CN104504127B - Degree of membership defining method and system for classification of power customers - Google Patents

Abstract

A kind of degree of membership defining method for classification of power customers and system, its method includes: obtains number of users and electric power achievement data corresponding to each user, and sets up raw data matrix; Data in described raw data matrix are normalized, it is thus achieved that normalized matrix; Initial cluster center is carried out particle coding by the coded system based on cluster centre, generates initial population; Initial population is set to the parent population of lateral cross, adopts lateral cross method and crossed longitudinally method to calculate the second population; Judge that whether iterations is more than setting iterations, if it is not, the second population is then set to the parent population of lateral cross, continues iteration; If so, particle corresponding for fitness minimum in second population being split as c cluster centre, and calculates the electric power achievement data of each user and the degree of membership of cluster centre, wherein, c represents default clusters number. This programme improves the accuracy rate and stability of determining degree of membership.

Description

Degree of membership defining method and system for classification of power customers

Technical field

The present invention relates to technical field of data processing, particularly relate to a kind of degree of membership defining method for classification of power customers and system.

Background technology

When carrying out classification of power customers, often electric power achievement data is processed. Electric power achievement data is a kind of achievement data in electric power data, such as, during using power consumption as classification of power customers index, then electric power achievement data can include power consumption, during using electric pressure as classification of power customers index, then electric power achievement data can include electric pressure. The analysis method of fuzzy clustering being introduced in classification of power customers is the new research direction of present stage a kind of comparison. It is determined by the electric power achievement data of user and the degree of membership of cluster centre, such that it is able to realize the classification to power consumer. Wherein, FCM cluster is a kind of conventional fuzzy clustering method, although FCM cluster has simple and quick advantage, but traditional FCM clustering procedure excessively relies on initial cluster center; FCM cluster process is based on gradient and declines, it is easy to be absorbed in local optimum; Traditional FCM clustering procedure classification time complexity is higher. Thus causing that the electric power achievement data of user is low with the accuracy rate of the degree of membership of cluster centre.

Summary of the invention

Based on this, it is necessary to for the problem that the electric power achievement data of user is low with the accuracy rate of the degree of membership of cluster centre, it is provided that a kind of degree of membership defining method for classification of power customers and system.

A kind of degree of membership defining method for classification of power customers, including:

From Electric Power Marketing System, obtain number of users and electric power achievement data corresponding to each user, and set up raw data matrix according to described number of users and electric power achievement data;

Data in described raw data matrix are normalized, it is thus achieved that normalized matrix;

Initial cluster center is carried out particle coding by the coded system based on cluster centre, generates initial population;

Initial population is set to the parent population of lateral cross, calculates the adaptive value of each particle in the parent population of described lateral cross according to described normalized matrix;

The adaptive value of the parent population according to lateral cross and its particle adopts lateral cross method to calculate the first population, first population is set to crossed longitudinally parent population, calculate the adaptive value of each particle in described crossed longitudinally parent population according to described normalized matrix, adopt crossed longitudinally method to calculate the second population according to the adaptive value of crossed longitudinally parent population and its particle;

Judge that whether iterations is more than setting iterations, if not, then the second population is set to the parent population of lateral cross, calculate the adaptive value of each particle in the parent population of this lateral cross according to described normalized matrix, adopt lateral cross method and crossed longitudinally method to be iterated calculating the second population; If so, particle corresponding for fitness minimum in second population being split as c cluster centre, and calculates the electric power achievement data of each user and the degree of membership of cluster centre, wherein, c represents default clusters number.

A kind of degree of membership for classification of power customers determines system, including:

Raw data matrix sets up module, for the electric power achievement data that acquisition number of users from Electric Power Marketing System and each user are corresponding, and sets up raw data matrix according to described number of users and electric power achievement data;

Normalized module, for being normalized the data in described raw data matrix, it is thus achieved that normalized matrix;

Initial population generation module, for initial cluster center being carried out particle coding based on the coded system of cluster centre, generates initial population;

Degree of membership determines module, for initial population is set to the parent population of lateral cross, calculates the adaptive value of each particle in the parent population of described lateral cross according to described normalized matrix; The adaptive value of the parent population according to lateral cross and its particle adopts lateral cross method to calculate the first population, first population is set to crossed longitudinally parent population, calculate the adaptive value of each particle in described crossed longitudinally parent population according to described normalized matrix, adopt crossed longitudinally method to calculate the second population according to the adaptive value of crossed longitudinally parent population and its particle; Judge that whether iterations is more than setting iterations, if not, then the second population is set to the parent population of lateral cross, calculate the adaptive value of each particle in the parent population of this lateral cross according to described normalized matrix, adopt lateral cross method and crossed longitudinally method to be iterated calculating the second population; If so, particle corresponding for fitness minimum in second population being split as c cluster centre, and calculates the electric power achievement data of each user and the degree of membership of cluster centre, wherein, c represents default clusters number.

