CN104318008A - Loose constraint genetic simplex algorithm based optimal design method for condenser - Google Patents

Abstract

The invention belongs to the field of nuclear power unit design, in particular to a loose constraint genetic simplex algorithm based optimal design method for a condenser. The method comprises the following steps: establishing a parameter evaluation model of the condenser; determining optimal variables, an objective function and a constraint function, and establishing an optimization model of the condenser; generating an initial population according to the optimization model; creating a loose restraint processing method for evaluating the individual quality; carrying out the genetic algorithm optimization of improved optimization strategy on the population, and generating a new population obtained after genetic evolution; generating the next generation of new population; utilizing the loose constraint processing method to evaluate the next generation of new population, and updating the restraint looseness; if meeting the termination condition, terminating the population evolution, and outputting the global optimal solution. The invention provides the loose restraint processing method for the complex nonlinear constraint problem of condenser optimization, and the loose restraint processing method is integrated into a hybrid optimization algorithm, so that the ability of optimization of the algorithm is greatly enhanced, and the effectiveness of condenser optimization design is ensured.

Description

A kind of condenser Optimization Design based on loose constraint heredity simplex algorithm

Technical field

The invention belongs to nuclear power unit design field, be specifically related to a kind of condenser Optimization Design based on loose constraint heredity simplex algorithm.

Background technology

Along with the development of new forms of energy, nuclear energy more and more comes into one's own as the new forms of energy that a kind of power density is high, clean, economic, applies also more and more extensive.Now, nuclear energy technology is generally crossed over to the third generation by the second generation, and security is more and more higher, and single heap power is increasing.Condenser is as the crucial heat transmission equipment in nuclear power unit, its bulking value is maximum in all devices, its performance also directly has influence on the thermal efficiency of nuclear power unit, the increase of single heap power causes the size of condenser and weight also increasing, and this brings difficulty to the layout of equipment and transport.In addition, atomic marine plant because space constraint also claimed structure is compact, volume is little, lightweight.But the selected main of current design of condenser scheme relies on the accumulation of previous experiences and the judgement of expert, also just chooses best design proposal from multiple candidate scheme to the optimization of design proposal.Consider that the design of condenser is very complicated engineering, and above-mentioned design optimization is a kind of semiempirical method for designing.This method cannot provide the optimal case in design space in theory, the preferred embodiments in several limited candidate scheme can only be obtained, what the quality of design result depended on the design experiences of deviser and candidate scheme to a great extent chooses level, and the efficiency of this design is simultaneously often not high.Therefore, Optimum Theory is applied to the design process of condenser, to the optimal design of the volume of condenser, weight and other performance index, uses intellectual technology to obtain best equipment de-sign scheme and have important practical significance.

When being optimized design to condenser, the parameter evaluation model of foundation combines the factor in all directions such as physical dimension, the strength of materials, thermal-hydraulic, belongs to complicated multivariate, the nonlinear optimal problem of multiple constraint.For this kind of problem, the calculating effect of traditional intelligent optimization algorithm is often undesirable.The genetic algorithm (GA) created by Univ Michigan-Ann Arbor USA John professor Holland is a kind of heuristic Swarm Evolution algorithm based on gene principle, is one of intelligent computation gordian technique, is widely used in engineering optimization field.Stronger ability of searching optimum is its internal characteristics attribute, but when processing complicated constrained optimization problem, genetic algorithm and innovatory algorithm thereof still exist the shortcoming that local search ability is weak, Searching efficiency is low, often can not meet the needs of Solve problems.Simplex algorithm (SM) is a kind of local search optimization algorithm of classics, has very strong local search ability, realize simple, fast convergence rate, but ability of searching optimum is poor.With strong complementarity due to genetic algorithm and simplex algorithm, and there is good bonding properties, simplex algorithm embeds in genetic algorithm by scholar Du Xiuli, Jiang Liping etc. (2006), form hybrid optimization algorithm, obtain performance boost to a certain extent, but the simple hybrid algorithm of this traditional constraints disposal route of only arranging in pairs or groups, when solving Complex Constraints optimization problem, effect is often undesirable, to complexity trial function (as scholar Runarsson propose g02 trial function) carry out optimizing time, be also difficult to search globally optimal solution.

