CN106874542A - A kind of hydraulic turbine impeller multi-state multi-objective optimization design of power method - Google Patents

Abstract

The present invention relates to the optimization design field of hydraulic, particularly a kind of hydraulic turbine impeller multi-state multi-objective optimization design of power method.The beneficial effects of the present invention are：Approximate prediction model is set up and updated to training BP neural network optimized algorithm, and the CFD for reducing big load is calculated, and is calculated with fewer number, obtains enough precision of predictions.The combination of BP neural network optimized algorithm and the multi-objective genetic algorithms of NSGA II, it is ensured that whole algorithm has taken into account the diversity of population while constantly search optimal solution set, and precision is also lifted.Solution is optimized to turbo wheel using the multi-objective genetic algorithms of NSGA II, the vibration that easily occurs when deviateing optimum operating condition in the difficulty and hydraulic turbine fortune that can preferably solve hydraulic turbine impeller traditional design and type selecting, the problems such as power output is unstable, efficiency is significantly relatively low, efficiency of turbine, axial force and radial load are taken into account, operating turbine ability can be effectively improved.

Description

A kind of hydraulic turbine impeller multi-state multi-objective optimization design of power method

Technical field

Patent of the present invention is related to the optimization design field of hydraulic, particularly a kind of many mesh of hydraulic turbine impeller multi-state Mark Optimization Design.

Background technology

Hydraulic turbine is the plant equipment that the pressure energy in liquid fluid working medium is converted to mechanical energy, using hydraulic turbine Liquid excess pressure recycling in technological process can be converted to mechanical energy and drive plant equipment, be a kind of energy regenerating dress Put, be now widely used for petrochemical industry be hydrocracked, the field such as Large-scale Ammonia Plant and desalinization.Technically, there is 20KW Recover energy, so that it may recycled with hydraulic turbine.Energy regenerating hydraulic turbine technology and application have important meaning to energy-saving and emission-reduction Justice.The special energy that hydraulic turbine mainly has reversion pump form, impact type, guide-vane and some advanced countries research and development at present is returned Receive hydraulic turbine.The basic arrangement of turbine retracting device has direct-drive type and assist type to arrange.Hydraulic turbine energy recycle device It is widely used, its research develops towards specialization, particularization, diversified direction.

The hydraulic turbine that the current country commonly uses with reversion pump based on, not only operational efficiency is relatively low, and high efficient district Relative narrower, the start-up course consuming time is long, and operating condition is unstable, constrains the development of energy regenerating engineering field.Pump is anti- Transfer to hydraulic turbine operation very sensitive to changes in flow rate, energy recovery efficiency decline when flow is higher than the 10% of optimum operating condition 50%, flow less than optimum operating condition 40% when, hydraulic turbine non-recovery power and deviate optimal operating condition point run when meeting There are the bad phenomenons such as vibrate and rotating speed, power output are unstable, it can be seen that hydraulic turbine has the change to operating condition More sensitive a series of problems, such as.Domestic and international some scholar's research find that hydraulic loss of the hydraulic turbine operationally in impeller is accounted for More than the 50% of total hydraulic loss, this just illustrate hydraulic turbine hydraulic performance it is not good enough main reason is that its impeller performance compared with Difference.And invert pump and make hydraulic turbine, standard is not usually reached when it operates in turbine operating mode, occur that efficiency is low, operation stability Difference, service life is short, there is serious noise and a series of problems, such as vibration.

Based on hydraulic turbine above series of problem, the present invention starts with from the impeller geometric parameter of turbine, with reference to BP nerves Network algorithm and multi-objective genetic algorithm NSGA- II carry out multi-state multiple-objection optimization solution to turbine, take into account multiple operating modes The performance of point, makes efficiency that the operation stability of each operating mode is ensure that while being improved.It is related to patent of the present invention through retrieval Patent have：Optimization design method of radial-flow-type hydraulic turbine (publication number：CN102608914A), the invention discloses a kind of runoff The whole machine Optimization Design of formula hydraulic turbine passage component, comprising unitary heating power optimization design, and whole machine Optimization Platform, will be multiple Miscellaneous Multi-variables optimum design PROBLEM DECOMPOSITION is subproblem that is multiple relatively independent but interacting, has both remained the spy of former problem Property, effectively reducing amount of calculation again, but it is only for single spot optimization, turbine is deviateing when optimum operating condition runs easy unstable phenomenon simultaneously It is not resolved；Consider the large-scale turbo-expander impeller blade design Optimization Design (publication number of defect： CN104331553A), on original impeller stress analysis foundation, defect factors are added, using generalized regression nerve networks and base Genetic optimization operation is carried out to impeller parameters in the multi-objective optimization algorithm of genetic algorithm, the optimal solution being evenly distributed is finally given Used as impeller blade Optimal Parameters, but its optimization process is more complicated cumbersome.