The above-mentioned degree of membership defining method for classification of power customers and system, by setting up raw data matrix, and be normalized, and follow-up is processed by normalized matrix, it is possible to improves treatment effeciency. initial cluster center is carried out particle coding by the coded system based on cluster centre, generates initial population. when first time iteration, initial population is set to the parent population of lateral cross, the adaptive value of each particle in the parent population of described lateral cross is calculated according to described normalized matrix, and adopt lateral cross method and crossed longitudinally method to calculate the second population, when being not reaching to set iterations, second population is set to the parent population of lateral cross, the adaptive value of each particle in the parent population of this lateral cross is calculated according to described normalized matrix, lateral cross method and crossed longitudinally method is adopted to be iterated calculating the second population, until reaching to set iterations, now can obtain the second population of optimum, particle corresponding for fitness minimum in second population is split as c cluster centre, thus obtaining the cluster centre of optimum, and calculate the electric power achievement data of each user and the degree of membership of cluster centre, improve the accuracy rate and stability of determining degree of membership.

Accompanying drawing explanation

Fig. 1 is the present invention schematic flow sheet for the degree of membership defining method embodiment of classification of power customers;

Fig. 2 is the present invention for the degree of membership of classification of power customers determines the structural representation of system embodiment.

Detailed description of the invention

Below in conjunction with embodiment and accompanying drawing, the present invention is described in further detail, but embodiments of the present invention are not limited to this.

As it is shown in figure 1, be used for the schematic flow sheet of the degree of membership defining method embodiment of classification of power customers for the present invention, including:

Step S101: obtain number of users and electric power achievement data corresponding to each user from Electric Power Marketing System, and set up raw data matrix according to described number of users and electric power achievement data;

Electric power achievement data is a kind of achievement data in electric power data. Original matrix X=(x can be set up_ab)_n��d. (1, n), n represents power consumer quantity altogether to a ��, and (1, d), d represents the quantity of electric power achievement data corresponding to each user, x to b ��_abRepresent the electric power achievement data of a user, b row.

Step S102: the data in described raw data matrix are normalized, it is thus achieved that normalized matrix;

This step in order that simplify subsequent calculations, improve computational efficiency.

Further, step S102 may include that

Below equation is adopted to calculate normalization data:

${x^{\′}}_{\mathrm{ab}}=\frac{x_{\mathrm{ab}}-\min x_{*b}}{\max x_{*b}-\min x_{*b}}$

Wherein, (1, n), n represents number of users to a ��, and (1, d), d represents the quantity of electric power achievement data corresponding to each user, minx to b ��_*bRepresent in raw data matrix minimum power achievement data, maxx in b row_*bRepresent in raw data matrix maximum power achievement data, x' in b row_abRepresent the normalization data corresponding to electric power achievement data of a user, b row;

Normalized matrix is obtained according to each normalization data. Such as, it is possible to X'=(x'_ij)_n��dRepresent normalized matrix.

Step S103: initial cluster center is carried out particle coding by the coded system based on cluster centre, generates initial population;

This step forms initial population based on the coded system of cluster centre, and namely each particle is formed (the classification number that c is data set) by c cluster centre. Data sample (normalized matrix) vector dimension is d, then the dimension of particle is c �� d, and the coding structure of particle v is:

Step S104: initial population is set to the parent population of lateral cross, calculates the adaptive value of each particle in the parent population of described lateral cross according to described normalized matrix;

In first time iterative process, it is possible to initial population is set to the parent population of lateral cross.

Further, the adaptive value step of each particle in the described parent population calculating described lateral cross according to described normalized matrix, including:

Step S1041: particle each in the parent population of lateral cross is split as c cluster centre respectively, it is thus achieved that the cluster centre that each particle is corresponding;

Step S1042: adopt below equation to calculate the degree of membership between electric power achievement data and the cluster centre that in described normalized matrix, each user is corresponding:

$u_{\mathrm{xy}}=1/\underset{k=1}{\overset{c}{\mathrm{\Σ}}}{\left(\frac{d_{\mathrm{xy}}}{d_{\mathrm{ky}}}\right)}^{\frac{2}{m-1}}$

Wherein, 1��x��c, 1��y��n, n represents that number of users, m represent fuzzy coefficient, d_xyRepresent the Euclidean distance between the electric power achievement data and x-th cluster centre that in described normalized matrix, y-th user is corresponding, d_kyRepresent the Euclidean distance between the electric power achievement data and kth cluster centre that in described normalized matrix, y-th user is corresponding, u_xyRepresent the degree of membership between each electric power achievement data corresponding for user y and cluster centre x in described normalized matrix;

Step S1043: adopt below equation to calculate the adaptive value of each particle in the parent population of lateral cross;

$\mathrm{fit}=\frac{1}{\underset{y=1}{\overset{n}{\mathrm{\Σ}}}\underset{x=1}{\overset{c}{\mathrm{\Σ}}}{\left(u_{\mathrm{xy}}\right)}^m{\left(d_{\mathrm{xy}}\right)}^2+1}$

Wherein, fit represents the adaptive value of the particle that cluster centre is corresponding.