Summary of the invention

The object of the present invention is to provide a kind of condenser Optimization Design based on loose constraint heredity simplex algorithm that fast and effeciently can search globally optimal solution.

The object of the present invention is achieved like this:

(1) condenser parameter evaluation model is set up;

(2) determine optimized variable, objective function and constraint function, set up the Optimized model of condenser;

(3) according to Optimized model, the population size NP of algorithm, constraint function number GN, simplex number of vertex K, maximum evolution iterations G is determined _max, make initial evolution iterations G=1, and produce initial population;

(4) create the loose constraint facture of the individual good and bad degree of evaluation, evaluate initial population, and calculate the initial constraint slackness ε of colony ₁;

(5) genetic algorithm optimization improving optimizing strategy is carried out to colony, produce the new colony after genetic evolution;

(6) continue to carry out simplex algorithm optimization to the new colony after genetic evolution, produce new colony of future generation;

(7) utilize loose constraint facture to evaluate the new colony of the next generation, evolution iterations G increases by 1, upgrades constraint slackness ε _g;

(8) judge whether to meet stopping criterion for iteration, i.e. G=G _maxif do not meet end condition and return step (4), enter the optimizing of next round Swarm Evolution, if meet end condition, then stop Swarm Evolution, export globally optimal solution.

Described step (4) comprises:

(4.1) target function value and the normalization constraint violation value of individual in population is solved;

(4.2) the constraint slackness of colony is solved;

(4.3) the loose constraint facture of the individual good and bad degree of evaluation is created;

(4.4) loose constraint facture is used to evaluate initial population.

Described step (5) comprises following steps:

(5.1) from colony, repeatedly select a pair individuality at random and not to match, parallel cross and variation operation is carried out to pairing individuality;

(5.2) two first wifves are merged into one group to individual, two intersection individualities, two variation individualities, utilize loose constraint facture to evaluate individuality in group;

(5.3) according to Swarm Evolution different times, elitist selection or part elitist selection are carried out to individuality in group of individuals, entered the new colony after genetic algorithm optimizing by the individuality selected;

(5.4) repeat above step, until after all individualities complete the operation of parallel cross and variation and elitist selection in colony, complete a genetic algorithm optimizing, produce the new colony after genetic algorithm optimizing.

Described step (6) comprising:

(6.1) in order the new colony after genetic algorithm optimizing is divided into NP/K group, forms NP/K simplex;

(6.2) at the different times of Swarm Evolution, simplex is carried out to the simplex algorithm optimizing of different strategies: if be in the prometaphase of Swarm Evolution, setting is as G≤0.8G _maxtime, each simplex carries out [K/4] secondary simplex algorithm local optimal searching all continuously; If be in the later stage of Swarm Evolution, setting is as G > 0.8G _maxtime, each simplex carries out K simplex algorithm local optimal searching all continuously;

(6.3) will the NP/K group individuality synthesis one large group of simplex algorithm optimizing be completed, form colony of new generation: if be in the later stage of Swarm Evolution, namely as G > 0.8G _maxtime, upset individual in population sequence.

In described step (4.3), loose constraint facture embodiment is:

For two individual a, b in certain generation colony, target function value is respectively f (x _a), f (x _b), normalization constraint violation value is respectively φ (x _a), φ (x _b), the constraint slackness of colony is ε, and its concrete evaluation method is as follows:

If a. individual a, b all meet loose constraint condition, i.e. φ (x _a)≤ε, φ (x _b)≤ε, then the individuality that target function value is less is dominant;

If b. body meets loose constraint condition one by one, another individuality does not meet loose constraint condition, then the individuality satisfied condition is dominant;

If c. two individualities all do not meet loose constraint condition, then the individuality that constraint violation value is less is dominant;

If d. φ (x _a)=φ (x _b), then the individuality that target function value is less is dominant.