The content of the invention

In order to solve to deviate optimal work in difficulty and hydraulic turbine operation of the hydraulic turbine impeller traditional design with type selecting The vibration that easily occurs during condition, the problems such as power output is unstable, efficiency is significantly relatively low, it is saturating that patent of the present invention provides a kind of fluid power Flat impeller multi-state multi-objective optimization design of power method, it is steady with the operation of inclined operating point the purpose is to improve hydraulic turbine operating efficiency It is qualitative.

Technical solution of the present invention is as follows：

Hydraulic turbine impeller of pump key geometric parameter is determined first as optimization design variable and sample number, using experiment side Method generates sample and it screen and obtains satisfactory sample point, and hydraulic turbine integrated automation is carried out to initial sample Moulding, mesh generation and CFD are calculated and are obtained respective performances parameter, set up optimization sample database, import BP neural network module The approximate agent model of optimized algorithm is set up after learning training, is finally embedded in the multi-objective genetic algorithms of NSGA- II, It is optimal as target with the efficiency under tri- flow rate working conditions points of 0.7Q, 1.0Q, 1.2Q, radial load and axial force, carry out genetic algorithm Optimizing solve, solve the overall optimal solution set of impeller.

A kind of hydraulic turbine impeller multi-state multi-objective optimization design of power method, implements step as follows：

Step1:Hydraulic turbine Impeller Design variable, object function and constraint are determined, with the outer diameter D of impeller₂, outlet is directly Footpath D₁, two inclination alphas of front and rear cover plate₁、α₂, two arc radius R on front and rear cover plate streamline₁、R₂, entrance width b₂, blade Outlet laying angle β₁, subtended angle of blade Φ and number of blade Z is design variable；Secondly, using test design method in design variable Space in generate design variable test sample；Finally, the variable in hydraulic turbine initial model is entered using Pro/E softwares Row batch Parametric designing；

Step2:The autoexec of grid and CFD software is made, realizes that all of test sample of hydraulic turbine is carried out certainly Dynamic mesh is divided and calculated with performance CFD, finally obtains the performance number of all models；

Step3:After obtaining the corresponding performance parameter of all sample patterns, geometric parameter and correspondence liquid according to test sample The performance parameter training BP neural network of power turbine, the |input paramete with design variable as approximate model, with corresponding property Energy parameter is output parameter, sets up approximate agent model；

Step4:Performance Evaluation and genetic evolutionary operations are carried out to each individuality in population, wherein individual Performance Evaluation Function is by training and be embedded in agent model in genetic algorithm and realize.Calculated using the multi-objective Genetics of NSGA- II When method is optimized to impeller geometric parameter, initial population is first randomly generated, the individual UVR exposure in initial population；Secondly, it is right Each individuality carries out Performance Evaluation in population, and performance parameter is related to object function；Finally, there are the gene of each individuality And genetic manipulation is carried out to population until meeting algorithm stop criterion according to its principle after object function, obtain multiple-objection optimization Pareto optimization disaggregation；

Step5:Pareto for obtaining optimizes disaggregation, using carrying out based on Evaluation formula selecting excellent to it, from this One is selected in Pareto optimal solution sets and be best suitable for the efficiently and smoothly operated optimal solution of hydraulic turbine multi-state, so as to obtain final The optimal hydraulic model of optimization design.

