CN101840200A - Adaptive processing method for optimizing dynamic data in dispatching control - Google Patents

Abstract

The invention provides an adaptive processing method for optimizing dynamic data in dispatching control. The method comprises the following steps of: outputting the preprocessed data to a main-control center by wavelet analysis, selecting different adaptive dispatching algorithms by using a controller and establishing a genetic algorithm in which a fuzzy rule is embedded to perform optimization operation, comparing the generated result with the result in a simulator, and selecting a dispatching scheme meeting the requirements as a system output; and if the deviation between the generated result and the result in the simulator is relatively great, correcting by using an identifier, entering the controller of the main-control center and training, and adjusting the algorithm until the deviation between the result outputted by the main-control center and the result in the simulator meets the requirements. In the method, after the actual data is preprocessed, by introducing an adaptive control mechanism, the slow change of the characteristics of the target is adapted to, the actual requirements of the dispatching control are preprocessed, and the method has certain advancement of the prediction of the future requirements, and simultaneously contributes to the whole realization of the optimization of a closed-loop supply chain.

Description

The adaptive processing method of dynamic data in a kind of Optimization Dispatching control

Technical field

The present invention relates to the adaptive processing method of dynamic data in a kind of Optimization Dispatching control.

Background technology

The Optimization Dispatching problem is prevalent in each research field, and people progressively go deep into the research of optimisation technique, and the method for solution optimization problem commonly used has exact algorithm and intelligent optimization algorithm.

The mathematic(al) representation minf (X) of optimization method, X ∈ E ⁿAnd satisfy certain constraint condition:

H (X)=0, g (X) 〉=0.X ∈ E ⁿX is expressed as the point (vector) in the theorem in Euclid space.

Continuity according to function f (X) can be divided into optimization problem two big classes: the optimization of continuous function and the optimization of discrete function.The latter also can be called combinatorial optimization problem.It is feasible theoretically that many combinatorial optimization problems are arranged, but in fact infeasible, and therefore, this class problem is called a NP-Hard difficult problem.Usually the traditional operational research Methods of utilization can only solve the optimal control scheduling problem of static data.And be a difficult problem of puzzlement for the processing of the dynamic data in the actual production always.

Summary of the invention

The present invention proposes the adaptive processing method of dynamic data in a kind of Optimization Dispatching control, the computing and the coupling between the data of its dynamic data are relatively accurate, the speed of service data is very fast, realize optimizing to the scheduling of the dynamic data under uncertain environment with based on the genetic algorithm of fuzzy rule, this method is simple and convenient.

The adaptive processing method of dynamic data in a kind of Optimization Dispatching control of the present invention, mainly be the raw data of input to be carried out wavelet analysis by wave filter, after involving noise after filtration and separating, with pretreated data output in the Master Control Center according to dynamically, static, services sets carries out classification and handles; After this Master Control Center receives static and dynamic data, select different adaptive scheduling algorithms and set up the genetic algorithm that embeds fuzzy rule data are optimized computing by controller wherein, and the data result that generates and the model in the emulator and data compared, if meet re-set target, then select satisfactory scheduling scheme to export as system; If the data result and the result error in the emulator that generate are bigger, do not meet re-set target, then the data result is revised by system model correction identifier, enter the controller training of Master Control Center simultaneously, and the algorithm in this controller adjusted, till the data result of Master Control Center output and the deviation between the result in the emulator meet re-set target.

Described adaptive scheduling algorithm has selected based on the fuzzy logic algorithm that quantizes fuzzy rule problem to be divided into critical path and non-critical path, and its concrete steps are as follows:

Step 1: the membership of calculating fuzzy factors

K=1,2,3, and round values Vac, Vmc, Vmnc;

Described Vac is meant the quantity that increases critical path, and Vmc is meant diminishbb client's quantity, and Vmnc is meant the quantity of diminishbb non-critical path;

Critical path collection and non-critical path collection are defined as V respectively _cAnd V _Nc, then the adaptive network dispatching algorithm is as follows:

Step (1): from node i=1 ..., n demarcates initial time T respectively _sWith concluding time T _e

Step (2): calculate and introduce comprehensive evaluation coefficient θ,

$\mathrm{\θ}=\mathrm{\γ}\·\underset{i=1}{\overset{N}{\mathrm{\Σ}}}(\mathrm{Te}-\mathrm{Ts})+\mathrm{\δ}\·\underset{i=1}{\overset{N}{\mathrm{\Σ}}}r+\mathrm{\σ}\·\frac{Q}{\mathrm{\ΣQ}}$

In the formula

R---node radius; Q---demand

Step (3): to n, determine that whether node is at critical path collection V from node i=1 _cWith non-critical path collection V _Nc