The parallel cross and variation mode of operation of described step (5.1) is, under the prerequisite not upsetting individual sequence, is sequentially matched between two by the individuality in colony, selects a pair pairing individual, carries out uniform crossover operator with total probability, produce pair of cross individual; Continue to carry out mutation operation to first wife to individuality with total probability, the method for mutation operation is, if be in Swarm Evolution in earlier stage, and setting G≤0.5G _maxtime, carry out even mutation operation, if be in the later stage of Swarm Evolution, setting G > 0.5G _maxtime, carry out Gaussian mutation operation, produce a pair variation individuality.

Improvement selection strategy in described step (5.3) is, if be in the prometaphase of Swarm Evolution, namely as G≤0.8G _maxtime, from group of individuals, Stochastic choice one non-optimal is individual, enters the new colony after genetic algorithm optimizing together with optimum individual; If be in the later stage of Swarm Evolution, namely as G > 0.8G _maxtime, the new colony after selecting optimum individual to enter genetic algorithm optimizing from group of individuals together with suboptimum individuality.

Beneficial effect of the present invention is:

1. abandoned the semiempirical method for designing of traditional condenser design, the Optimum Theory in artificial intelligence has been introduced in the design of condenser, the best design of condenser can be obtained;

2. for the complex nonlinear restricted problem that condenser is optimized, propose loose constraint facture, and be dissolved in hybrid optimization algorithm, greatly strengthen the optimizing ability of algorithm, ensure the validity of condenser optimal design.

Accompanying drawing explanation

Fig. 1 is the structural drawing of the condenser evaluation model that the present invention proposes;

Fig. 2 is the process flow diagram of the hybrid optimization algorithm that the present invention proposes.

Embodiment

Below in conjunction with accompanying drawing, the present invention is described further.

The present invention is a kind of condenser Optimization Design based on loose constraint heredity pure algorithm.Comprise the following steps: the parameter evaluation model setting up condenser; According to condenser problem to be optimized, determine optimized variable, objective function and constraint function, set up the Optimized model of condenser; According to condenser Optimized model, determine population parameter, construct random initial population; Set up the loose constraint facture of the individual good and bad degree of evaluation, in conjunction with condenser parameter evaluation model evaluation initial population, and calculating initially retrains slackness; First colony is carried out to the genetic algorithm global optimizing of parallel cross and variation operation, then carry out simplex local optimal searching; Judge whether the hybrid algorithm condition of convergence meets, if do not meet, continue iteration optimizing, if meet, optimize and terminate and export optimizing design scheme.The present invention establishes a whole set of technical scheme of the condenser optimal design such as condenser parameter evaluation model, condenser Optimized model and optimized algorithm, genetic algorithm and simplex algorithm not only organically combine by the hybrid optimization algorithm that wherein the present invention proposes, and for the complex nonlinear confinement features of condenser optimal design, propose loose constraint facture, substantially increase ability of searching optimum and the local search ability of hybrid optimization algorithm, for effective enforcement of condenser optimizing design scheme provides strong technical support.

The concrete implementation step of technical scheme of the present invention is as follows:

Step 1: according to designing requirement, sets up condenser parameter evaluation model;

Step 2: according to problem to be optimized, determines optimized variable, objective function and constraint function, sets up the Optimized model of condenser;

Step 3: according to Optimized model, determines the population size NP of algorithm, constraint function number GN, simplex number of vertex K, maximum evolution iterations G _maxetc. parameter, make initial evolution iterations G=1, and produce initial population at random;

Step 4: the loose constraint facture creating the individual good and bad degree of evaluation, evaluates initial population, and calculates the initial constraint slackness ε of colony ₁;

Step 5: carry out the genetic algorithm optimization improving optimizing strategy to colony, produces the new colony after genetic evolution;

Step 6: continue the new colony after to genetic evolution and carry out simplex algorithm optimization, produces new colony of future generation;

Step 7: utilize loose constraint facture to evaluate the new colony of the next generation, evolution iterations G increases by 1, upgrades constraint slackness ε _g;

Step 8: judge whether to meet stopping criterion for iteration, i.e. G=G _maxif do not meet end condition and return step 4, enter the optimizing of next round Swarm Evolution, if meet end condition, then stop Swarm Evolution, export globally optimal solution.