Owing to each particle can be split as c cluster centre, by calculating the degree of membership of electric power achievement data and these cluster centres, and adopt the formula in step S1043 can calculate the adaptive value of this particle.

Step S105: adopt lateral cross method to calculate the first population according to the adaptive value of the parent population of lateral cross and its particle, first population is set to crossed longitudinally parent population, calculate the adaptive value of each particle in described crossed longitudinally parent population according to described normalized matrix, adopt crossed longitudinally method to calculate the second population according to the adaptive value of crossed longitudinally parent population and its particle;

Calculating the adaptive value of each particle in crossed longitudinally parent population can adopt the method for S1041 to S1043 to be determined.

Step S106: judge that whether iterations is more than setting iterations, if it is not, enter step S107, if so, enters step S108;

Step S107: the second population is set to the parent population of lateral cross, calculates the adaptive value of each particle in the parent population of this lateral cross, and returns step S105 according to described normalized matrix;

When first time iteration, the parent population of lateral cross is initial population; During pth time iteration, the second population obtained when the parent population of lateral cross is-1 iteration of pth. Wherein, p >=2. The first population obtained when crossed longitudinally parent population is current iteration.

Step S108: particle corresponding for fitness minimum in the second population being split as c cluster centre, and calculates the electric power achievement data of each user and the degree of membership of cluster centre, wherein, c represents default clusters number.

Particle minimum for fitness in second population is split the cluster centre obtained is Optimal cluster centers. By the degree of membership between electric power achievement data and cluster centre that user each in normalized matrix is corresponding, the degree of membership between electric power achievement data and the cluster centre that each user in raw data matrix is corresponding can be obtained, degree of membership can be divided into a class more than setting electric power achievement data corresponding to degree of membership, and then realize the classification to user.

The present embodiment is by setting up raw data matrix, and is normalized, and follow-up is processed by normalized matrix, it is possible to improve treatment effeciency. initial cluster center is carried out particle coding by the coded system based on cluster centre, generates initial population. when first time iteration, initial population is set to the parent population of lateral cross, the adaptive value of each particle in the parent population of described lateral cross is calculated according to described normalized matrix, and adopt lateral cross method and crossed longitudinally method to calculate the second population, when being not reaching to set iterations, second population is set to the parent population of lateral cross, the adaptive value of each particle in the parent population of this lateral cross is calculated according to described normalized matrix, lateral cross method and crossed longitudinally method is adopted to be iterated calculating the second population, until reaching to set iterations, now can obtain the second population of optimum, particle corresponding for fitness minimum in second population is split as c cluster centre, thus obtaining the cluster centre of optimum, and calculate the electric power achievement data of each user and the degree of membership of cluster centre, improve the accuracy rate and stability of determining degree of membership.

Wherein in an embodiment, a kind of lateral cross method of concrete introduction.Concrete, the adaptive value of the described parent population according to lateral cross and its particle adopts lateral cross method to calculate the first population step, including:

A1: the particle in the parent population of lateral cross is carried out two neither repeated combinations, it is thus achieved that particle pair, and be numbered;

This step is particle individualities all in population to carry out two neither repeat pairing, total N/2 pair, and is numbered, it is therefore an objective to follow-up can take out successively.

When first time iteration, the parent population of lateral cross is initial population; During pth time iteration, the second population obtained when the parent population of lateral cross is-1 iteration of pth. Second population is properly termed as again the crossed longitudinally solution that is dominant, it is possible to use DS_vcRepresent.

