lambda;}}_3\\ {\mathrm{\&lambda;}}_4\end{array}\right]\&CenterDot;\mathrm{\&xi;}+\left[\begin{array}{c}{\mathrm{\&delta;}}_1\\ {\mathrm{\&delta;}}_2\\ {\mathrm{\&delta;}}_3\\ {\mathrm{\&delta;}}_4\end{array}\right]$

In above formula: x ₁, x ₂, x ₃, x ₄for image transforms seven measurable variables that obtain; y ₁, y ₂, y ₃y ₁₉for user expectation, the perception of client to quality, client is to the perception being worth, customer satisfaction, client's complaint, these six variablees of customer loyalty transform and tested and assessed variable; λ _i, λ _ijbe regression coefficient, represent the influence degree between variable; δ _i, ε _ierror for model;

(4), by the trend method of moving average, it is the variation tendency function that transverse axis, evaluation number are the longitudinal axis that the history evaluation result of evaluation index is built to evaluation time by moving average repeatedly, predicts the evaluation number in next cycle;

The trend method of moving average;

If Single moving average number is double moving average number computing formula be:

$M_t^{\left(2\right)}=\frac{M_t^{\left(1\right)}+M_{t-1}^{\left(1\right)}+\&CenterDot;\&CenterDot;\&CenterDot;+M_{t-N+1}^{\left(1\right)}}{N}=M_{t-1}^{\left(2\right)}+\frac{M_t^{\left(1\right)}-M_{t-N}^{\left(1\right)}}{N}$

Establish again time series y ₁, y ₂..., y _tfrom certain period, start to there are trends of straight line, and think that future period also changes by these trends of straight line, can establish this trends of straight line forecast model to be:

${\hat{y}}_{t+T}=a_t+b_{t}T$

In formula, t is current epoch number; T is the epoch number to time span of forecast by current epoch number t, i.e. the time of model extrapolation after t; it is the predicted value of t+T phase; a _tfor intercept; b _tfor slope.A _tb _tbe called again smoothing factor.

According to moving average, can obtain intercept a _twith slope b _tcomputing formula be:

$a_t={2M}_t^{\left(1\right)}-M_t^{\left(2\right)}$

$b_t=\frac{2}{N-1}(M_t^{\left(1\right)}-M_t^{\left(2\right)});$

(5) redefining of index at the same level, while having desired value to be sky in index at the same level, redefines the new system into higher level's index by the index at the same level of non-NULL, carries out weight reallocation, guarantees that the weight sum of index at the same level equals 1 all the time;

By weight fitting process, by there is no the weighted value of index at the same level of the index of statistics, do not reallocate, form new system, according to the relation that influences each other between index, by step analysis algorithm, recalculate index weights, guarantee that the final higher level's index weights calculating is 1, avoid losing without the index weights of statistics.

Claims (3)

1. the evaluation method that power supply quality efficiently moves, is characterized in that comprising that step is as follows:

(1) analytic hierarchy process; Define the mutual significance level between index at the same level, adopt analytical hierarchy process to calculate the weight of index at the same level in higher level's index, the differentiation that solves index at the same level significance level in higher level's index;

Wherein, described index is divided into internal indicator and external indicator,

Internal indicator includes: service, bill service, the service of client's calibration and 95598 services are accepted in new clothes and increase-volume and change electricity consumption service, troublshooting service, consulting inquiry service, complaint, report and suggestion;

External indicator includes: outside overall target and outside special index,

Wherein, outside overall target further comprises: satisfaction, quality perception, value perception, image appraisal, user expectation, client's complaint, customer loyalty;

Wherein, outside special index comprises: power supply reliability, the quality of power supply, new clothes, increase-volume and change electricity consumption service, troublshooting service, consulting inquiry service, complaint, report and suggestion are accepted have a power failure informing service, Very Important Person of service, paying service, bill service, client's arrearage and stopped rationing the power supply and inform and serve and electricity consumption guide service;

(2) build multistage linear function, convert the raw data results of desk evaluation index to desired value that 0-100 quantizes;

(3) by the relation that affects of the structure variable in customer satisfaction computation model, calculate, then by factor of influence substitution customer satisfaction computation model, customer satisfaction result of calculation is revised;

(4), by the trend method of moving average, it is the variation tendency function that transverse axis, evaluation number are the longitudinal axis that the history evaluation result of evaluation index is built to evaluation time by moving average repeatedly, predicts the evaluation number in next cycle;

(5) redefining of index at the same level, while having desired value to be sky in index at the same level, redefines the new system into higher level's index by the index at the same level of non-NULL, carries out weight reallocation, guarantees that the weight sum of index at the same level equals 1 all the time.

2. the evaluation method of the efficient operation of power supply quality according to claim 1, is characterized in that: the weight in described step (2) adopts definitely important, very important, important, important a little and five ranks of no less important.

3. the evaluation method of the efficient operation of power supply quality according to claim 1, is characterized in that: the calculating in described step (3) and the specifically employing PLS method of revising are calculated and revised.