Step (4): calculating target function value f (x)

$f\left(x\right)=\mathrm{\β}(T_{\mathrm{ij}}-{\mathrm{MT}}_{\mathrm{ij}})+\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{d}_{\mathrm{ij}}$

Step 2: the degree of membership of calculating the fuzzy decision rule

K=1,2,3,4; Specific algorithm is as follows:

If

Be fuzzy decision D _kDegree of membership, k=1,2,3,4, consider of the influence of three fuzzy factors to this decision-making, utilize decision-making expert's knowledge and experience to obtain:

${\mathrm{\&mu;}}_{{\mathrm{<figure-callout\; id="1"\; label="D"\; filenames="HSA00000055604400031.png"\; state="\{\{state\}\}">D</figure-callout>}}_1}={\mathrm{\&mu;}}_{{\tilde{F}}_1}^c\&CircleTimes;{\mathrm{\&mu;}}_{{\tilde{F}}_2}$

${\mathrm{\&mu;}}_{D_2}={\mathrm{\&mu;}}_{{\mathrm{<figure-callout\; id="1"\; label="F"\; filenames="HSA00000055604400031.png"\; state="\{\{state\}\}">F</figure-callout>}}_1}\&CircleTimes;{\mathrm{\&mu;}}_{F_3}$

${\mathrm{\&mu;}}_{D_3}={\mathrm{\&mu;}}_{{\tilde{F}}_1}\&CircleTimes;{\mathrm{\&mu;}}_{F_2}^c\&CircleTimes;{\mathrm{\&mu;}}_{F_3}^c$

${\mathrm{\&mu;}}_{D_4}={\mathrm{\&mu;}}_{F_1}^c\&CircleTimes;{\mathrm{\&mu;}}_{F_2}^c\&CircleTimes;{\mathrm{\&mu;}}_{F_3}^c$

Wherein, μ ^c=1-μ, Symbol is an operator

${\mathrm{\&mu;}}_1\&CircleTimes;{\mathrm{\&mu;}}_2=\frac{{\mathrm{\&mu;}}_1{\mathrm{\&mu;}}_2}{\mathrm{\&delta;}+(1-\mathrm{\&delta;})({\mathrm{\&mu;}}_1+{\mathrm{\&mu;}}_2-{\mathrm{\&mu;}}_1{\mathrm{\&mu;}}_2)}$

Step 3: calculate selected decision-making k

From obtaining

Then decision-making can be drawn by following formula,

$k=\arg \max ({\mathrm{\&mu;}}_{{\mathrm{<figure-callout\; id="1"\; label="D"\; filenames="HSA00000055604400031.png"\; state="\{\{state\}\}">D</figure-callout>}}_1},{\mathrm{\&mu;}}_{D_2},{\mathrm{\&mu;}}_{D_3},{\mathrm{\&mu;}}_{D_4})$

Step 4: the path to the decision-making selected and selection adds respectively, reducing, if decision-making then is add operation for critical path, non-critical path then is reducing, and the adaptive network dispatching algorithm in the invocation step 1, dispatches the calculating target function value again;

Step 5: calculating target function value: f (x) specifically is calculated as:

Step 6: return f (x) and selection result X=[X1, X2 ..., Xi ..., Xn] to step 5;

Genetic algorithm and specific implementation step that described foundation embeds fuzzy rule comprise:

Step 1, determine the gene code mode,, genetic operator, select to calculate slightly and stopping criterion just when function

(1) gene code mode: be natural number coding at the client;

(2) just when function

f(l)＝Fmax-F(1)+a，1＝1，2…，NP (1)

Described NP is a population number, Fmax=max{F (1), and l=1,2 ..., NP}, a are little positive numbers;

(3) genetic operator: adopt 2 bracketing methods, the center section of promptly getting two chromosome correspondences carries out corresponding exchange, and mutation operation is the change that the chromosomal part of random choose is worth;

(4) select to calculate slightly: adopt gyroscope wheel method and elite's selection strategy, elite's selection strategy is meant selects a gene arbitrarily in population of new generation, replace with gene best among the previous generation, and the selection probability of the 3rd gene is in a new generation:

$P_k\left(i\right)=\frac{f_k-f_k\left(i\right)}{\underset{i=1}{\overset{\mathrm{NP}}{\mathrm{\&Sigma;}}}\{f_k-f_k\left(i\right)\}}$ f _k＝max{f _k(i)，i＝1，2，…，NP}

(2)

(5) stopping criterion: adopt the stopping criterion of greatest iteration number as algorithm;

Step 2, based on the solution procedure of the genetic algorithm that embeds fuzzy rule

(1) setup parameter: population scale NP, greatest iteration algebraically NG, crossover probability Pc, variation probability P m;