A2: order takes out particle pair successively by number, and adopts the below equation d to particle pair under the lateral cross probability set₀Dimension performs horizontal line and intersects:

MS_hc(i,d₀)=r₁��X(i,d₀)+(1-r₁)��X(j,d₀)+c₁��(X(i,d₀)-X(j,d₀))

MS_hc(j,d₀)=r₁��X(j,d₀)+(1-r₂)��X(i,d₀)+c₁��(X(j,d₀)-X(i,d₀))

Wherein, d₀�� (1, D), D=c �� d, d represent the quantity of electric power achievement data corresponding to each user; I �� (1, N); J �� (1, N), N represents initial population particle number; r₁,r₂For the uniform random number on [0,1]; c₁,c₂For the uniform random number between [-1,1]; X (i, d₀) represent i-th particle d₀Dimension data; X (j, d₀) represent jth particle d₀Dimension data; MS_hc(i,d₀) and MS_hc(j,d₀) represent the golden mean of the Confucian school solution after i-th particle and jth particle lateral cross respectively;

Repeat the above steps, it is possible to golden mean of the Confucian school solution corresponding after obtaining this particle centering lateral cross. Again other particles are carried out same treatment, obtain the golden mean of the Confucian school solution that each particle is corresponding.

A3: golden mean of the Confucian school solution corresponding for each particle is saved in the first matrix, calculate the adaptive value of each golden mean of the Confucian school solution in described first matrix, and the adaptive value corresponding with each particle in the parent population of lateral cross compares, particle little for fitness is stored in the first population.

This step in order that perform competition operator, it is thus achieved that after lateral cross being dominant solution DS_hc(i.e. the first population).

Further, a kind of crossed longitudinally method is also disclosed, concrete, described adopt crossed longitudinally method to calculate the second population step according to crossed longitudinally parent population, including:

B1: often one-dimensional by the particle of crossed longitudinally parent population is normalized;

The solution DS that is dominant of lateral cross in crossed longitudinally parent population and current iteration process_hc��

Further, below equation is adopted to be normalized by the often one-dimensional of particle of crossed longitudinally parent population:

$X^k(i,d_1)=\frac{X^{k-1}(i,d_1)-P_{d1\min }}{P_{d1\max }-P_{d1\min }}$

In formula, d₁�� (1, D), P_d1maxIt is d₁The upper limit of dimension control variable, P_d1minIt is d₁The lower limit of dimension control variable, k is current iteration number of times, X^k(i,d₁) represent the d of i-th particle when iterations is k₁Dimension data, X^k-1(i,d₁) represent the d of i-th particle when iterations is k-1₁Dimension data.

B2: each dimension is carried out two neither repeated combinations, and is numbered;

Dimensions all in population are carried out two and neither repeat pairing (total N/2 to) by this step.

B3: order takes out every a pair successively by number;

Take out each to data. Such as, it is possible to be d₁Peacekeeping d₂The data that dimension is corresponding.

B4: adopt below equation to perform crossed longitudinally to each to data according to setting crossed longitudinally rate:

MS_vc(i,d₁)=r X (i, d₁)+(1-r)��X(i,d₂)

In formula, X (i, d₁) represent i-th particle d₁Dimension data, X (i, d₂) represent i-th particle d₂Dimension data, d₁,d₂�� (1, D), d₁��d₂, D=c �� d, d represents the quantity of electric power achievement data corresponding to each user;I �� (1, N); R is the uniform random number on [0,1]; MS_vc(i,d₁) represent i-th particle d₁Dimension offspring data, according to each offspring data of i-th particle obtain i-th particle crossed longitudinally after initial golden mean of the Confucian school solution;

Wherein, i-th particle each in data, MS_vc(i,d₁) represent i-th particle d₁Dimension offspring data, by X (i, d₁) it is updated to MS_vc(i,d₁); The d of i-th particle₂Dimension offspring data is d₂Dimension data X (i, d₂), namely do not update.

Repeat step B4 can obtain each particle crossed longitudinally after initial golden mean of the Confucian school solution.

B5: the initial golden mean of the Confucian school solution after crossed longitudinally for each particle is carried out renormalization, obtain the golden mean of the Confucian school solution of correspondence, and save it in the second matrix, calculate the adaptive value of each golden mean of the Confucian school solution in described second matrix, and compare with the adaptive value of each particle in crossed longitudinally parent population, being stored in the second population by particle little for fitness, wherein, crossed longitudinally parent population is the first population that lateral cross method obtains.

This step performs Competitive Algorithms, it is thus achieved that the solution DS that is dominant after crossed longitudinally_vc��

Further, adopt below equation that the initial golden mean of the Confucian school solution after crossed longitudinally for each particle is carried out renormalization:

MS_vc'(i,d₁)=MS_vc(i,d₁)��(P_d1max-P_d1min)+P_d1min

Wherein, MS_vc(i,d₁) represent i-th particle d₁Dimension offspring data, MS_vc'(i,d₁) represent i-th particle d₁Data after dimension offspring data renormalization.