Condenser parameter evaluation model in described step 1 will consider the performance requirement of condenser, design criteria, safety standard, thermal-hydraulic and build process etc. and require to set up mathematics physics model;

Condenser Optimized model in described step 2 will consider thermal technology, physical dimension, strength of materials restriction and safety criterion to set up constraint function; Optimized variable to be established by the sensitivity level of thermal technology and structure parameters influence according to operation characteristic and the performance index such as weight, volume; According to designing requirement, the optimal design of condenser to be converted into single goal or multi-objective optimization question;

Described step 4 comprises following steps:

Step 4.1: the target function value and the normalization constraint violation value that solve individual in population;

Step 4.2: the constraint slackness solving colony;

Step 4.3: the loose constraint facture creating the individual good and bad degree of evaluation;

Step 4.4: use loose constraint facture to evaluate initial population.

Described step 5 comprises following steps:

Step 5.1: repeatedly select a pair individuality at random and not and match from colony, carries out parallel cross and variation operation to pairing individuality;

Step 5.2: two first wifves are merged into one group to individual, two intersection individualities, two variation individualities, utilizes loose constraint facture to evaluate individuality in group;

Step 5.3: according to Swarm Evolution different times, carries out elitist selection or part elitist selection to individuality in group of individuals, is entered the new colony after genetic algorithm optimizing by the individuality selected;

Step 5.4: repeat above step, until after all individualities complete the operation of parallel cross and variation and elitist selection in colony, complete a genetic algorithm optimizing, produces the new colony after genetic algorithm optimizing.

Described step 6 comprises following steps:

Step 6.1: under the prerequisite not upsetting individual sequence, in order the new colony after genetic algorithm optimizing is divided into NP/K group, forms NP/K simplex;

Step 6.2: at the different times of Swarm Evolution, carries out the simplex algorithm optimizing of different strategies to simplex.If be in the prometaphase of Swarm Evolution, setting is as G≤0.8G _maxtime, each simplex carries out [K/4] secondary simplex algorithm local optimal searching all continuously, belongs to the Local Search of the more shallow degree of depth; If be in the later stage of Swarm Evolution, setting is as G > 0.8G _maxtime, each simplex carries out K simplex algorithm local optimal searching all continuously, belongs to degree of depth Local Search.

Step 6.3: will the NP/K group individuality synthesis one large group of simplex algorithm optimizing be completed, and form colony of new generation.If be in the later stage of Swarm Evolution, namely as G > 0.8G _maxtime, upset individual in population sequence.

In described step 4.3, loose constraint facture embodiment is, for two individual a, b in certain generation colony, target function value is respectively f (x _a), f (x _b), normalization constraint violation value is respectively φ (x _a), φ (x _b), the constraint slackness of colony is ε, and its concrete evaluation method is as follows:

If individual a, b all meet loose constraint condition, i.e. φ (x _a)≤ε, φ (x _b)≤ε, then the individuality that target function value is less is dominant;

If body meets loose constraint condition one by one, another individuality does not meet loose constraint condition, then the individuality satisfied condition is dominant;

If two individualities all do not meet loose constraint condition, then the individuality that constraint violation value is less is dominant;

If φ is (x _a)=φ (x _b), then the individuality that target function value is less is dominant;

The parallel cross and variation mode of operation of described step 5.1 is, under the prerequisite not upsetting individual sequence, is sequentially matched between two by the individuality in colony, selects a pair pairing individual, carries out uniform crossover operator with total probability, produce pair of cross individual; Continue to carry out mutation operation to first wife to individuality with total probability, the method for mutation operation is, if be in Swarm Evolution in earlier stage, and setting G≤0.5G _maxtime, carry out even mutation operation, if be in the later stage of Swarm Evolution, setting G > 0.5G _maxtime, carry out Gaussian mutation operation, produce a pair variation individuality.

Described step 5.3 improves selection strategy, if be in the prometaphase of Swarm Evolution, namely as G≤0.8G _maxtime, from group of individuals, Stochastic choice non-optimal individuality, enters the new colony after genetic algorithm optimizing together with optimum individual; If be in the later stage of Swarm Evolution, namely as G > 0.8G _maxtime, the new colony after selecting optimum individual to enter genetic algorithm optimizing from group of individuals together with suboptimum individuality.