(2) call the adaptive network dispatching algorithm, calculate comprehensive evaluation coefficient θ, and definite critical path and non-critical path;

(3) it is as follows to produce NP chromosomal initial population at random:

Z=[X1, X2 ... Xi ..., X (N)], Xj (i) ≠ Xk (i), any j, k, i=1,2 ..., NP

Wherein Xj (i) is a natural number that is not more than Ni, establishes in fact that the number of iterations of population is k=0, and initial feasible solution is X*=X (1), and objective function F *=Q, Q are a big integer;

(4) satisfy commentaries on classics (9) if judge whether to satisfy end condition, otherwise change (5);

(5) for chromosome x (i), i=1,2 ..., NP, carry out as follows:

Operation 1: call the adaptive network dispatching algorithm, calculate comprehensive evaluation coefficient θ;

Operation 2: call the fuzzy decision rule, return target function value f (i) and memory selection X (i);

Operation 3: the target function value that record is maximum, minimum

fmax＝max{f(Xi)，i＝1，2，...，NP}

fmin＝min{f(Xi)，i＝1，2，...，NP}

Make i ^*=arg{f (Xi)=fmin}, X (i ^*) be the chromosome of minimum target functional value

(6)IF?f ^*≥fmin，then?f ^*＝fmin，X ^*＝X(i ^*)；

(7) calculate just when function and every gene just when, calculate it according to the adaptive value of gene and select probability, and select the big or small picked at random gene of probability to duplicate;

(8) intersect and the computing that makes a variation, produce new generation population, and substitute a new gene at random, change (4) with gene best among the previous generation;

(9) output result, record f* and X* finish as optimum solution.

Result in the controlling schemes that described result who generates by the genetic algorithm optimization of the fuzzy rule that embeds and practical problems require compares, usually adopt least square method that the data in the system model are revised, thereby the algorithm in the Master Control Center is adjusted, adjust back output and optimize the parameter of result, by realize training and adjustment as the intersection factor of parameter and mutagenic factor algorithm in the Master Control Center as Master Control Center.

Because the present invention carries out wavelet analysis by filtrator to raw data, filter and realize the noise separation, help the processing of the mass data in the scheduling controlling practical problems, can improve the accuracy of the Dynamic Data Processing of Optimization Dispatching algorithm; By the design Master Control Center, select different adaptive algorithms, the data after handling are optimized the random search algorithm, try to achieve operation result; The prediction of pre-service and tomorrow requirement is carried out in the current demand of scheduling controlling certain advance, helps the integral body of closed loop supply chain optimization to realize simultaneously; The result who has generated is set up mathematical model, thereby set up emulator, do the check of related data; By the data dispatching in Master Control Center design-calculated operation result and the emulator is compared, thereby the data in the system model are revised, formed the result of dynamically optimized scheduling.

The present invention is on the operating mechanism basis of static scheduling system, proposition is based on the multiple-objection optimization dispatching system of the Adaptive Identification type of intelligent algorithm, both guaranteed the optimal selection of scheduling scheme, can obtain scheduling scheme again, determine the characteristic of real system, as dispensing delivery period, transportation cost according to practical problems, offer emulator and realize accurately simulation real system, by introducing identifier, as the self-correcting controlling mechanism, the slow variation of the dynamic data of adaption object characteristic; In addition, the present invention is in system's input raw data stage, through the pre-service of filtrator realization to real data, self-adaptive controlled by reference making mechanism, but the slow variation of adaption object characteristic.

Description of drawings

Fig. 1 is a workflow diagram of the present invention;

Fig. 2 is the fundamental diagram of Master Control Center of the present invention;

Fig. 3 is applied to the workflow diagram of path optimization's scheduling in the city logistics delivery system for the present invention;

Fig. 4 is applied to the functional block diagram of path optimization's scheduling in the city logistics delivery system for the present invention;

Fig. 5 is applied to the operational flowchart of path optimization's scheduling in the city logistics delivery system for the present invention.

The invention will be further described below in conjunction with embodiment.

Embodiment

As shown in Figure 1, 2, the adaptive processing method of dynamic data in a kind of Optimization Dispatching control of the present invention, mainly be the raw data of input to be carried out wavelet analysis by wave filter, after involving noise after filtration and separating, with pretreated data output in the Master Control Center according to dynamically, static, services sets carries out classification and handles; After this Master Control Center receives static and dynamic data, select different adaptive scheduling algorithms and set up the genetic algorithm that embeds fuzzy rule data are optimized computing by controller wherein, and the data result that generates and the model in the emulator and data compared, if meet re-set target, then select satisfactory scheduling scheme to export as system; If the data result and the result error in the emulator that generate are bigger, do not meet re-set target, then the data result is revised by system model correction identifier, enter the controller training of Master Control Center simultaneously, and the algorithm in this controller adjusted, till the data result of Master Control Center output and the deviation between the result in the emulator meet re-set target.

