CN105488589A - Genetic simulated annealing algorithm based power grid line loss management evaluation method - Google Patents

Abstract

The invention discloses a genetic simulated annealing algorithm based power grid line loss management evaluation method. The method comprises: step 1: modeling a factor influencing the line loss of a power grid and establishing an N-dimensional index vector; step 2: randomly selecting k samples from a generated sample set and performing floating-point coding on chromosomes; step 3: performing initialization; step 4: performing sample clustering on initial populations according to clustering centers and calculating a fitness value of each individual; step 5: performing selection, crossing and mutation operations on individuals in the populations, calculating the fitness values of the individuals, and performing simulated annealing algorithm operation to generate a new population; step 6: when an evolutional frequency is less than a maximum evolutional frequency S, returning to the step 5; otherwise, going to the step 7; step 7: if the algorithm is ended, decoding optimal individuals to obtain optimal clustering number and clustering center, performing sample clustering, and dividing a line loss management level of a to-be-evaluated power grid into different levels; and step 8: performing evaluation according to a line loss value.

Description

A kind of grid line loss administrative evaluation method based on Global Genetic Simulated Annealing Algorithm

Technical field

The present invention relates to a kind of grid line loss administrative evaluation method based on Global Genetic Simulated Annealing Algorithm.

Background technology

Due to the difference of each side such as electric network composition, socio-economic factor, there is larger difference in the line loss level between different electrical network, in order to more objective evaluation Controlling line loss level, several echelon need be marked off according to the different attribute of electrical network, electrical network like Attribute class is gathered in an echelon and evaluates, separately between each echelon do not interfere with each other, it is in the nature a kind of cluster.

K-MEANS algorithm is a kind of simple and effective clustering algorithm, but because it have employed heuristic in the process finding cluster centre, comparatively responsive to the selection of initial cluster center, and, this algorithm needs to specify cluster number in advance, and often accurately cannot learn the cluster number of sample in actual applications, therefore, traditional K-MEANS clustering algorithm is easily absorbed in locally optimal solution.

Genetic algorithm (GeneticAlgorithm, GA) be the computation model of simulating the natural selection of Darwinian evolutionism and the biological evolution process of genetic mechanisms, it is a kind of method by simulating nature evolutionary process search optimum solution, which introduce the operations such as selection, crossover and mutation, optimum solution can be obtained in a random way, but its local optimal searching scarce capacity, and there is precocious defect.Simulated annealing (SimulatedAnnealing, SA) from a certain higher initial temperature, with the continuous decline of temperature parameter, locally optimal solution can be probability jump out and be finally tending towards global optimum, there is very strong local search ability.Clustering algorithm based on Global Genetic Simulated Annealing Algorithm then combines the advantage of genetic algorithm and simulated annealing, can obtain globally optimal solution with larger probability.

Summary of the invention

For the problems referred to above, the invention provides a kind of grid line loss administrative evaluation method based on Global Genetic Simulated Annealing Algorithm, cluster number in partition process dynamic conditioning in algorithm operational process is determined, compensate for the deficiency that traditional K-MEANS clustering algorithm is responsive to initial cluster center, be difficult to determine in advance cluster number, the division result of global optimum can be obtained with larger probability, to grid line loss administrative evaluation work, there is certain reference value.

For realizing above-mentioned technical purpose, reach above-mentioned technique effect, the present invention is achieved through the following technical solutions:

Based on a grid line loss administrative evaluation method for Global Genetic Simulated Annealing Algorithm, it is characterized in that, comprise the steps:

Step one: mathematical modeling is carried out on the factor affecting grid line loss level, comprise electric pressure and level, circuit average length, sectional area of wire, distribution transforming status of equipment, reactive-load compensation configuration, duration of load application distribution, unit power transformation capacity load rate, the maximum natural load or burden without work coefficient of electrical network, dividing potential drop electricity sales amount, can't harm electricity accounting, rural area area accounting, non-industrial GDP accounting and power density, and set up grid line loss N dimension indicator vector according to application needs and the quality of data, wherein, 1≤N≤13;

Step 2: in the sample set that grid line loss N dimension indicator vector generates, randomly draw k sample as initial cluster center, adopt the chromosome floating-point code mode based on cluster centre, every bar chromosomal coding C is: C=kc _λ... c _k, gene c _λbe certain grid line loss N dimensional vector that λ cluster centre is corresponding, the division evaluation method of the individual corresponding a kind of Controlling line loss level of each dyeing;

Step 3: initialization algorithm controling parameters, comprises population at individual number M, maximum evolution number of times S, crossover probability P _c, mutation probability P _m, temperature cooling ratio k _c, annealing initial temperature T ₀, annealing final temperature T _end;