Carry out Volume Optimal Design for the condenser of Qinshan first phase nuclear power station as parent form below, the enforcement of piling the condenser Optimization Design based on loose constraint heredity simplex algorithm of the present invention is described in detail.

Step 1: according to designing requirement, sets up condenser parameter evaluation model;

Condenser parameter evaluation model will consider the performance requirement of condenser, design criteria, safety standard, the requirement such as thermal-hydraulic and build process, thermal conduction study rule-of-thumb relation and condenser result of industrial test is utilized to set up thermal-hydraulic physical arrangement model, this model comprises heat Balance Calculation module, and drag evaluation module, vibration check module and condensation vessel volume computing module.Detailed process is see Fig. 2.

Step 2: determine optimized variable, objective function and constraint function, sets up the Optimized model of condenser;

Establishing optimized variable according to operation characteristic and condensation vessel volume by the sensitivity level of thermal technology and structure parameters influence is:

(a) cooling tube outside diameter d _o(mm), variation range [20,30];

(b) cooling tube pitch s, variation range [24,38];

(c) chilled water flow velocity u (m/s), variation range [2,4];

(d) condenser pressure P (KPa), variation range [4,6];

Can be write as following form:

x＝(d _o,s,u,P)＝(x ¹,x ²,...,x ^D) (1)

Require that establishing constraint condition is according to equipment performance, the strength of materials, safety criterion etc.:

(a) Water in Condenser resistance Δ P _f(KPa) restriction range is 60≤Δ P _f≤ 80, can be write as following form

g ₁(x)-60≥0 (2)

g ₂(x)-80≤0 (3)

(b) condensed fluid degree of supercooling Δ T _sub(DEG C) restriction range 0≤Δ T _sub≤ 1.5, can be write as following form

g ₃(x)≥0 (4)

g ₄(x)-1.5≤0 (5)

C () cooling tube wall thickness δ (mm), restriction range 0.5≤δ≤1.3, can be write as following form

g ₅(x)-0.5≥0 (6)

g ₆(x)-1.3≤0 (7)

D () condenser shell length breadth ratio L/W, restriction range 1.0≤L/W≤2.5, can be write as following form

g ₇(x)-1.0≥0 (8)

g ₈(x)-2.5≤0 (9)

(e) cooling tube lay ratio s/d _o, restriction range 1.4≤s/d _o≤ 1.6, can be write as following form

g ₉(x)-1.4≥0 (10)

g ₁₀(x)-1.6≤0 (11)

Objective function can be expressed as:

$\min V\left(x\right)=\underset{g_i\left(x\right)\≥0}{\min }V(d_o,s,u,P)---\left(12\right)$

Namely be variable with x, solve under the restriction of constraint condition (2) ~ (11), the minimum value of objective function V (x).

Step 3: according to mathematical model, setting population size NP=30, constraint function number GN=10, simplex number of vertex K=D+1=5, maximum iteration time G _max=1000, the value of initial evolution iterations G is 1, and produces initial population at random;

Each variable to be optimized is D dimension space vector, is called the individuality in colony, explicit at variable to be optimized

Restriction range interior random value carrys out initialization population at individual x _{i, 1}:

$x_{i,1}=(x_{i,1}^1,x_{i,1}^2,...,x_{i,1}^D)---\left(13\right)$

$x_{i,1}^j=x_{\min }^j+\mathrm{rand}\left(\mathrm{0,1}\right)\·(x_{\max }^j-x_{\min }^j)---\left(14\right)$

In formula, represent higher limit and the lower limit of the jth dimension component of individual vector respectively,

i＝1,2,...,NP，j＝1,2,...,D。

Then initial population is:

X _G＝1＝{x _1,1,x _2,1,...,x _NP,1} (15)

Step 4: use loose constraint facture to evaluate initial population, and solve the lax ε of initial constraint ₁;

Step 4.1: solve target function value individual in initial population and normalization constraint violation value;

Concrete grammar is as follows:

For individual x _{i, 1}, using formula (12) calculates its target function value f (x _{i, 1}).Constraint function formula (2) ~ (11) are converted to g _kthe form of (x)≤0, then calculate individual x _{i, 1}constraint function value g _k(x _{i, 1}).Wherein, i=1,2 ..., NP, k=1,2 ..., GN.