Below algorithm steps be for the dispatching algorithm design of the fuzzy rule set up, the i.e. scheduling rule that in the controller of Master Control Center, designs, the scheduling that this algorithm helps to solve the dynamic data under the uncertain factor condition is handled, and the scheduling rule of foundation and mechanism help the solution of practical problems.Here selected problem to be divided into critical path and non-critical path based on the fuzzy logic algorithm that quantizes fuzzy rule, thus the adaptive network dispatching algorithm of setting up.Its concrete steps are as follows:

Step 1: the membership of calculating fuzzy factors

K=1,2,3, and round values Vac, Vmc, Vmnc;

Described Vac is meant the quantity that increases critical path, and Vmc is meant diminishbb client's quantity, and Vmnc is meant the quantity of diminishbb non-critical path;

Critical path collection and non-critical path collection are defined as V respectively _cAnd V _Nc, the adaptive network dispatching algorithm is as follows:

Step (1): from node i=1 ..., n demarcates initial time T respectively _sWith concluding time T _e

Step (2): calculate and introduce comprehensive evaluation coefficient θ,

$\mathrm{\&theta;}=\mathrm{\&gamma;}\&CenterDot;\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}(\mathrm{Te}-\mathrm{Ts})+\mathrm{\&delta;}\&CenterDot;\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}r+\mathrm{\&sigma;}\&CenterDot;\frac{Q}{\mathrm{\&Sigma;Q}}$

In the formula

R---node radius; Q---demand.

Step (3): to n, determine that whether node is at critical path collection V from node i=1 _cWith non-critical path collection V _Nc

Step (4): calculating target function value f (x)

$f\left(x\right)=\mathrm{\&beta;}(T_{\mathrm{ij}}-{\mathrm{MT}}_{\mathrm{ij}})+\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_{\mathrm{ij}}$

Step 2: the degree of membership of calculating the fuzzy decision rule

K=1,2,3,4; Specific algorithm is as follows:

If

Be fuzzy decision D _kDegree of membership, k=1,2,3,4.Consider of the influence of three fuzzy factors, utilize decision-making expert's knowledge and experience to obtain decision-making:

${\mathrm{\&mu;}}_{{\mathrm{<figure-callout\; id="1"\; label="D"\; filenames="HSA00000055604400031.png"\; state="\{\{state\}\}">D</figure-callout>}}_1}={\mathrm{\&mu;}}_{{\tilde{F}}_1}^c\&CircleTimes;{\mathrm{\&mu;}}_{{\tilde{F}}_2}$

${\mathrm{\&mu;}}_{D_2}={\mathrm{\&mu;}}_{{\mathrm{<figure-callout\; id="1"\; label="F"\; filenames="HSA00000055604400031.png"\; state="\{\{state\}\}">F</figure-callout>}}_1}\&CircleTimes;{\mathrm{\&mu;}}_{F_3}$

${\mathrm{\&mu;}}_{D_3}={\mathrm{\&mu;}}_{{\tilde{F}}_1}\&CircleTimes;{\mathrm{\&mu;}}_{F_2}^c\&CircleTimes;{\mathrm{\&mu;}}_{F_3}^c$

${\mathrm{\&mu;}}_{D_4}={\mathrm{\&mu;}}_{F_1}^c\&CircleTimes;{\mathrm{\&mu;}}_{F_2}^c\&CircleTimes;{\mathrm{\&mu;}}_{F_3}^c$

Wherein, μ ^c=1-μ,

Symbol is an operator

${\mathrm{\&mu;}}_1\&CircleTimes;{\mathrm{\&mu;}}_2=\frac{{\mathrm{\&mu;}}_1{\mathrm{\&mu;}}_2}{\mathrm{\&delta;}+(1-\mathrm{\&delta;})({\mathrm{\&mu;}}_1+{\mathrm{\&mu;}}_2-{\mathrm{\&mu;}}_1{\mathrm{\&mu;}}_2)}$

Step 3: calculate selected decision-making k.

From obtaining

Then decision-making can be drawn by following formula,

$k=\arg \max ({\mathrm{\&mu;}}_{{\mathrm{<figure-callout\; id="1"\; label="D"\; filenames="HSA00000055604400031.png"\; state="\{\{state\}\}">D</figure-callout>}}_1},{\mathrm{\&mu;}}_{D_2},{\mathrm{\&mu;}}_{D_3},{\mathrm{\&mu;}}_{D_4})$

Step 4: the path to the decision-making selected and selection adds respectively, reducing, if decision-making then is add operation for critical path, non-critical path then is reducing, and the adaptive network dispatching algorithm in the invocation step 1, dispatches the calculating target function value again.