Step 4: to the variant individuality of initial population, respectively according to cluster centre, carries out sample clustering according to Euclidean distance minimum principle, calculates the fitness value of variant individuality;

Step 5: select the individuality in population, intersect, mutation operation, calculates its fitness value to the new individuality produced, carries out simulated annealing operation, generate new population;

Step 6: when evolution number of times is less than maximum evolution number of times S, return step 5; Otherwise, forward step 7 to;

Step 7: if temperature index is lower than final temperature T _end, then algorithm stops, and decodes to the optimum individual in existing population, obtains best cluster number and cluster centre, then carries out sample clustering by the Controlling line loss horizontal division of electrical network to be evaluated in different echelons; If temperature index is higher than final temperature T _end, then perform cooling and operate and return step 5;

Step 8: in the same echelon of line loss management level, the size according to line loss value is evaluated, and the height of the corresponding Controlling line loss level of height of line loss value, evaluates separate between each echelon.

First by calculating line loss influence factor index, form grid line loss indicator vector sample set, randomly draw some samples as initial cluster center, carry out the floating-point code of variable cluster centre, and cluster number is added in chromosome coding as first gene position, solve optimum individual and decode to determine cluster number and cluster centre, by the Controlling line loss horizontal division of electrical network to be evaluated in different echelons, in same echelon, carry out administrative evaluation according to the size of line loss value, its management level are more excellent for less then the thinking of line loss value.The present invention has merged the thought of genetic algorithm and simulated annealing in the process evaluated, eliminate responsive to initial cluster center in traditional clustering algorithm, to be easily absorbed in local optimum defect, and without the need to specifying cluster number in advance, the similar division result of the line loss of global optimum can be found with larger probability.

Preferably, in step 4, for k cluster centre selected in chromosome, by grid line loss indicator vector x _lcluster centre c is included into according to the principle that Euclidean distance is minimum _λ, that is:

||x _l-c _λ||＝min||x _l-c _j||(j＝1,2,...,k)

Wherein, x _lfor the N dimensional vector that each influence factor index of grid line loss is formed, and x _l=(x _l1, x _l2..., x _lN), x _l1~ x _lNcorrespond respectively to N number of influence factor index value.

Wherein, the cluster object definition fitness that, between class distance minimum according to inter-object distance is maximum is wherein $J=\frac{\underset{\mathrm{\λ}=1}{\overset{k}{\Σ}}\underset{x_l\∈c_{\mathrm{\λ}}}{\Σ}\left|\right|x_l-c_{\mathrm{\λ}}\left|\right|}{1+\underset{m\≠n}{\Σ}\left|\right|c_m-c_n\left|\right|}.$

The invention has the beneficial effects as follows: the clustering algorithm based on genetic simulated annealing is applied in grid line loss administrative evaluation work by the present invention, first electrical network to be evaluated is divided into several different echelon, then carries out management level evaluation according to the size of line loss value in same echelon.Cluster number in partition process dynamic conditioning in algorithm operational process is determined, compensate for the deficiency that traditional K-MEANS clustering algorithm is responsive to initial cluster center, be difficult to determine in advance cluster number, the division result of global optimum can be obtained with larger probability, to grid line loss administrative evaluation work, there is certain reference value.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of a kind of grid line loss administrative evaluation method based on Global Genetic Simulated Annealing Algorithm of the present invention.

Embodiment

Below in conjunction with accompanying drawing and specific embodiment, technical solution of the present invention is described in further detail, can better understand the present invention to make those skilled in the art and can be implemented, but illustrated embodiment is not as a limitation of the invention.

Based on a grid line loss administrative evaluation method for Global Genetic Simulated Annealing Algorithm, as shown in Figure 1, comprise the steps:

Step one: mathematical modeling is carried out on the factor affecting grid line loss level, comprise electric pressure and level, circuit average length, sectional area of wire, distribution transforming status of equipment, reactive-load compensation configuration, duration of load application distribution, unit power transformation capacity load rate, the maximum natural load or burden without work coefficient of electrical network, dividing potential drop electricity sales amount, can't harm electricity accounting, rural area area accounting, non-industrial GDP accounting and power density, and set up grid line loss N dimension indicator vector according to application needs and the quality of data, wherein, 1≤N≤13;

Step 2: in the sample set that grid line loss N dimension indicator vector generates, randomly draw k sample as initial cluster center, adopt the chromosome floating-point code mode based on cluster centre, every bar chromosome is made up of the coding of k cluster centre and clusters number k, and to be C be the chromosome floating-point code based on cluster centre: C=kc _λ... c _k, gene c _λbe certain grid line loss N dimensional vector that λ cluster centre is corresponding, k is also this chromosomal cluster number, the division evaluation method of the individual corresponding a kind of Controlling line loss level of each dyeing;