Defining i-th individual normalization constraint violation value function is:

${\tilde{v}}_k\left(x_i\right)=\left\{\begin{array}{cc}0,& g_k\left(x_{i,1}\right)\≤0\\ \frac{g_k\left(x_{i,1}\right)}{g_k^{\max }\left(x\right)},& g_k\left(x_{i,1}\right)>0\end{array}\right.---\left(16\right)$

In formula, for individualities all in contemporary community are about the maximal value of the constraint function value of a kth constraint condition.

The normalization constraint violation value defining i-th individuality total is:

${\mathrm{\φ}}_1\left(x_{i,1}\right)=\underset{k=1}{\overset{\mathrm{GN}}{\mathrm{\Σ}}}{\tilde{v}}_k\left(x_{i,1}\right)---\left(17\right)$

Step 4.2: the constraint slackness solving initial population;

Definition initial constraint slackness is:

${\mathrm{\ϵ}}_1=\frac{\underset{i=1}{\overset{\mathrm{NP}}{\mathrm{\Σ}}}{\mathrm{\φ}}_1\left(x_{i,1}\right)}{\mathrm{NP}}---\left(18\right)$

Step 4.3: use loose constraint facture to evaluate initial population.

Step 5: carry out the genetic algorithm optimization improving optimizing strategy to colony, produces the new colony after genetic evolution;

Step 5.1: parallel cross and variation operation is carried out to individual in population;

Under the prerequisite not upsetting individual sequence, G is sequentially matched between two for the individuality in colony, select the individual x of a pair pairing _i,G, x _{i+1, G}carry out parallel cross and variation operation, produce the individual x of two intersections _i,G', x _{i+1, G}' and the individual x of two variations _i,G", x _{i+1, G}", method is as follows:

Intersection is produced individual by uniform crossover operator:

${x_{i,G}^j}^{\′}={\mathrm{\α}}_1\·x_{i+1,G}^j+(1-{\mathrm{\α}}_1)\·x_{i,G}^j---\left(19\right)$

${x_{i+1,G}^j}^{\′}={\mathrm{\α}}_2\·x_{i,G}^j+(1-{\mathrm{\α}}_2)\·x_{i+1,G}^j---\left(20\right)$

Variation individuality is produced by even variation or Gaussian mutation operation:

${x_{i,G}^p}^{\′\′}=\left\{\begin{array}{cc}{x}_{\min }^p+{\mathrm{\β}}_1\·(x_{\max }^p-x_{\min }^p),& G\≤500\\ x_{i,G}^p+{\mathrm{\χ}}_1\·\mathrm{Min}(x_{\max }^p-x_{i,G}^p,x_{i,G}^p-x_{\min }^p),& G>500\end{array}\right.---\left(21\right)$

${x_{i+1,G}^p}^{\′\′}=\left\{\begin{array}{cc}{x}_{\min }^p+{\mathrm{\β}}_2\·(x_{\max }^p-x_{\min }^p),& G\≤500\\ x_{i+1,G}^p+{\mathrm{\χ}}_2\·\mathrm{Min}(x_{\max }^p-x_{i+1,G}^p,x_{i+1,G}^p-x_{\min }^p),& G>500\end{array}\right.---\left(22\right)$

In formula, be G for i-th individual jth dimension component in colony; α ₁, α ₂, β ₁, β ₂it is the stochastic distribution real number between 0 to 1; χ ₁=N (0,0.25), χ ₂=N (0,0.25), is the Gaussian distribution of parameter μ=1, δ=0.5; P is variant sites, and span is random integers between 1 to D (comprise 1 and D); it is the function getting smaller value in a, b.

Step 5.2: by individual x _i,G, x _{i+1, G}, x _i,G', x _{i+1, G}', x _i,G", x _{i+1, G}" merge into one group, utilize loose constraint facture to evaluate individuality in group;

Step 5.3: when evolutionary generation G≤800, carries out part elitist selection, and from group of individuals, Stochastic choice one non-optimal is individual, enters the new colony after genetic algorithm optimizing together with optimum individual; As G > 800, carry out elitist selection, the new colony after selecting optimum individual to enter genetic algorithm optimizing from group of individuals together with suboptimum individuality.