Step 5: calculating target function value: f (x) specifically is calculated as:

Step 6: return f (x) and selection result X=[X1, X2 ..., Xi ..., Xn] to step 5.

Following arthmetic statement goes out the adaptive scheduling algorithm's who sets up at the controller of Master Control Center performing step, and the genetic algorithm of promptly selecting to set up based on fuzzy rule is selected more excellent process of separating.The fuzzy scheduling rule that above-mentioned first section algorithm selected is the performing step that embeds second section genetic algorithm, and for preparing algorithm, the two is a continuous relationship.

Genetic algorithm and specific implementation step that described foundation embeds fuzzy rule comprise:

Step 1, determine the gene code mode,, genetic operator, select to calculate slightly and stopping criterion just when function

(1) gene code mode: be natural number coding at the client;

(2) just when function

f(l)＝Fmax-F(l)+a，l＝1，2...，NP (1)

Described NP is a population number, Fmax=max{F (l), and l=1,2 ..., NP}, a are little positive numbers.

(3) genetic operator: adopt 2 bracketing methods, the center section of promptly getting two chromosome correspondences carries out corresponding exchange, and mutation operation is the change that the chromosomal part of random choose is worth.

(4) select to calculate slightly: adopt gyroscope wheel method and elite's selection strategy, elite's selection strategy is meant selects a gene arbitrarily in population of new generation, replace with gene best among the previous generation, and the selection probability of the 3rd gene is in a new generation:

$P_k\left(i\right)=\frac{f_k-f_k\left(i\right)}{\underset{i=1}{\overset{\mathrm{NP}}{\mathrm{\&Sigma;}}}\{f_k-f_k\left(i\right)\}}$ f _k＝max{f _k(i)，i＝1，2，…，NP}

(2)

(5) stopping criterion: adopt the stopping criterion of greatest iteration number as algorithm.

Step 2, based on the solution procedure of the genetic algorithm that embeds fuzzy rule

(1) setup parameter: population scale NP, greatest iteration algebraically NG, crossover probability Pc, variation probability P m;

(2) call the adaptive network dispatching algorithm, calculate comprehensive evaluation coefficient θ, and definite critical path and non-critical path;

(3) it is as follows to produce NP chromosomal initial population at random:

Z=[X1, X2 ... Xi ..., X (N)], Xj (i) ≠ Xk (i), any j, k, i=1,2 ..., NP

Wherein Xj (i) is a natural number that is not more than Ni, establishes in fact that the number of iterations of population is k=0, and initial feasible solution is X*=X (1), and objective function F *=Q, Q are a big integer;

(4) satisfy commentaries on classics (9) if judge whether to satisfy end condition, otherwise change (5);

(5) for chromosome x (i), i=1,2 ..., NP, carry out as follows:

Operation 1: call the adaptive network dispatching algorithm and draw comprehensive evaluation coefficient θ;

Operation 2: call the fuzzy decision rule, return target function value f (i) and memory selection X (i);

Operation 3: the target function value that record is maximum, minimum

fmax＝max{f(Xi)，i＝1，2，...，NP}

fmin＝min{f(Xi)，i＝1，2，...，NP}

Make i ^*=arg{f (Xi)=fmin}, X (i ^*) be the chromosome of minimum target functional value.

(6)IF?f ^*≥fmin，thenf ^*＝fmin，X ^*＝X(i ^*)；

(7) calculate just when function and every gene just when, calculate it according to the adaptive value of gene and select probability, and select the big or small picked at random gene of probability to duplicate;

(8) intersect and the computing that makes a variation, produce new generation population, and substitute a new gene at random, change (4) with gene best among the previous generation;

(9) output result, record f* and X* finish as optimum solution.

The genetic algorithm that described employing embeds fuzzy rule is a kind of of genetic algorithm, its objective is in reorganization to select to embed additional local optimum module in the circulation, makes it to move to last locally optimal solution.This algorithm divides three phases to realize: the phase one is utilized the genetic algorithm optimization routing of providing and delivering; Subordinate phase utilizes the adaptive network dispatching algorithm to calculate its corresponding index; Phase III utilizes the fuzzy rule that embeds to carry out chromosomal adjustment.

It in the described emulator system model of setting up by algorithm for design.To compare by the data dispatching in Master Control Center design-calculated operation result (dynamic data of computing in the Master Control Center) and the emulator (being the data that static data is concentrated), if the re-set target of not meeting, then carry out the correction of System Discrimination type, till the output that generates data fit real data system by emulator.