Wherein, the design variable in above step, parameter constraints and object function are as follows：

(1) design variable：X=[D₁,D₂,α₁,α₂,R₁,R₂,b₂,β₁,Φ,Z]^T,

(2) design variable constraints is respectively：

0.8Φ⁰≤Φ≤1.2Φ⁰；

INT(0.6Z⁰)≤Z≤INT(1.4Z⁰)；

The initial value of geometric parameter in formulaThe tradition such as the velocity-coefficient method by pump or turbine class product Design method is designed or type selecting conversion gained.

(3) object function：

1) it is up to object function with the efficiency under tri- flow rate working conditions points of 0.7Q, 1.0Q, 1.2Q：

X-dimension is 10 design variable vector in formula；

The object function of F (X)-optimization；

O-initial value；

I-represent each flow rate working conditions point of 0.7Q, 1.0Q, 1.3Q, 1,2,3；

W_iThe weight factor of-each operating point, is determined by super-transitive approximate method；

P-penalty factor；

H-turbine pressure head；

2) with the minimum object function of radial load of tri- flow rate working conditions points of 0.7Q, 1.0Q, 1.2Q：

3) with the minimum object function of axial force of tri- flow rate working conditions points of 0.7Q, 1.0Q, 1.3Q：

Wherein, the test method in Step1 is surpassed using the Latin hypercube method for designing of optimization with an improved random Latin Cube design uniformity so that the factor and response fitting it is more accurate, the method for designing have preferably it is space filling And harmony.

Wherein, the Evaluation formula in Step5 realizes that step is as follows：

(1) the Pareto optimal solution sets for obtaining NSGA-II algorithms are translated into decision-making square as scheme collection Battle array, standardization processing is carried out using vector standardization to decision matrix, obtains specified decision matrix B=(b_ij)_n×m, n is side Case number, m is attribute number；

(2) difference method is adjusted to determine the m subjective preferences weight vector ω=(ω of attribute using expert₁,ω₂,…,ω_m)^T；

(3) the objective weight μ=(μ of attribute is determined based on specified decision matrix and using information Entropy Method₁,μ₂,…,μ_m )^T；

(4) subjective and objective weight is combined using least square method method, obtains comprehensive subjective and objective making decision Combination w=(the w of every attribute₁,w₂,…,w_m)^T。

The beneficial effect of patent of the present invention is：

1) Approximate prediction model is set up and updated to training BP neural network optimized algorithm, and the CFD for reducing big load is calculated, Calculated with fewer number, obtain enough precision of predictions.

2) combination of BP neural network optimized algorithm and the multi-objective genetic algorithms of NSGA- II, it is ensured that whole algorithm exists Constantly while search optimal solution set, the diversity of population is taken into account, precision is also lifted.

3) solution is optimized to turbo wheel using the multi-objective genetic algorithms of NSGA- II, can preferably solves fluid power saturating Vibration, the power output easily occurred when deviateing optimum operating condition in flat impeller traditional design and difficulty and the hydraulic turbine fortune of type selecting It is unstable, the problems such as efficiency is significantly relatively low, take into account efficiency of turbine, axial force and radial load, turbine can be effectively improved in multiplexing Service ability under condition.

Brief description of the drawings

Fig. 1 is a kind of general flow chart of hydraulic turbine impeller multi-state multi-objective optimization design of power method in the present invention；

Specific embodiment

The present invention is further elaborated below according to accompanying drawing：

It is as shown in Figure 1 a kind of main-process stream of hydraulic turbine impeller multi-state multi-objective optimization design of power method in the present invention Figure, its key step flow is as follows：

Step1:Hydraulic turbine Impeller Design variable, object function and constraint are determined, with the outer diameter D of impeller₂, outlet is directly Footpath D₁, two inclination alphas of front and rear cover plate₁、α₂, two arc radius R on front and rear cover plate streamline₁、R₂, entrance width b₂, blade Outlet laying angle β₁, subtended angle of blade Φ and number of blade Z is design variable；Secondly, using test design method in design variable Space in generate design variable test sample；Finally, the variable in hydraulic turbine initial model is entered using Pro/E softwares Row batch Parametric designing；

Step2:The autoexec of grid and CFD software is made, realizes that all of test sample of hydraulic turbine is carried out certainly Dynamic mesh is divided and calculated with performance CFD, finally obtains the performance number of all models；