Each iterative process all can update the first population and the second population, makes the second population finally given optimum. Before whole scheme, it is also possible to set clusters number c, fuzzy coefficient m, Population Size N, control variable number D (i.e. dimension c �� d), maximum iteration time, crossed longitudinally rate p_vc, lateral cross rate p_hc;

Various technical characteristics in embodiment of above can arbitrarily be combined, as long as the combination between feature is absent from conflict or contradiction, but as space is limited, describe one by one, therefore the various technical characteristics in above-mentioned embodiment be arbitrarily combined falling within the scope of this disclosure.

The present invention is directed to the deficiency of traditional fuzzy C mean algorithm, utilize crossover algorithm in length and breadth to find correct cluster centre, effectively accelerate the convergence rate of clustering algorithm, and avoid being absorbed in local optimum. The present invention program may be used for large power customers data carry out mining analysis objective, science, it is achieved that the become more meticulous classification more comprehensive and accurate to large power customers. The present invention controls that parameter is few and easily operated enforcement, and same kind of classification problem is had important using value.

Based on the above-mentioned degree of membership defining method for classification of power customers, the present invention also provides for a kind of degree of membership for classification of power customers and determines system, as in figure 2 it is shown, determine the structural representation of system embodiment for the degree of membership of classification of power customers for the present invention, including:

Raw data matrix sets up module 210, for the electric power achievement data that acquisition number of users from Electric Power Marketing System and each user are corresponding, and sets up raw data matrix according to described number of users and electric power achievement data;

Normalized module 220, for being normalized the data in described raw data matrix, it is thus achieved that normalized matrix;

Initial population generation module 230, for initial cluster center being carried out particle coding based on the coded system of cluster centre, generates initial population;

Degree of membership determines module 240, for initial population is set to the parent population of lateral cross, calculates the adaptive value of each particle in the parent population of described lateral cross according to described normalized matrix;The adaptive value of the parent population according to lateral cross and its particle adopts lateral cross method to calculate the first population, first population is set to crossed longitudinally parent population, calculate the adaptive value of each particle in described crossed longitudinally parent population according to described normalized matrix, adopt crossed longitudinally method to calculate the second population according to the adaptive value of crossed longitudinally parent population and its particle; Judge that whether iterations is more than setting iterations, if not, then the second population is set to the parent population of lateral cross, calculate the adaptive value of each particle in the parent population of this lateral cross according to described normalized matrix, adopt lateral cross method and crossed longitudinally method to be iterated calculating the second population; If so, particle corresponding for fitness minimum in second population being split as c cluster centre, and calculates the electric power achievement data of each user and the degree of membership of cluster centre, wherein, c represents default clusters number.

Wherein in an embodiment, described degree of membership determines module, is used for:

Particle in the parent population of lateral cross is carried out two neither repeated combinations, it is thus achieved that particle pair, and be numbered;

Order takes out particle pair successively by number, and adopts the below equation d to particle pair under the lateral cross probability set₀Dimension performs horizontal line and intersects:

MS_hc(i,d₀)=r₁��X(i,d₀)+(1-r₁)��X(j,d₀)+c₁��(X(i,d₀)-X(j,d₀))

MS_hc(j,d₀)=r₁��X(j,d₀)+(1-r₂)��X(i,d₀)+c₁��(X(j,d₀)-X(i,d₀))

Wherein, d₀�� (1, D), D=c �� d, d represent the quantity of electric power achievement data corresponding to each user; I �� (1, N); J �� (1, N), N represents initial population particle number; r₁,r₂For the uniform random number on [0,1]; c₁,c₂For the uniform random number between [-1,1]; X (i, d₀) represent i-th particle d₀Dimension data; X (j, d₀) represent jth particle d₀Dimension data; MS_hc(i,d₀) and MS_hc(j,d₀) represent the golden mean of the Confucian school solution after i-th particle and jth particle lateral cross respectively;

Golden mean of the Confucian school solution corresponding for each particle is saved in the first matrix, calculate the adaptive value of each golden mean of the Confucian school solution in described first matrix, and the adaptive value corresponding with each particle in the parent population of lateral cross compares, particle little for fitness is stored in the first population.

Wherein in an embodiment, described degree of membership determines module, is used for:

Often one-dimensional by the particle of crossed longitudinally parent population is normalized, and each dimension carries out two neither repeated combinations, and is numbered;

Order takes out every a pair successively by number;

Below equation is adopted to perform crossed longitudinally to each to data according to setting crossed longitudinally rate:

MS_vc(i,d₁)=r X (i, d₁)+(1-r)��X(i,d₂)

In formula, X (i, d₁) represent i-th particle d₁Dimension data, X (i, d₂) represent i-th particle d₂Dimension data, d₁,d₂�� (1, D), d₁��d₂, D=c �� d, d represents the quantity of electric power achievement data corresponding to each user; I �� (1, N); R is the uniform random number on [0,1]; MS_vc(i,d₁) represent i-th particle d₁Dimension offspring data, according to each offspring data of i-th particle obtain i-th particle crossed longitudinally after initial golden mean of the Confucian school solution;

Initial golden mean of the Confucian school solution after crossed longitudinally for each particle is carried out renormalization, obtain the golden mean of the Confucian school solution of correspondence, and save it in the second matrix, calculate the adaptive value of each golden mean of the Confucian school solution in described second matrix, and compare with the adaptive value of each particle in crossed longitudinally parent population, being stored in the second population by particle little for fitness, wherein, crossed longitudinally parent population is the first population that lateral cross method obtains.