Step 5.4: after carrying out parallel cross and variation operation and elitist selection to all pairing individualities, complete a genetic algorithm optimizing, produces the new colony after genetic algorithm optimizing.

Step 6: continue the new colony after to genetic evolution and carry out simplex algorithm optimization, produces new colony of future generation;

Step 6.1: under the prerequisite not upsetting individual sequence, in order the new colony after genetic algorithm optimizing is divided into 6 groups, forms 6 simplexs;

Step 6.2: when evolutionary generation G≤800, each simplex carries out 3 simplex algorithm local optimal searching all continuously, and as G > 800, each simplex carries out 6 simplex algorithm local optimal searching all continuously.

Step 6.3: will 6 groups of individual synthesis one large groups of simplex algorithm optimizing be completed, and form colony X of new generation _g+1={ x _{1, G+1}, x _{2, G+1}..., x _{nP, G+1}.As G > 800, upset individual in population sequence.

Step 7: utilize loose constraint facture to evaluate the new colony of the next generation, the value of evolution iterations G increases by 1, upgrades constraint slackness ε _g;

Wherein ε _gmore new formula as follows:

${\mathrm{\&epsiv;}}_G=\left\{\begin{array}{cc}{\mathrm{\&epsiv;}}_0*{(1-\frac{G}{G_0})}^3,& G<G_0\\ 0& G\&GreaterEqual;G_0\end{array}\right.---\left(23\right)$

In formula, G ₀=400, for retraining lax algebraically.

Step 8: judge whether to meet stopping criterion for iteration, i.e. G=1000, if do not meet end condition to return step 4, enters the optimizing of next round Swarm Evolution, if meet end condition, then stops Swarm Evolution, exports globally optimal solution.

Table one lists and is optimized the result after design for Qinshan first phase condenser, result shows, heat-transfer pipe external diameter and pitch is increased in appropriateness, after appropriateness increases chilled water flow velocity and design of condenser pressure, under the condition keeping former condenser exchange capability of heat and other performance index constant, its volume can be reduced 19.2%.This is not changing material composition, only when change structure parameter and operational factor, can meet the design proposal of the minimum condensation vessel volume of performance requirement.

Table 1: Qinshan first phase condensation vessel volume optimizes result of calculation

Parameter	Unit	Qinshan first phase parent form value	Value after optimizing	Relative change rate
					d0	mm	25	26.71	6.91％
s	─	35	37.79	7.92％
					u	m/s	2.3	2.786	21.1％
P	kPa	4.9	5.62	15.0％
					V	m 3	4733.89	3824.98	－19.2％

Claims (7)

1., based on a condenser Optimization Design for loose constraint heredity simplex algorithm, it is characterized in that, comprise the following steps:

(1) condenser parameter evaluation model is set up;

(2) determine optimized variable, objective function and constraint function, set up the Optimized model of condenser;

(3) according to Optimized model, the population size NP of algorithm, constraint function number GN, simplex number of vertex K, maximum evolution iterations G is determined _max, make initial evolution iterations G=1, and produce initial population;

(4) create the loose constraint facture of the individual good and bad degree of evaluation, evaluate initial population, and calculate the initial constraint slackness ε of colony ₁;

(5) genetic algorithm optimization improving optimizing strategy is carried out to colony, produce the new colony after genetic evolution;

(6) continue to carry out simplex algorithm optimization to the new colony after genetic evolution, produce new colony of future generation;

(7) utilize loose constraint facture to evaluate the new colony of the next generation, evolution iterations G increases by 1, upgrades constraint slackness ε _g;

(8) judge whether to meet stopping criterion for iteration, i.e. G=G _maxif do not meet end condition and return step (4), enter the optimizing of next round Swarm Evolution, if meet end condition, then stop Swarm Evolution, export globally optimal solution.