Step3:After obtaining the corresponding performance parameter of all sample patterns, geometric parameter and correspondence liquid according to test sample The performance parameter training BP neural network of power turbine, the |input paramete with design variable as approximate model, with corresponding property Energy parameter is output parameter, sets up approximate agent model；

Step4:Performance Evaluation and genetic evolutionary operations are carried out to each individuality in population, wherein individual Performance Evaluation Function is by training and be embedded in agent model in genetic algorithm and realize.Calculated using the multi-objective Genetics of NSGA- II When method is optimized to impeller geometric parameter, initial population is first randomly generated, the individual UVR exposure in initial population；Secondly, it is right Each individuality carries out Performance Evaluation in population, and performance parameter is related to object function；Finally, there are the gene of each individuality And genetic manipulation is carried out to population until meeting algorithm stop criterion according to its principle after object function, obtain multiple-objection optimization Pareto optimization disaggregation；

Step5:Pareto for obtaining optimizes disaggregation, using carrying out based on Evaluation formula selecting excellent to it, from this One is selected in Pareto optimal solution sets and be best suitable for the efficiently and smoothly operated optimal solution of hydraulic turbine multi-state, so as to obtain final The optimal hydraulic model of optimization design.

Wherein, the design variable in above step, parameter constraints and object function are as follows：

(1) design variable：X=[D₁,D₂,α₁,α₂,R₁,R₂,b₂,β₁,Φ,Z]^T,

(2) design variable constraints is respectively：

0.8Φ⁰≤Φ≤1.2Φ⁰；

INT(0.6Z⁰)≤Z≤INT(1.4Z⁰)；

The initial value of geometric parameter in formulaThe tradition such as the velocity-coefficient method by pump or turbine class product Design method is designed or type selecting conversion gained.

(3) object function：

1) it is up to object function with the efficiency under tri- flow rate working conditions points of 0.7Q, 1.0Q, 1.2Q：

X-dimension is 10 design variable vector in formula；

The object function of F (X)-optimization；

O-initial value；

I-represent each flow rate working conditions point of 0.7Q, 1.0Q, 1.3Q, 1,2,3；

W_iThe weight factor of-each operating point, is determined by super-transitive approximate method；

P-penalty factor；

H-turbine pressure head；

2) with the minimum object function of radial load of tri- flow rate working conditions points of 0.7Q, 1.0Q, 1.2Q：

3) with the minimum object function of axial force of tri- flow rate working conditions points of 0.7Q, 1.0Q, 1.3Q：

Wherein, the test method in Step1 is surpassed using the Latin hypercube method for designing of optimization with an improved random Latin Cube design uniformity so that the factor and response fitting it is more accurate, the method for designing have preferably it is space filling And harmony.

Wherein, the Evaluation formula in Step5 realizes that step is as follows：

(1) the Pareto optimal solution sets for obtaining NSGA-II algorithms are translated into decision-making square as scheme collection Battle array, standardization processing is carried out using vector standardization to decision matrix, obtains specified decision matrix B=(b_ij)_n×m, n is side Case number, m is attribute number；

(2) difference method is adjusted to determine the m subjective preferences weight vector ω=(ω of attribute using expert₁,ω₂,…,ω_m)^T；

(3) the objective weight μ=(μ of attribute is determined based on specified decision matrix and using information Entropy Method₁,μ₂,…,μ_m )^T；

(4) subjective and objective weight is combined using least square method method, obtains comprehensive subjective and objective making decision Combination w=(the w of every attribute₁,w₂,…,w_m)^T。

The invention is not restricted to above-described embodiment, also comprising other embodiments in the range of present inventive concept and variation.

Claims (5)

1. a kind of hydraulic turbine impeller multi-state multi-objective optimization design of power method, it is characterised in that：Hydraulic turbine pump is determined first Impeller key geometric parameter generates sample and it is carried out to screen as optimization design variable and sample number, using test method To satisfactory sample point, hydraulic turbine integrated automation moulding, mesh generation and CFD are carried out to initial sample and calculates acquisition Respective performances parameter, sets up optimization sample database, imports BP neural network module and optimized algorithm is set up after learning training Approximate agent model, is finally embedded in the multi-objective genetic algorithms of NSGA- II, with tri- flow works of 0.7Q, 1.0Q, 1.2Q The optimal optimizing solution for being target, carrying out genetic algorithm of efficiency, radial load and axial force under condition point, solves impeller overall most Excellent disaggregation.