The degree of membership for classification of power customers of the present invention determines that the degree of membership defining method for classification of power customers of system and the present invention is one to one, above-mentioned all determine system embodiment suitable in the degree of membership for classification of power customers for the correlation technique feature in the degree of membership defining method embodiment of classification of power customers and technique effect thereof, do not repeat them here.

Embodiment described above only have expressed the several embodiments of the present invention, and it describes comparatively concrete and detailed, but therefore can not be interpreted as the restriction to the scope of the claims of the present invention. It should be pointed out that, for the person of ordinary skill of the art, without departing from the inventive concept of the premise, it is also possible to making some deformation and improvement, these broadly fall into protection scope of the present invention. Therefore, the protection domain of patent of the present invention should be as the criterion with claims.

Claims (10)

1. the degree of membership defining method for classification of power customers, it is characterised in that including:

From Electric Power Marketing System, obtain number of users and electric power achievement data corresponding to each user, and set up raw data matrix according to described number of users and electric power achievement data;

Data in described raw data matrix are normalized, it is thus achieved that normalized matrix;

Initial cluster center is carried out particle coding by the coded system based on cluster centre, generates initial population;

Initial population is set to the parent population of lateral cross, calculates the adaptive value of each particle in the parent population of described lateral cross according to described normalized matrix;

The adaptive value of the parent population according to lateral cross and its particle adopts lateral cross method to calculate the first population, first population is set to crossed longitudinally parent population, calculate the adaptive value of each particle in described crossed longitudinally parent population according to described normalized matrix, adopt crossed longitudinally method to calculate the second population according to the adaptive value of crossed longitudinally parent population and its particle;

Judge to adopt lateral cross method and crossed longitudinally method whether to be iterated calculating the iterations of the second population more than setting iterations, if not, then the second population is set to the parent population of lateral cross, calculate the adaptive value of each particle in the parent population of this lateral cross according to described normalized matrix, adopt lateral cross method and crossed longitudinally method to be iterated calculating the second population; If so, particle corresponding for fitness minimum in second population being split as c cluster centre, and calculates the electric power achievement data of each user and the degree of membership of cluster centre, wherein, c represents default clusters number.

2. the degree of membership defining method for classification of power customers according to claim 1, it is characterised in that the adaptive value of the described parent population according to lateral cross and its particle adopts lateral cross method to calculate the first population step, including:

Particle in the parent population of lateral cross is carried out two neither repeated combinations, it is thus achieved that particle pair, and be numbered;

Order takes out particle pair successively by number, and adopts the below equation d to particle pair under the lateral cross probability set₀Dimension performs horizontal line and intersects:

MS_hc(i,d₀)=r₁��X(i,d₀)+(1-r₁)��X(j,d₀)+c₁��(X(i,d₀)-X(j,d₀))

MS_hc(j,d₀)=r₁��X(j,d₀)+(1-r₂)��X(i,d₀)+c₁��(X(j,d₀)-X(i,d₀))

Wherein, d₀�� (1, D), D=c �� d, d represent the quantity of electric power achievement data corresponding to each user; I �� (1, N); J �� (1, N), N represents initial population particle number; r₁,r₂For the uniform random number on [0,1]; c₁,c₂For the uniform random number between [-1,1]; X (i, d₀) represent i-th particle d₀Dimension data; X (j, d₀) represent jth particle d₀Dimension data; MS_hc(i,d₀) and MS_hc(j,d₀) represent the golden mean of the Confucian school solution after i-th particle and jth particle lateral cross respectively;

Golden mean of the Confucian school solution corresponding for each particle is saved in the first matrix, calculate the adaptive value of each golden mean of the Confucian school solution in described first matrix, and the adaptive value corresponding with each particle in the parent population of lateral cross compares, particle little for fitness is stored in the first population.