Simulation result among Fig. 1 and re-set target refer to is meant that respectively the result in the controlling schemes that result that the genetic algorithm optimization by the fuzzy rule that embeds generates and practical problems require compares, usually adopt least square method that the data in the system model are revised, thereby model optimization algorithm among Fig. 1 is adjusted.Adjust back output and optimize the parameter of result's (satisfying objective function optimal value f (x), average, maximal value, superior rate), realize training and adjustment algorithm in the Master Control Center by intersecting parameters such as the factor and mutagenic factor as Master Control Center.The selection and the design itself of described embedding rules optimization of fuzzy method have embodied adaptive processing procedure.

Below in conjunction with the city logistics delivery system in detail concrete application of the present invention is described in detail, the information flow of dynamic data self-adaptive processing control mainly comprises following several core link in the path optimization's scheduling of promptly providing and delivering.The information flow of following algorithm for design is the key component of system.

To shown in Figure 4, the design core of described city logistics delivery system is the solution of the dynamically optimized scheduling problem of LRP as Fig. 3.The work that the scheduler program of its core will be finished in the dynamic data infosystem of city logistics delivery system has: the data of calling in according to control program select delivery home-delivery center, choose vehicle, prestowage vehicle, arrange the time of departure and select traffic route.If all home-delivery centers are regarded as potential facility, and actual road virtual is become traffic route, this scheduler program is exactly in fact an optimizing process that has the LRP of time restriction.The specific implementation of city logistics delivery system varies, this mainly is by professional self the location decision of dispensing, client such as service is enterprise or ordinarily resident, the goods of dispensing is the single relatively bulky goods of kind or short run goods various in style or the like, and these all will influence the specific implementation of its infosystem.But the LRP optimization method as its core is constant, and it has very big stability.Each logistics distribution system all will relate to decision processes such as addressing, arrangement route.Join on time under the prerequisite of goods satisfied by the method that the present invention proposes, not only controlled the cost of logistics distribution, also satisfied simultaneously client's demand, therefore the adaptive processing method that proposes of the present invention is effectively, and this method can promote the use of in the production reality in other field of Optimization Dispatching.

Reasonably city logistics dispensing scheduling should be satisfied following three targets: 1. transportation on time, and promptly the goods time period is on request transported in client's hand; 2. total cost (comprise the haulage track cost, set up and the fixed cost in operation warehouse, obtain means of transport cost etc.) minimum; 3. the vehicle transport total path is the shortest.

In order to solve these problems of city logistics dispensing, the present invention is on the abstract single goal LRP model basis of the LRP of forefathers' research, foundation meets the logistics distribution Model for Multi-Objective Optimization, and adopts heuritic approach and genetic algorithm, from multi-level solution city logistics dispensing Optimization Dispatching design problem.It mainly is to propose at the problem that the location of the facility in the logistics activity, traffic route are arranged.

As Fig. 2 and shown in Figure 5, this logistics distribution system input customer order information is to customer information processing module (not shown), and carry out wavelet analysis by filtrator, finish pre-service to the customer order original input data, and deposit pretreated data in dynamic data set with unified format, operation after being convenient to.This dynamic data concentrates the data that comprise to have: the client connects ETCD estimated time of commencing discharging, place, assortment of article, quantity etc., and these data deposit in along with each customer order, and all these dynamic datas also are kept at the Business Information data centralization as historical summary.

Concentrate regular good data to call in Master Control Center dynamic data then, this Master Control Center is also accepted the data from the static data collection simultaneously.This static data centralized stores be the data that seldom change, position, the number of vehicles of each home-delivery center, type, delivered payload capability as home-delivery center, information such as the numeral of transport routes, Customer Location, these information are not forever constant, and much smaller with respect to the frequency of its change of dynamic data.After this Master Control Center receives static and dynamic data, begin to start dispensing path optimization dispatching algorithm, and set up model bank by the processing rule in the controller based on the distribution network of the genetic algorithm of fuzzy rule, simultaneously the prior model in the emulator is compared, then export result of calculation if meet re-set target; If do not meet then, thereby realize scheduling to constraint conditions such as dispensing delivery period, transportation cost by the further correction model of system model correction identifier.Described result of calculation is the decision information of relevant goods delivery, this decision information mainly be included in which or which home-delivery center arrange cargo transport, which car cargo transport of each home-delivery center, each car load which client goods, dress what, when dispatch a car, the traffic route of each car etc.

This Master Control Center is issued corresponding home-delivery center to these decision informations, instructs home-delivery center to arrange the goods delivery business.Simultaneously, Master Control Center is passed to the customer information processing module to information such as partial data such as acknowledgement of orders, vehicle time of arrival, condition of loadings, and module becomes these information translation friendly form to pass to the client thus.Also to mail to the Business Information data set to these decision informations in addition, preserve as historical summary.