Step 3: initialization algorithm controling parameters, comprises population at individual number M, maximum evolution number of times S, crossover probability P _c, mutation probability P _m, temperature cooling ratio k _c, annealing initial temperature T ₀, annealing final temperature T _end;

Step 4: to the variant individuality of initial population, respectively according to cluster centre, carries out sample clustering according to Euclidean distance minimum principle, calculates the fitness value of variant individuality;

Preferably, in this step, for k cluster centre selected in chromosome, by grid line loss indicator vector x _lcluster centre c is included into according to the principle that Euclidean distance is minimum _λ, that is:

||x _l-c _λ||＝min||x _l-c _j||(j＝1,2,...,k)

Wherein, x _lfor the N dimensional vector that each influence factor index of grid line loss is formed, and x _l=(x _l1, x _l2..., x _lN), x _l1~ x _lNcorrespond respectively to N number of influence factor index value.

The maximum cluster object definition fitness of, between class distance minimum according to inter-object distance is wherein $J=\frac{\underset{\mathrm{\λ}=1}{\overset{k}{\Σ}}\underset{x_l\∈c_{\mathrm{\λ}}}{\Σ}\left|\right|x_l-c_{\mathrm{\λ}}\left|\right|}{1+\underset{m\≠n}{\Σ}\left|\right|c_m-c_n\left|\right|}.$

Step 5: select the individuality in population, intersect, mutation operation, calculates its fitness value to the new individuality produced, carries out simulated annealing operation, generate new population;

Step 6: when evolution number of times is less than maximum evolution number of times S, return step 5; Otherwise, forward step 7 to;

Step 7: if temperature index is lower than final temperature T _end, then algorithm stops, and decodes to the optimum individual in existing population, obtains best cluster number and cluster centre, then carries out sample clustering by the Controlling line loss horizontal division of electrical network to be evaluated in different echelons; If temperature index is higher than final temperature T _end, then perform cooling and operate and return step 5;

Step 8: in the same echelon of line loss management level, the size according to line loss value is evaluated, and the height of the corresponding Controlling line loss level of height of line loss value, evaluates separate between each echelon.

General, in step one, mathematical modeling is carried out on the factor affecting grid line loss level, is specially:

(1) electric pressure and level Y _dYDJcalculated with mathematical model formula be:

Y _DYDJ＝ΣLoss _i

Wherein, each electric pressure in i=500 ~ 10kV; Loss _ifor i electric pressure dividing potential drop line loss per unit in electrical network gamut to be divided;

(2) circuit average length Y _xLCDcalculated with mathematical model formula be:

$Y_{XLCD}=\frac{\ΣL_i\×{10}^2/i^2}{\ΣN_i}$

Wherein: L _ifor the line length under electrical network i electric pressure to be divided; N _ifor electrical network i electric pressure circuit total number to be divided;

(3) sectional area of wire Y _dXJMcalculated with mathematical model formula:

Y _DXJM＝ΣLoss _iq _θ(q _iaL _iθa+q _ibL _iθb+q _icL _iθc)

Wherein: θ is wire type sequence number, when θ=1, represent pole line, during θ=2, represent cable; q _θfor wire type weight, by the length ratio of pole line and cable, if certain electrical network is without cable, then q ₁get 100%; Loss _ifor i electric pressure dividing potential drop line loss per unit in electrical network gamut to be divided; L _{i θ a}, L _{i θ b}, L _{i θ c}for the i electric pressure overhead transmission line of electrical network to be divided or the cross section of cable and the ratio of electrical network i electric pressure line length to be divided, wire is divided into I cross section, II cross section, cross section, three kinds, III cross section by sectional area size, the sectional area that wherein sectional area in I cross section is less than the sectional area in II cross section, the sectional area in II cross section is less than III cross section, a, b, c represent I cross section, II cross section, III cross section respectively; q _ia, q _ib, q _icrepresent i electric pressure a respectively, weight coefficient that b, c cross section wire affects line loss.