A kind of hydraulic turbine impeller multi-state multi-objective optimization design of power method, implements step as follows：

Step1:Determine hydraulic turbine Impeller Design variable, object function and parameter constraints；Secondly, using experimental design Method generates the test sample of design variable in the space of design variable；Finally, it is initial to hydraulic turbine using Pro/E softwares Variable in model carries out batch Parametric designing；

Step2:The autoexec of grid and CFD software is made, realizes that all of test sample of hydraulic turbine carries out Autonet Lattice are divided and calculated with performance CFD, finally obtain the performance number of all models；

Step3:After obtaining the corresponding performance parameter of all sample patterns, the geometric parameter and correspondence fluid power according to test sample are saturating Flat performance parameter training BP neural network, the |input paramete with design variable as approximate model is joined with corresponding performance Number is output parameter, sets up approximate agent model；

Step4:Performance Evaluation and genetic evolutionary operations are carried out to each individuality in population, wherein individual performance evaluation function It is by training and be embedded in agent model in genetic algorithm and realize.Using the multi-objective genetic algorithms pair of NSGA- II When impeller geometric parameter is optimized, initial population is first randomly generated, the individual UVR exposure in initial population；Secondly, to population In each individuality carry out Performance Evaluation, performance parameter is related to object function；Finally, have each individuality gene and Genetic manipulation is carried out to population until meeting algorithm stop criterion according to its principle after object function, multiple-objection optimization is obtained Pareto optimizes disaggregation；

Step5:Pareto for obtaining optimizes disaggregation, using carrying out based on Evaluation formula selecting excellent to it, from the Pareto One is selected in optimal solution set and be best suitable for the efficiently and smoothly operated optimal solution of hydraulic turbine multi-state, so as to obtain final optimization pass set The optimal hydraulic model of meter.

2. as described in claim 1 a kind of hydraulic turbine impeller multi-state multi-objective optimization design of power method, it is characterised in that： The design variable, parameter constraints and object function are as follows：

(1) design variable：X=[D₁,D₂,α₁,α₂,R₁,R₂,b₂,β₂,Φ,Z]^T,

X-dimension is 10 design variable vector；

D₁- turbo wheel outlet diameter, m；

D₂- turbo wheel inlet diameter, m；

α₁- turbo wheel front shroud inclination angle, °；

α₂- turbo wheel back shroud inclination angle, °；

R₁- turbo wheel front shroud arc radius, m；

R₂- turbo wheel back shroud arc radius, m；

b₂- turbo wheel inlet axial width, m；

β₂- turbo wheel vane inlet laying angle, °；

Φ-turbo wheel subtended angle of blade, °；

Z-turbo wheel the number of blade, piece；

(2) design variable constraints is respectively：

$0.8D_1^0\≤D_1\≤1.2D_1^0;$

$0.8D_2^0\≤D_2\≤1.2D_2^0;$

$0.95{\mathrm{\α}}_1^0\≤{\mathrm{\α}}_1\≤1.05{\mathrm{\α}}_1^0;$

$0.95{\mathrm{\α}}_2^0\≤{\mathrm{\α}}_2\≤1.05{\mathrm{\α}}_2^0;$

$0.8R_1^0\≤R_1\≤1.2R_1^0;$

$0.8R_2^0\≤R_2\≤1.2R_2^0;$

$0.8b_2^0\≤b_2\≤1.2b_2^0;$

$0.8{\mathrm{\β}}_2^0\≤{\mathrm{\β}}_2\≤1.2{\mathrm{\β}}_2^0;$

0.8Φ⁰≤Φ≤1.2Φ⁰；

INT(0.6Z⁰)≤Z≤INT(1.4Z⁰)；

The initial value of geometric parameter in formulaThe traditional designs such as the velocity-coefficient method by pump or turbine class product Method is designed or type selecting conversion gained.