The present invention is on the operating mechanism basis of static scheduling system, proposition is based on the multiple-objection optimization dispatching system of the Adaptive Identification type of intelligent algorithm, both guaranteed the optimal selection of scheduling scheme, can obtain scheduling scheme again, determine the characteristic of real system, as dispensing delivery period, transportation cost according to practical problems, offer emulator and realize accurately simulation real system, by introducing identifier, as the self-correcting controlling mechanism, the slow variation of the dynamic data of adaption object characteristic; In addition, the present invention is in system's input raw data stage, through the pre-service of filtrator realization to real data, self-adaptive controlled by reference making mechanism, but the slow variation of adaption object characteristic.

Claims (3)

1. the adaptive processing method of dynamic data during an Optimization Dispatching is controlled, it is characterized in that: mainly be the raw data of input to be carried out wavelet analysis by wave filter, after involving noise after filtration and separating, with pretreated data output in the Master Control Center according to dynamically, static, services sets carries out classification and handles; After this Master Control Center receives static and dynamic data, select different adaptive scheduling algorithms and set up the genetic algorithm that embeds fuzzy rule data are optimized computing by controller wherein, and the data result that generates and the model in the emulator and data compared, if meet re-set target, then select satisfactory scheduling scheme to export as system; If the data result and the result error in the emulator that generate are bigger, do not meet re-set target, then the data result is revised by system model correction identifier, enter the controller training of Master Control Center simultaneously, and the algorithm in this controller adjusted, till the data result of Master Control Center output and the deviation between the result in the emulator meet re-set target.

2. the adaptive processing method of dynamic data in a kind of Optimization Dispatching control according to claim 1, it is characterized in that: described adaptive scheduling algorithm has selected based on the fuzzy logic algorithm that quantizes fuzzy rule, problem is divided into critical path and non-critical path, and its concrete steps are as follows:

Step 1: the membership of calculating fuzzy factors

K=1,2,3, and round values Vac, Vmc, Vmnc;

Described Vac is meant the quantity that increases critical path, and Vmc is meant diminishbb client's quantity, and Vmnc is meant the quantity of diminishbb non-critical path;

Critical path collection and non-critical path collection are defined as V respectively _cAnd V _Nc, then the adaptive network dispatching algorithm is as follows:

Step (1): from node i=1 ..., n demarcates initial time T respectively _sWith concluding time T _e

Step (2): calculate and introduce comprehensive evaluation coefficient θ,

$\mathrm{\&theta;}=\mathrm{\&gamma;}\&CenterDot;\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}(\mathrm{Te}-\mathrm{Ts})+\mathrm{\&delta;}\&CenterDot;\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}r+\mathrm{\&sigma;}\&CenterDot;\frac{Q}{\mathrm{\&Sigma;Q}}$

In the formula

R---node radius; Q---demand

Step (3): to n, determine that whether node is at critical path collection V from node i=1 _cWith non-critical path collection V _Nc

Step (4): calculating target function value f (x)

$f\left(x\right)=\mathrm{\&beta;}(T_{\mathrm{ij}}-{\mathrm{MT}}_{\mathrm{ij}})+\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{d}_{\mathrm{ij}}$

Step 2: the degree of membership of calculating the fuzzy decision rule K=1,2,3,4; Specific algorithm is as follows:

If

Be fuzzy decision D _kDegree of membership, k=1,2,3,4, consider of the influence of three fuzzy factors to this decision-making, utilize decision-making expert's knowledge and experience to obtain:

${\mathrm{\&mu;}}_{D_1}={\mathrm{\&mu;}}_{{\tilde{F}}_1}^c\&CircleTimes;{\mathrm{\&mu;}}_{{\tilde{F}}_2}$

${\mathrm{\&mu;}}_{D_2}={\mathrm{\&mu;}}_{F_1}\&CircleTimes;{\mathrm{\&mu;}}_{F_3}$

${\mathrm{\&mu;}}_{D_3}={\mathrm{\&mu;}}_{{\tilde{F}}_1}\&CircleTimes;{\mathrm{\&mu;}}_{F_2}^c\&CircleTimes;{\mathrm{\&mu;}}_{F_3}^c$

${\mathrm{\&mu;}}_{D_4}={\mathrm{\&mu;}}_{F_1}^c\&CircleTimes;{\mathrm{\&mu;}}_{F_2}^c\&CircleTimes;{\mathrm{\&mu;}}_{F_3}^c$