(4) distribution transforming status of equipment Y _pBZKcalculated with mathematical model formula:

Y _PBZK＝q _dT _d+q _eT _e+q _fT _f

Wherein: T _d, T _e, T _fbe respectively the capacity of distribution transform ratio of high consumption model, common model, energy-conservation model, high consumption model is below S7, and common model is S9 ~ S11, and energy-conservation model is more than S11, and the capacity merging of wherein non-crystaline amorphous metal change and single-phase change is included into more than S11 model and is participated in calculating;

(5) reactive-load compensation configuration Y _wGBCcalculated with mathematical model formula be:

Y _WGBC＝ΣLoss _i(1-W _i)

Wherein: Loss _ifor i electric pressure dividing potential drop line loss per unit in electrical network gamut to be divided; W _ifor the i electric pressure reactive-load compensation configuration coefficients of each electrical network, main transformer configure the ratio of condenser capacity and main transformer capacity;

(6) load space distribution Y _fHSJcalculated with mathematical model formula:

$Y_{FHSJ}=\underset{i=1}{\overset{12}{\Σ}}{\mathrm{\ψ}}_i$

Wherein: ψ _ifor the moon load degree of uniformity of each electrical network, the ratio of front-month maximum peak-valley difference and monthly average load.

(7) unit power transformation capacity load rate Y _zBFZcalculated with mathematical model formula be:

$Y_{ZBFZ}=\frac{{\mathrm{\η}}_i{G}_i}{G_{500}+G_{330}+G_{220}+G_{110}}$

${\mathrm{\η}}_i=\frac{P_{0i}}{P_i/B_i}+\frac{P_{1i}}{S_i^2}\×\frac{P_i}{B_i}$

Wherein: G _ifor power transmission amount under i electric pressure transformer; G ₅₀₀, G ₃₃₀, G ₂₂₀, G ₁₁₀the lower power transmission amount of corresponding 500kV, 330kV, 220kV, 110kV electric pressure transformer respectively; η _ifor the proportion of goods damageds of i electric pressure transformer; P _0ifor the representative value of i electric pressure transformer noload losses; P _1ifor the representative value of i electric pressure transformer load loss; S _ifor the representative value of i electric pressure transformer rated capacity; P _ifor sending power under i electric pressure transformer; B _ifor the capacity of i electric pressure transformer;

(8) maximum natural load or burden without work coefficient Y _zRWGcalculated with mathematical model formula:

$Y_{ZRWG}=1/\frac{Q}{P}$

Wherein: Q is the maximum reactive power capability of electrical network, P is the maximum tracking burden with power of electrical network;

The maximum reactive power capability of electrical network is:

Q＝Q _G+Q _C+Q _R+Q _L

Wherein: Q _gfor the reactive power of generator, Q _cfor capacitive reactive power compensates total volume, Q _rfor adjacent net inputs or outputs idle, Q _lfor the charge power of circuit and cable;

(9) dividing potential drop electricity sales amount Y _fYDLindex calculate formula:

$Y_{FYDL}=\Σ\frac{A_i{\mathrm{Loss}}_i}{A}$

Wherein: Loss _ifor i electric pressure dividing potential drop line loss per unit in electrical network gamut to be divided; A _ifor electrical network i electric pressure delivery; A is total delivery of electrical network;

(10) harmless electricity Y _wSDLcalculated with mathematical model formula:

Y _WSDL＝1-W

Wherein: W is that each electrical network can't harm electricity accounting;

(11) rural area area accounting Y _nCMJcalculated with mathematical model formula be:

Y _NCMJ＝S’

Wherein: S ' is each electrical network rural area area accounting;

(12) non-industrial GDP accounting Y _fGYcalculated with mathematical model formula be:

Y _FGY＝G

Wherein: G is the non-industrial GDP accounting of each electrical network;

(13) power density Y _gDMDcalculated with mathematical model formula be:

$Y_{GDMD}=\frac{A}{\mathrm{\η}}$

Wherein: A is total delivery of electrical network, η is each mains supply area.

Set forth the similar splitting scheme of line loss of the present invention for the sake of simplicity, suppose that dividing object is electrical network A, electrical network B, electrical network C, electrical network D, electrical network E and electrical network F, then a concrete embodiment is as follows:

S1, mathematical modeling is carried out on all kinds of factors affecting grid line loss level, set up grid line loss multidimensional index vector.Influence factor comprises electric pressure and level, circuit average length, sectional area of wire, distribution transforming status of equipment, reactive-load compensation configuration, duration of load application distribution, unit power transformation capacity load rate, maximum natural load or burden without work coefficient, dividing potential drop electricity sales amount, can't harm electricity, rural area area accounting, non-industrial GDP accounting and power density, sets up N dimension indicator vector (1≤N≤13) according to application needs and the quality of data;

For simplified schematic, choose electric pressure and level, circuit average length, sectional area of wire, distribution transforming status of equipment four influence factors, set up 4 dimension indicator vectors according to mathematical formulae mentioned above, be respectively:

X ₁=(x ₁₁, x ₁₂, x ₁₃, x ₁₄), x ₂=(x ₂₁, x ₂₂, x ₂₃, x ₂₄), x ₃=(x ₃₁, x ₃₂, x ₃₃, x ₃₄), x ₄=(x ₄₁, x ₄₂, x ₄₃, x ₄₄), x ₅=(x ₅₁, x ₅₂, x ₅₃, x ₅₄), x ₆=(x ₆₁, x ₆₂, x ₆₃, x ₆₄), with x ₁for example, x ₁₁, x ₁₂, x ₁₃, x ₁₄be respectively electric pressure and level index, circuit average length index, sectional area of wire index and the distribution transforming status of equipment index of electrical network A.X ₂, x ₃, x ₄, x ₅and x ₆then be respectively the indicator vector of electrical network B, electrical network C, electrical network D, electrical network E and electrical network F.