3. the degree of membership defining method for classification of power customers according to claim 1, it is characterised in that described adopt crossed longitudinally method to calculate the second population step according to crossed longitudinally parent population, including:

Often one-dimensional by the particle of crossed longitudinally parent population is normalized, and each dimension carries out two neither repeated combinations, and is numbered;

Order takes out every a pair successively by number;

Below equation is adopted to perform crossed longitudinally to each to data according to setting crossed longitudinally rate:

MS_vc(i,d₁)=r X (i, d₁)+(1-r)��X(i,d₂)

In formula, X (i, d₁) represent i-th particle d₁Dimension data, X (i, d₂) represent i-th particle d₂Dimension data, d₁,d₂�� (1, D), d₁��d₂, D=c �� d, d represents the quantity of electric power achievement data corresponding to each user; I �� (1, N); R is the uniform random number on [0,1]; MS_vc(i,d₁) represent i-th particle d₁Dimension offspring data, according to each offspring data of i-th particle obtain i-th particle crossed longitudinally after initial golden mean of the Confucian school solution;

Initial golden mean of the Confucian school solution after crossed longitudinally for each particle is carried out renormalization, obtain the golden mean of the Confucian school solution of correspondence, and save it in the second matrix, calculate the adaptive value of each golden mean of the Confucian school solution in described second matrix, and compare with the adaptive value of each particle in crossed longitudinally parent population, being stored in the second population by particle little for fitness, wherein, crossed longitudinally parent population is the first population that lateral cross method obtains.

4. according to claims 1 to 3 any one for the degree of membership defining method of classification of power customers, it is characterised in that described data in described raw data matrix are normalized, it is thus achieved that normalized matrix step includes:

Below equation is adopted to calculate normalization data:

${x^{\′}}_{ab}=\frac{x_{ab}-{\mathrm{minx}}_{*b}}{{\mathrm{maxx}}_{*b}-{\mathrm{minx}}_{*b}}$

Wherein, (1, n), n represents number of users to a ��, and (1, d), d represents the quantity of electric power achievement data corresponding to each user, minx to b ��_*bRepresent in raw data matrix minimum power achievement data, maxx in b row_*bRepresent in raw data matrix maximum power achievement data, x' in b row_abRepresent the normalization data corresponding to electric power achievement data of a user, b row;

Normalized matrix is obtained according to each normalization data.

5. the degree of membership defining method for classification of power customers according to claims 1 to 3 any one, it is characterised in that the adaptive value step of each particle in the described parent population calculating described lateral cross according to described normalized matrix, including:

Particle each in the parent population of lateral cross is split as c cluster centre respectively, it is thus achieved that the cluster centre that each particle is corresponding;

Below equation is adopted to calculate the degree of membership between electric power achievement data and the cluster centre that in described normalized matrix, each user is corresponding:

$u_{xy}=1/\underset{k=1}{\overset{c}{\mathrm{\Σ}}}{\left(\frac{d_{xy}}{d_{ky}}\right)}^{\frac{2}{m-1}}$

Wherein, 1��x��c, 1��y��n, n represents that number of users, m represent fuzzy coefficient, d_xyRepresent the Euclidean distance between the electric power achievement data and x-th cluster centre that in described normalized matrix, y-th user is corresponding, d_kyRepresent the Euclidean distance between the electric power achievement data and kth cluster centre that in described normalized matrix, y-th user is corresponding, u_xyRepresent the degree of membership between each electric power achievement data corresponding for user y and cluster centre x in described normalized matrix;

Below equation is adopted to calculate the adaptive value of each particle in the parent population of lateral cross;

$fit=\frac{1}{\underset{y=1}{\overset{n}{\Σ}}\underset{x=1}{\overset{c}{\Σ}}{\left(u_{xy}\right)}^m{\left(d_{xy}\right)}^2+1}$

Wherein, fit represents the adaptive value of the particle that cluster centre is corresponding.

6. the degree of membership defining method for classification of power customers according to claim 3, it is characterised in that adopt below equation to be normalized by the often one-dimensional of particle of crossed longitudinally parent population:

$X^k(i,d_1)=\frac{X^{k-1}(i,d_1)-P_{d1min}}{P_{d1max}-P_{d1min}}$

In formula, d₁�� (1, D), P_d1maxIt is d₁The upper limit of dimension control variable, P_d1minIt is d₁The lower limit of dimension control variable, k is current iteration number of times, X^k(i,d₁) represent the d of i-th particle when iterations is k₁Dimension data, X^k-1(i,d₁) represent the d of i-th particle when iterations is k-1₁Dimension data.

7. the degree of membership defining method for classification of power customers according to claim 6, it is characterised in that adopt below equation that the initial golden mean of the Confucian school solution after crossed longitudinally for each particle is carried out renormalization:

MS_vc'(i,d₁)=MS_vc(i,d₁)��(P_d1max-P_d1min)+P_d1min

Wherein, MS_vc(i,d₁) represent i-th particle d₁Dimension offspring data, MS_vc'(i,d₁) represent i-th particle d₁Data after dimension offspring data renormalization.