(3) object function：

1) it is up to object function with the efficiency under tri- flow rate working conditions points of 0.7Q, 1.0Q, 1.2Q：

$\left\{\begin{array}{c}X={\[D_1,D_2,{\mathrm{\α}}_1,{\mathrm{\α}}_2,R_1,R_2,b_2,{\mathrm{\β}}_2,\mathrm{\Φ},Z\]}^T;\\ f_1\left(X\right)=Min\[\underset{i=1}{\overset{3}{\Σ}}{W}_i({\mathrm{\η}}_i^0-{\mathrm{\η}}_i)+\underset{i=1}{\overset{2}{\Σ}}\underset{j=i+1}{\overset{3}{\Σ}}\left|{\mathrm{\η}}_j-{\mathrm{\η}}_i\right|+P\underset{i=1}{\overset{3}{\Σ}}\left|Min(0,\mathrm{\Δ}H-|H_i-H_i^0\left|\right)\right|\]\end{array}\right.$

X-dimension is 10 design variable vector in formula；

The object function of F (X)-optimization；

O-initial value；

I-represent each flow rate working conditions point of 0.7Q, 1.0Q, 1.3Q, 1,2,3；

W_iThe weight factor of-each operating point, is determined by super-transitive approximate method；

P-penalty factor；

H-turbine pressure head；

2) with the minimum object function of radial load of tri- flow rate working conditions points of 0.7Q, 1.0Q, 1.2Q：

$f_2\left(X\right)=Max\[\underset{i=1}{\overset{3}{\Σ}}{W}_i({\mathrm{Fr}}_i^0-{\mathrm{Fr}}_i)+\underset{i=1}{\overset{2}{\Σ}}\underset{j=i+1}{\overset{3}{\Σ}}\left|{\mathrm{Fr}}_j-{\mathrm{Fr}}_i\right|+P\underset{i=1}{\overset{3}{\Σ}}\left|Min(0,\mathrm{\Δ}H-|H_i-H_i^0\left|\right)\right|\]$

3) with the minimum object function of axial force of tri- flow rate working conditions points of 0.7Q, 1.0Q, 1.2Q：

$f_3\left(X\right)=Max\[\underset{i=1}{\overset{3}{\Σ}}{W}_i({\mathrm{Fz}}_i^0-{\mathrm{Fz}}_i)+\underset{i=1}{\overset{2}{\Σ}}\underset{j=i+1}{\overset{3}{\Σ}}\left|{\mathrm{Fz}}_j-{\mathrm{Fz}}_i\right|+P\underset{i=1}{\overset{3}{\Σ}}\left|Min(0,\mathrm{\Δ}H-|H_i-H_i^0\left|\right)\right|\].$

3. as described in claim 1 a kind of hydraulic turbine impeller multi-state multi-objective optimization design of power method, it is characterised in that： The initial value of the geometric parameter constraintThe traditional designs such as the velocity-coefficient method by pump or turbine class product Method is designed or type selecting conversion gained.

4. as described in claim 1 a kind of hydraulic turbine impeller multi-state multi-objective optimization design of power method, it is characterised in that： Test method in the Step1 is designed using the Latin hypercube method for designing of optimization with an improved random Latin hypercube Uniformity so that the fitting of the factor and response is more accurate, and the method for designing has preferable space filling and harmony.

5. as described in claim 1 a kind of hydraulic turbine impeller multi-state multi-objective optimization design of power method, it is characterised in that： Evaluation formula in Step5 realizes that step is as follows：

(1) the Pareto optimal solution sets for obtaining NSGA-II algorithms are translated into decision matrix as scheme collection, profit Standardized with vector carries out standardization processing to decision matrix, obtains specified decision matrix B=(b_ij)_n×m, n is scheme Number, m is attribute number；

(2) difference method is adjusted to determine the m subjective preferences weight vector ω=(ω of attribute using expert₁,ω₂,…,ω_m)^T；

(3) the objective weight μ=(μ of attribute is determined based on specified decision matrix and using information Entropy Method₁,μ₂,…,μ_m)^T；

(4) subjective and objective weight is combined using least square method method, obtains the items of comprehensive subjective and objective making decision Combination w=(the w of attribute₁,w₂,…,w_m)^T。