S2, grid line loss multidimensional index vector generate sample set in, randomly draw k sample as initial cluster center, adopt the floating-point code mode based on cluster centre, every bar chromosome is made up of the coding of k cluster centre and clusters number k; The division evaluation method of the individual corresponding a kind of Controlling line loss level of each dyeing;

Such as, randomly drawing sample x ₄and x ₆as initial cluster center, then chromosome 1 is encoded to 2x ₄x ₆; If sample drawn x ₁, x ₃and x ₅as cluster centre, then chromosome 2 is encoded to 3x ₁x ₃x ₅; If sample drawn x ₁, x ₃as cluster centre, then chromosome 3 is encoded to 2x ₁x ₃.

S3, initialization algorithm controling parameters, comprise population at individual number M, maximum evolution number of times S, crossover probability P _c, mutation probability P _m, temperature cooling ratio k _c, annealing initial temperature T ₀, annealing final temperature T _end;

S4, variant individuality to initial population, respectively according to cluster centre, carry out sample clustering by Euclidean distance minimum principle, calculate the fitness value of variant individuality;

For chromosome 2, calculate all the other grid line loss indicator vectors x respectively ₂, x ₄and x ₆apart from the Euclidean distance of each cluster centre, and carry out cluster according to the principle that Euclidean distance is minimum, can obtain:

||x ₂-x ₃||＝min||x ₂-x _j||(j＝1,3,5)

||x ₄-x ₁||＝min||x ₄-x _j||(j＝1,3,5)

||x ₆-x ₃||＝min||x ₆-x _j||(j＝1,3,5)

Namely initial clustering result is: { x ₁, x ₄, { x ₂, x ₃, x ₆, { x ₅.

The cluster result of chromosome 1 and 3 is respectively: { x ₃, x ₄, x ₅, { x ₂, x ₁, x ₆and { x ₁, x ₄, x ₆, { x ₂, x ₃, x ₅.

The cluster object definition fitness that, between class distance minimum according to inter-object distance is maximum is again wherein, in population, 3 chromosomes are respectively calculated as follows:

$J_1=\frac{\underset{\mathrm{\λ}=1}{\overset{k}{\Σ}}\underset{x_l\∈c_{\mathrm{\λ}}}{\Σ}\left|\right|x_l-c_{\mathrm{\λ}}\left|\right|}{1+\underset{m\≠n}{\Σ}\left|\right|c_m-c_n\left|\right|}=\frac{\left|\right|x_3-x_4\left|\right|+\left|\right|x_5-x_4\left|\right|+\left|\right|x_1-x_6\left|\right|+\left|\right|x_2-x_6\left|\right|}{1+\left|\right|x_4-x_6\left|\right|}$

$J_2=\frac{\underset{\mathrm{\λ}=1}{\overset{k}{\Σ}}\underset{x_l\∈c_{\mathrm{\λ}}}{\Σ}\left|\right|x_l-c_{\mathrm{\λ}}\left|\right|}{1+\underset{m\≠n}{\Σ}\left|\right|c_m-c_n\left|\right|}=\frac{\left|\right|x_4-x_1\left|\right|+\left|\right|x_2-x_3\left|\right|+\left|\right|x_6-x_3\left|\right|}{1+\left|\right|x_1-x_3\left|\right|+\left|\right|x_1-x_5\left|\right|+\left|\right|x_3-x_5\left|\right|}$

$J_3=\frac{\underset{\mathrm{\λ}=1}{\overset{k}{\Σ}}\underset{x_l\∈c_{\mathrm{\λ}}}{\Σ}\left|\right|x_l-c_{\mathrm{\λ}}\left|\right|}{1+\underset{m\≠n}{\Σ}\left|\right|c_m-c_n\left|\right|}=\frac{\left|\right|x_4-x_1\left|\right|+\left|\right|x_6-x_1\left|\right|+\left|\right|x_2-x_3\left|\right|+\left|\right|x_5-x_3\left|\right|}{1+\left|\right|x_1-x_3\left|\right|}.$