8. determine system for the degree of membership of classification of power customers for one kind, it is characterised in that including:

Raw data matrix sets up module, for the electric power achievement data that acquisition number of users from Electric Power Marketing System and each user are corresponding, and sets up raw data matrix according to described number of users and electric power achievement data;

Normalized module, for being normalized the data in described raw data matrix, it is thus achieved that normalized matrix;

Initial population generation module, for initial cluster center being carried out particle coding based on the coded system of cluster centre, generates initial population;

Degree of membership determines module, for initial population is set to the parent population of lateral cross, calculates the adaptive value of each particle in the parent population of described lateral cross according to described normalized matrix; The adaptive value of the parent population according to lateral cross and its particle adopts lateral cross method to calculate the first population, first population is set to crossed longitudinally parent population, calculate the adaptive value of each particle in described crossed longitudinally parent population according to described normalized matrix, adopt crossed longitudinally method to calculate the second population according to the adaptive value of crossed longitudinally parent population and its particle; Judge to adopt lateral cross method and crossed longitudinally method whether to be iterated calculating the iterations of the second population more than setting iterations, if not, then the second population is set to the parent population of lateral cross, calculate the adaptive value of each particle in the parent population of this lateral cross according to described normalized matrix, adopt lateral cross method and crossed longitudinally method to be iterated calculating the second population; If so, particle corresponding for fitness minimum in second population being split as c cluster centre, and calculates the electric power achievement data of each user and the degree of membership of cluster centre, wherein, c represents default clusters number.

9. the degree of membership for classification of power customers according to claim 8 determines system, it is characterised in that described degree of membership determines module, is used for:

Particle in the parent population of lateral cross is carried out two neither repeated combinations, it is thus achieved that particle pair, and be numbered;

Order takes out particle pair successively by number, and adopts the below equation d to particle pair under the lateral cross probability set₀Dimension performs horizontal line and intersects:

MS_hc(i,d₀)=r₁��X(i,d₀)+(1-r₁)��X(j,d₀)+c₁��(X(i,d₀)-X(j,d₀))

MS_hc(j,d₀)=r₁��X(j,d₀)+(1-r₂)��X(i,d₀)+c₁��(X(j,d₀)-X(i,d₀))

Wherein, d₀�� (1, D), D=c �� d, d represent the quantity of electric power achievement data corresponding to each user; I �� (1, N); J �� (1, N), N represents initial population particle number; r₁,r₂For the uniform random number on [0,1];C₁,c₂For the uniform random number between [-1,1]; X (i, d₀) represent i-th particle d₀Dimension data; X (j, d₀) represent jth particle d₀Dimension data; MS_hc(i,d₀) and MS_hc(j,d₀) represent the golden mean of the Confucian school solution after i-th particle and jth particle lateral cross respectively;

Golden mean of the Confucian school solution corresponding for each particle is saved in the first matrix, calculate the adaptive value of each golden mean of the Confucian school solution in described first matrix, and the adaptive value corresponding with each particle in the parent population of lateral cross compares, particle little for fitness is stored in the first population.

10. the degree of membership for classification of power customers according to claim 8 determines system, it is characterised in that described degree of membership determines module, is used for:

Often one-dimensional by the particle of crossed longitudinally parent population is normalized, and each dimension carries out two neither repeated combinations, and is numbered;

Order takes out every a pair successively by number;

Below equation is adopted to perform crossed longitudinally to each to data according to setting crossed longitudinally rate:

MS_vc(i,d₁)=r X (i, d₁)+(1-r)��X(i,d₂)

In formula, X (i, d₁) represent i-th particle d₁Dimension data, X (i, d₂) represent i-th particle d₂Dimension data, d₁,d₂�� (1, D), d₁��d₂, D=c �� d, d represents the quantity of electric power achievement data corresponding to each user; I �� (1, N); R is the uniform random number on [0,1]; MS_vc(i,d₁) represent i-th particle d₁Dimension offspring data, according to each offspring data of i-th particle obtain i-th particle crossed longitudinally after initial golden mean of the Confucian school solution;

Initial golden mean of the Confucian school solution after crossed longitudinally for each particle is carried out renormalization, obtain the golden mean of the Confucian school solution of correspondence, and save it in the second matrix, calculate the adaptive value of each golden mean of the Confucian school solution in described second matrix, and compare with the adaptive value of each particle in crossed longitudinally parent population, being stored in the second population by particle little for fitness, wherein, crossed longitudinally parent population is the first population that lateral cross method obtains.