Wherein, μ ^c=1-μ,

Symbol is an operator

${\mathrm{\&mu;}}_1\&CircleTimes;{\mathrm{\&mu;}}_2=\frac{{\mathrm{\&mu;}}_1{\mathrm{\&mu;}}_2}{\mathrm{\&delta;}+(1-\mathrm{\&delta;})({\mathrm{\&mu;}}_1+{\mathrm{\&mu;}}_2-{\mathrm{\&mu;}}_1{\mathrm{\&mu;}}_2)}$

Step 3: calculate selected decision-making k

From obtaining

Then decision-making can be drawn by following formula,

$k=\arg \max ({\mathrm{\&mu;}}_{D_1},{\mathrm{\&mu;}}_{D_3},{\mathrm{\&mu;}}_{D_3},{\mathrm{\&mu;}}_{D_4})$

Step 4: the path to the decision-making selected and selection adds respectively, reducing, if decision-making then is add operation for critical path, non-critical path then is reducing, and the adaptive network dispatching algorithm in the invocation step 1, dispatches the calculating target function value again;

Step 5: calculating target function value: f (x) specifically is calculated as:

Step 6: return f (x) and selection result X=[X1, X2 ..., Xi ..., Xn] and to step 5;

Genetic algorithm and specific implementation step that described foundation embeds fuzzy rule comprise:

Step 1, determine the gene code mode,, genetic operator, select to calculate slightly and stopping criterion just when function

(1) gene code mode: be natural number coding at the client;

(2) just when function

f(l)＝Fmax-F(l)+a，1＝1，2…，NP (1)

Described NP is a population number, Fmax=max{F (l), and l=1,2 ..., NP}, a are little positive numbers;

(3) genetic operator: adopt 2 bracketing methods, the center section of promptly getting two chromosome correspondences carries out corresponding exchange, and mutation operation is the change that the chromosomal part of random choose is worth;

(4) select to calculate slightly: adopt gyroscope wheel method and elite's selection strategy, elite's selection strategy is meant selects a gene arbitrarily in population of new generation, replace with gene best among the previous generation, and the selection probability of the 3rd gene is in a new generation:

$P_k\left(i\right)=\frac{f_k-f_k\left(i\right)}{\underset{i=1}{\overset{\mathrm{NP}}{\mathrm{\&Sigma;}}}\{f_k-f_k\left(i\right)\}}$ f _k＝max{f _k(i)，i＝1，2，…，NP}

(2)

(5) stopping criterion: adopt the stopping criterion of greatest iteration number as algorithm;

Step 2, based on the solution procedure of the genetic algorithm that embeds fuzzy rule

(1) setup parameter: population scale NP, greatest iteration algebraically NG, crossover probability Pc, variation probability P m;

(2) call the adaptive network dispatching algorithm, calculate comprehensive evaluation coefficient θ, and definite critical path and non-critical path;

(3) it is as follows to produce NP chromosomal initial population at random:

Z=[X1, X2 ... Xi ..., X (N)], Xj (i) ≠ Xk (i), any j, k, i=1,2 ..., NP

Wherein Xj (i) is a natural number that is not more than Ni, establishes in fact that the number of iterations of population is k=0, and initial feasible solution is X*=X (l), and objective function F *=Q, Q are a big integer;

(4) satisfy commentaries on classics (9) if judge whether to satisfy end condition, otherwise change (5);

(5) for chromosome x (i), i=1,2 ..., NP, carry out as follows:

Operation 1: call the adaptive network dispatching algorithm, calculate comprehensive evaluation coefficient θ;

Operation 2: call the fuzzy decision rule, return target function value f (i) and memory selection X (i);

Operation 3: the target function value that record is maximum, minimum

fmax＝max{f(Xi)，i＝1，2，…，NP}

fmin＝min{f(Xi)，i＝1，2，…，NP}

Make i*=arg{f (Xi)=fmin}, X (i*) is the chromosome of minimum target functional value

(6)IF?f*≥fmin，then?f*＝fmin，X*＝X(i*)；

(7) calculate just when function and every gene just when, calculate it according to the adaptive value of gene and select probability, and select the big or small picked at random gene of probability to duplicate;

(8) intersect and the computing that makes a variation, produce new generation population, and substitute a new gene at random, change (4) with gene best among the previous generation;

(9) output result, record f* and X* finish as optimum solution.

3. the adaptive processing method of dynamic data in a kind of Optimization Dispatching control according to claim 1, it is characterized in that: the result in the controlling schemes that described result who generates by the genetic algorithm optimization of the fuzzy rule that embeds and practical problems require compares, usually adopt least square method that the data in the system model are revised, thereby the algorithm in the Master Control Center is adjusted, adjust back output and optimize the parameter of result, by realize training and adjustment as the intersection factor of parameter and mutagenic factor algorithm in the Master Control Center as Master Control Center.

Images