S5, the individuality in population selected, intersect, mutation operation, its fitness value is calculated to the new individuality produced, carries out simulated annealing operation, generate new population;

According to the chromosomal fitness value sequence that S4 calculates be: f ₂> f ₁> f ₃,

Select operation: chromosome 2 can be chosen by roulette wheel selection and chromosome 1 enters population of future generation;

Interlace operation: can adopt single-point bracketing method, random generation natural number 2, as point of crossing, produces a Probability p at random, is greater than the crossover probability P in S3 as p _ctime, intersect from chromosome the 2nd gene position, and first gene position newly individual according to the length amendment after intersecting.The new chromosome coding that this step produces is 2x ₁x ₆, 3x ₄x ₃x ₅;

Mutation operation: produce a natural number 3 at random as change point, random generation Probability p ', the mutation probability P in S3 is greater than as p ' _mtime, chromosomal 3rd gene position is made a variation, replaces with the sample vector that the random number produced is corresponding, suppose that the new individual UVR exposure that variation produces is 3x ₄x ₁x ₅;

Calculate new 3 the chromosomal fitness value f ' produced _i, carry out simulated annealing operation, adopt Metropolis acceptance criterion, if f ' _i> f _i, then the old individuality of new individual replacement is accepted, if f ' _i< f _i, then with probability accept new individual to replace old individuality, k is Boltzmann constant, generates new population to be: chromosome 1 is encoded 3x ₁x ₃x ₅, chromosome 2 encodes 3x ₄x ₁x ₅, chromosome 3 encodes 2x ₁x ₆.

S6, when evolution number of times is less than maximum evolution number of times, return S5; Otherwise, forward S7 to;

If S7 temperature index is lower than final temperature, then algorithm stops, and decodes to the optimum individual in existing population, obtains best cluster number and cluster centre, then carries out sample clustering by the Controlling line loss horizontal division of electrical network to be evaluated in different echelons; If temperature index is higher than final temperature, then performs cooling and operate and return S5;

When temperature index is lower than final temperature T _endtime, algorithm stops, and now existing Evolution of Population is: chromosome 1 is encoded 3x ₁x ₃x ₅, chromosome 2 encodes 3x ₄x ₁x ₆, chromosome 3 encodes 2x ₁x ₄, calculate fitness value sequence for f ₃> f ₁> f ₂, then chromosome 3 is optimum individual, and can obtain best cluster centre number to its decoding is 2, and cluster centre is sample x ₁and x ₄.Divide 6 grid line loss situations according to this cluster condition, can obtain best division result is echelon 1{x ₁, x ₃and echelon 2{x ₂, x ₄, x ₅, x ₆.

S8, in the same echelon of line loss management level, the size according to line loss value is evaluated, and the electrical network that line loss value is less then thinks that Controlling line loss level is more excellent, evaluates separate between each echelon;

For echelon 1, if the line loss value of electrical network 3 is less than electrical network 1, then think that the Controlling line loss level of electrical network 3 is relatively excellent.

Clustering algorithm based on genetic simulated annealing is applied in grid line loss administrative evaluation work by the present invention, first electrical network to be evaluated is divided into several different echelon, then carries out management level evaluation according to the size of line loss value in same echelon.Cluster number in partition process dynamic conditioning in algorithm operational process is determined, when the between class distance of cluster is identical, the inter-object distance based on the clustering algorithm of Global Genetic Simulated Annealing Algorithm can be starkly lower than K-MEANS algorithm, and effect is more excellent.Compensate for the deficiency that traditional K-MEANS clustering algorithm is responsive to initial cluster center, be difficult to determine in advance cluster number, the division result of global optimum can be obtained with larger probability, to grid line loss administrative evaluation work, there is certain reference value.

These are only the preferred embodiments of the present invention; not thereby the scope of the claims of the present invention is limited; every utilize instructions of the present invention and accompanying drawing content to do equivalent structure or equivalent flow process conversion; or be directly or indirectly used in the technical field that other are relevant, be all in like manner included in scope of patent protection of the present invention.