2. a kind of condenser Optimization Design based on loose constraint heredity simplex algorithm according to claim 1, it is characterized in that, described step (4) comprises:

(4.1) target function value and the normalization constraint violation value of individual in population is solved;

(4.2) the constraint slackness of colony is solved;

(4.3) the loose constraint facture of the individual good and bad degree of evaluation is created;

(4.4) loose constraint facture is used to evaluate initial population.

3. a kind of condenser Optimization Design based on loose constraint heredity simplex algorithm according to claim 1, it is characterized in that, described step (5) comprises following steps:

(5.1) from colony, repeatedly select a pair individuality at random and not to match, parallel cross and variation operation is carried out to pairing individuality;

(5.2) two first wifves are merged into one group to individual, two intersection individualities, two variation individualities, utilize loose constraint facture to evaluate individuality in group;

(5.3) according to Swarm Evolution different times, elitist selection or part elitist selection are carried out to individuality in group of individuals, entered the new colony after genetic algorithm optimizing by the individuality selected;

(5.4) repeat above step, until after all individualities complete the operation of parallel cross and variation and elitist selection in colony, complete a genetic algorithm optimizing, produce the new colony after genetic algorithm optimizing.

4. a kind of condenser Optimization Design based on loose constraint heredity simplex algorithm according to claim 1, it is characterized in that, described step (6) comprising:

(6.1) in order the new colony after genetic algorithm optimizing is divided into NP/K group, forms NP/K simplex;

(6.2) at the different times of Swarm Evolution, simplex is carried out to the simplex algorithm optimizing of different strategies: if be in the prometaphase of Swarm Evolution, setting is as G≤0.8G _maxtime, each simplex carries out [K/4] secondary simplex algorithm local optimal searching all continuously; If be in the later stage of Swarm Evolution, setting is as G > 0.8G _maxtime, each simplex carries out K simplex algorithm local optimal searching all continuously;

(6.3) will the NP/K group individuality synthesis one large group of simplex algorithm optimizing be completed, form colony of new generation: if be in the later stage of Swarm Evolution, namely as G > 0.8G _maxtime, upset individual in population sequence.

5. a kind of condenser Optimization Design based on loose constraint heredity simplex algorithm according to claim 2, it is characterized in that, in described step (4.3), loose constraint facture embodiment is:

For two individual a, b in certain generation colony, target function value is respectively f (x _a), f (x _b), normalization constraint violation value is respectively φ (x _a), φ (x _b), the constraint slackness of colony is ε, and its concrete evaluation method is as follows:

If a. individual a, b all meet loose constraint condition, i.e. φ (x _a)≤ε, φ (x _b)≤ε, then the individuality that target function value is less is dominant;

If b. body meets loose constraint condition one by one, another individuality does not meet loose constraint condition, then the individuality satisfied condition is dominant;

If c. two individualities all do not meet loose constraint condition, then the individuality that constraint violation value is less is dominant;

If d. φ (x _a)=φ (x _b), then the individuality that target function value is less is dominant.

6. a kind of hereditary simplex optimization algorithm based on improving optimizing strategy according to claim 3, it is characterized in that the parallel cross and variation mode of operation of described step (5.1) is, under the prerequisite not upsetting individual sequence, individuality in colony is sequentially matched between two, select a pair pairing individual, carry out uniform crossover operator with total probability, produce pair of cross individual; Continue to carry out mutation operation to first wife to individuality with total probability, the method for mutation operation is, if be in Swarm Evolution in earlier stage, and setting G≤0.5G _maxtime, carry out even mutation operation, if be in the later stage of Swarm Evolution, setting G > 0.5G _maxtime, carry out Gaussian mutation operation, produce a pair variation individuality.

7. a kind of condenser Optimization Design based on loose constraint heredity simplex algorithm according to claim 5, is characterized in that the improvement selection strategy in described step (5.3) is, if be in the prometaphase of Swarm Evolution, namely as G≤0.8G _maxtime, from group of individuals, Stochastic choice one non-optimal is individual, enters the new colony after genetic algorithm optimizing together with optimum individual; If be in the later stage of Swarm Evolution, namely as G > 0.8G _maxtime, the new colony after selecting optimum individual to enter genetic algorithm optimizing from group of individuals together with suboptimum individuality.