Claims (6)

1., based on a grid line loss administrative evaluation method for Global Genetic Simulated Annealing Algorithm, it is characterized in that, comprise the steps:

Step one: mathematical modeling is carried out on the factor affecting grid line loss level, comprise electric pressure and level, circuit average length, sectional area of wire, distribution transforming status of equipment, reactive-load compensation configuration, duration of load application distribution, unit power transformation capacity load rate, the maximum natural load or burden without work coefficient of electrical network, dividing potential drop electricity sales amount, can't harm electricity accounting, rural area area accounting, non-industrial GDP accounting and power density, and set up grid line loss N dimension indicator vector according to application needs and the quality of data, wherein, 1≤N≤13;

Step 2: in the sample set that grid line loss N dimension indicator vector generates, randomly draw k sample as initial cluster center, adopt the chromosome floating-point code mode based on cluster centre, every bar chromosomal coding C is: C=kc _λ... c _k, gene c _λbe certain grid line loss N dimensional vector that λ cluster centre is corresponding, the division evaluation method of the individual corresponding a kind of Controlling line loss level of each dyeing;

Step 3: initialization algorithm controling parameters, comprises population at individual number M, maximum evolution number of times S, crossover probability P _c, mutation probability P _m, temperature cooling ratio k _c, annealing initial temperature T ₀, annealing final temperature T _end; Step 4: to the variant individuality of initial population, respectively according to cluster centre, carries out sample clustering according to Euclidean distance minimum principle, calculates the fitness value of variant individuality;

Step 5: select the individuality in population, intersect, mutation operation, calculates its fitness value to the new individuality produced, carries out simulated annealing operation, generate new population;

Step 6: when evolution number of times is less than maximum evolution number of times S, return step 5; Otherwise, forward step 7 to;

Step 7: if temperature index is lower than final temperature T _end, then algorithm stops, and decodes to the optimum individual in existing population, obtains best cluster number and cluster centre, then carries out sample clustering by the Controlling line loss horizontal division of electrical network to be evaluated in different echelons; If temperature index is higher than final temperature T _end, then perform cooling and operate and return step 5;

Step 8: in the same echelon of line loss management level, the size according to line loss value is evaluated, and the height of the corresponding Controlling line loss level of height of line loss value, evaluates separate between each echelon.

2. a kind of grid line loss administrative evaluation method based on Global Genetic Simulated Annealing Algorithm according to claim 1, is characterized in that, in step 4, for k cluster centre selected in chromosome, by grid line loss indicator vector x _lcluster centre c is included into according to the principle that Euclidean distance is minimum _λ, that is:

||x _l-c _λ||＝min||x _l-c _j||(j＝1,2,...,k)

Wherein, x _lfor the N dimensional vector that each influence factor index of grid line loss is formed, and x _l=(x _l1, x _l2..., x _lN), x _l1~ x _lNcorrespond respectively to N number of influence factor index value.

3. a kind of grid line loss administrative evaluation method based on Global Genetic Simulated Annealing Algorithm according to claim 2, is characterized in that, in step 4, the maximum cluster object definition fitness of, between class distance minimum according to inter-object distance is $f=\frac{1}{1+J},$ Wherein $J=\frac{\underset{\mathrm{\&lambda;}=1}{\overset{k}{\&Sigma;}}\underset{x_l\&Element;c_{\mathrm{\&lambda;}}}{\&Sigma;}\left|\right|x_l-c_{\mathrm{\&lambda;}}\left|\right|}{1+\underset{m\&NotEqual;n}{\&Sigma;}\left|\right|c_m-c_n\left|\right|}.$

4. a kind of grid line loss administrative evaluation method based on Global Genetic Simulated Annealing Algorithm according to claim 1, is characterized in that, carry out mathematical modeling in step one on the factor affecting grid line loss level, wherein: electric pressure and level Y _dYDJcalculated with mathematical model formula be:

Y _DYDJ＝ΣLoss _i

Wherein, each electric pressure in i=500 ~ 10kV; Loss _ifor i electric pressure dividing potential drop line loss per unit in electrical network gamut to be divided;

Circuit average length Y _xLCDcalculated with mathematical model formula be:

$Y_{XLCD}=\frac{{\mathrm{\&Sigma;L}}_i\&times;{10}^2/i^2}{{\mathrm{\&Sigma;N}}_i}$

Wherein: L _ifor the line length under electrical network i electric pressure to be divided; N _ifor electrical network i electric pressure circuit total number to be divided;

Sectional area of wire Y _dXJMcalculated with mathematical model formula:

Y _DXJM＝ΣLoss _iq _θ(q _iaL _iθa+q _ibL _iθb+q _icL _iθc)

Wherein: θ is wire type sequence number, when θ=1, represent pole line, during θ=2, represent cable; q _θfor wire type weight, by the length ratio of pole line and cable, if certain electrical network is without cable, then q ₁get 100%; Loss _ifor i electric pressure dividing potential drop line loss per unit in electrical network gamut to be divided; L _{i θ a}, L _{i θ b}, L _{i θ c}for the i electric pressure overhead transmission line of electrical network to be divided or the cross section of cable and the ratio of electrical network i electric pressure line length to be divided, wire is divided into I cross section, II cross section, cross section, three kinds, III cross section by sectional area size, the sectional area that wherein sectional area in I cross section is less than the sectional area in II cross section, the sectional area in II cross section is less than III cross section, a, b, c represent I cross section, II cross section, III cross section respectively; q _ia, q _ib, q _icrepresent i electric pressure a respectively, weight coefficient that b, c cross section wire affects line loss.

5. a kind of grid line loss administrative evaluation method based on Global Genetic Simulated Annealing Algorithm according to claim 4, is characterized in that, carry out mathematical modeling in step one on the factor affecting grid line loss level, wherein: distribution transforming status of equipment Y _pBZKcalculated with mathematical model formula:

Y _PBZK＝q _dT _d+q _eT _e+q _fT _f

Wherein: T _d, T _e, T _fbe respectively the capacity of distribution transform ratio of high consumption model, common model, energy-conservation model, high consumption model is below S7, and common model is S9 ~ S11, and energy-conservation model is more than S11, and the capacity merging of wherein non-crystaline amorphous metal change and single-phase change is included into more than S11 model and is participated in calculating; Reactive-load compensation configuration Y _wGBCcalculated with mathematical model formula be:

Y _WGBC＝ΣLoss _i(1-W _i)

Wherein: Loss _ifor i electric pressure dividing potential drop line loss per unit in electrical network gamut to be divided; W _ifor the i electric pressure reactive-load compensation configuration coefficients of each electrical network, main transformer configure the ratio of condenser capacity and main transformer capacity;

Load space distribution Y _fHSJcalculated with mathematical model formula:

$Y_{FHSJ}=\underset{i=1}{\overset{12}{\&Sigma;}}{\mathrm{\&psi;}}_i$

Wherein: ψ _ifor the moon load degree of uniformity of each electrical network, the ratio of front-month maximum peak-valley difference and monthly average load.

6. a kind of grid line loss administrative evaluation method based on Global Genetic Simulated Annealing Algorithm according to claim 5, is characterized in that, carry out mathematical modeling in step one on the factor affecting grid line loss level, wherein: unit power transformation capacity load rate Y _zBFZcalculated with mathematical model formula be:

$Y_{ZBFZ}=\frac{{\mathrm{\&eta;}}_i{G}_i}{G_{500}+G_{330}+G_{220}+G_{110}}$

${\mathrm{\&eta;}}_i=\frac{P_{0i}}{P_i/B_i}+\frac{P_{1i}}{S_i^2}\&times;\frac{P_i}{B_i}$

Wherein: G _ifor power transmission amount under i electric pressure transformer; G ₅₀₀, G ₃₃₀, G ₂₂₀, G ₁₁₀the lower power transmission amount of corresponding 500kV, 330kV, 220kV, 110kV electric pressure transformer respectively; η _ifor the proportion of goods damageds of i electric pressure transformer; P _0ifor the representative value of i electric pressure transformer noload losses; P _1ifor the representative value of i electric pressure transformer load loss; S _ifor the representative value of i electric pressure transformer rated capacity; P _ifor sending power under i electric pressure transformer; B _ifor the capacity of i electric pressure transformer; Maximum natural load or burden without work coefficient Y _zRWGcalculated with mathematical model formula:

$Y_{ZRWG}=1/\frac{Q}{P}$

Wherein: Q is the maximum reactive power capability of electrical network, P is the maximum tracking burden with power of electrical network;

The maximum reactive power capability of electrical network is:

Q＝Q _G+Q _C+Q _R+Q _L

Wherein: Q _gfor the reactive power of generator, Q _cfor capacitive reactive power compensates total volume, Q _rfor adjacent net inputs or outputs idle, Q _lfor the charge power of circuit and cable;

Dividing potential drop electricity sales amount Y _fYDLindex calculate formula:

$Y_{FYDL}=\mathrm{\&Sigma;}\frac{A_i{\mathrm{Loss}}_i}{A}$

Wherein: Loss _ifor i electric pressure dividing potential drop line loss per unit in electrical network gamut to be divided; A _ifor electrical network i electric pressure delivery; A is total delivery of electrical network;

Harmless electricity Y _wSDLcalculated with mathematical model formula:

Y _WSDL＝1-W

Wherein: W is that each electrical network can't harm electricity accounting;

Rural area area accounting Y _nCMJcalculated with mathematical model formula be:

Y _NCMJ＝S’

Wherein: S ' is each electrical network rural area area accounting;

Non-industrial GDP accounting Y _fGYcalculated with mathematical model formula be:

Y _FGY＝G

Wherein: G is the non-industrial GDP accounting of each electrical network;

Power density Y _gDMDcalculated with mathematical model formula be:

$Y_{GDMD}=\frac{A}{\mathrm{\&eta;}}$

Wherein: A is total delivery of electrical network, η is each mains supply area.