CN112948997B - Multi-objective adaptive clustering optimization method and system - Google Patents
Multi-objective adaptive clustering optimization method and system Download PDFInfo
- Publication number
- CN112948997B CN112948997B CN202110207836.9A CN202110207836A CN112948997B CN 112948997 B CN112948997 B CN 112948997B CN 202110207836 A CN202110207836 A CN 202110207836A CN 112948997 B CN112948997 B CN 112948997B
- Authority
- CN
- China
- Prior art keywords
- target
- constraint
- temperature
- deviation
- clustering
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 61
- 238000005457 optimization Methods 0.000 title claims abstract description 40
- 230000003044 adaptive effect Effects 0.000 title claims description 17
- 239000011159 matrix material Substances 0.000 claims abstract description 100
- 238000004458 analytical method Methods 0.000 claims abstract description 24
- 230000008569 process Effects 0.000 claims abstract description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 15
- 238000010438 heat treatment Methods 0.000 claims description 8
- 238000012546 transfer Methods 0.000 claims description 8
- 238000007621 cluster analysis Methods 0.000 claims description 7
- 239000007788 liquid Substances 0.000 claims description 7
- 238000010276 construction Methods 0.000 claims description 6
- 238000005192 partition Methods 0.000 claims description 6
- 239000012530 fluid Substances 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 3
- 238000013461 design Methods 0.000 abstract description 24
- 230000035945 sensitivity Effects 0.000 abstract description 3
- 239000013598 vector Substances 0.000 description 6
- 230000006872 improvement Effects 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 230000000875 corresponding effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000002596 correlated effect Effects 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 230000005611 electricity Effects 0.000 description 2
- KEBHLNDPKPIPLI-UHFFFAOYSA-N hydron;2-(3h-inden-4-yloxymethyl)morpholine;chloride Chemical compound Cl.C=1C=CC=2C=CCC=2C=1OCC1CNCCO1 KEBHLNDPKPIPLI-UHFFFAOYSA-N 0.000 description 2
- 230000008878 coupling Effects 0.000 description 1
- 230000001808 coupling effect Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000000265 homogenisation Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000002195 synergetic effect Effects 0.000 description 1
- 239000011215 ultra-high-temperature ceramic Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Hardware Design (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Complex Calculations (AREA)
Abstract
The invention provides a multi-target self-adaptive clustering optimization method and a system, wherein the method comprises the following steps: carrying out priority classification on the N targets according to different clusters; generating a target correlation characteristic matrix according to the N targets with the divided priorities; and performing target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a final target clustering result. According to the method, the relation of each target can be distinguished based on the final target clustering result under the condition that the related subject and related field knowledge are not solved, and for the target clustering cluster with strong correlation, the sensitivity of each target to variables is almost consistent, so that redundant targets which have little influence on the result can be deleted in the process of optimizing design, and the design complexity is reduced; for the target cluster with weak correlation, setting the target cluster with weak correlation as the same priority during optimization design, and regarding the target cluster as the same type of target; for the non-correlation target, a target needing priority guarantee is found during design, and a higher priority is set to obtain a better scheme.
Description
Technical Field
The invention relates to the technical field of multi-objective optimization, in particular to a multi-objective self-adaptive cluster optimization method and system.
Background
The complex product is a complex integrating multiple discipline knowledge, such as a heat exchanger. Due to the complex coupling relationship between disciplines, designers need to consider the interaction and coupling effect between subsystems simultaneously when designing, and trade off between conflicting target requirements. But it is difficult for designers to master the knowledge in various areas of each subject and to make reliable, flexible and modifiable design decisions.
Multi-objective problems can be divided into two categories: a method for performance and design improvement of solution algorithms. In the papers on solution algorithms, the main focus is to explore the Pareto solution set and evaluate the performance of the algorithm by criteria such as solution optimality, solution diversity and computational power. But focus on the solution algorithm with limitations, focusing only on identifying the vicinity of the Pareto frontier, and lack discussion and decision support on how to use solutions in the vicinity of Pareto. For approaches to design improvement, more attention is often paid to improving the design or gaining more knowledge about the problem. The limitation of this approach is that it relies too heavily on disciplinary knowledge or case analysis of the relevant professions to explore the composition of objects or the decomposition of problems, so that the versatility and reusability of the method is reduced. There is therefore a need for a method that combines both algorithmic and solution improvements.
Disclosure of Invention
The invention aims to provide a multi-target self-adaptive clustering optimization method and a multi-target self-adaptive clustering optimization system so as to improve the universality and reusability.
In order to achieve the above object, the present invention provides a multi-objective adaptive cluster optimization method, including:
step S1: carrying out priority classification on the N targets according to different clusters;
step S2: generating a target correlation characteristic matrix according to the N targets with the priority;
and step S3: and performing target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a final target clustering result.
Optionally, the performing priority ranking on the N targets according to different clusters specifically includes:
step S11: constructing a multi-target description problem based on multiple targets;
step S12: constructing a compromise decision problem according to the multi-target description problem;
step S13: and according to the last target clustering result, carrying out priority classification on the N targets according to different clusters by utilizing a priority deviation function in the compromise decision problem.
Optionally, the generating a target correlation characteristic matrix according to the prioritized N targets specifically includes:
step S21: initializing, and setting the fraction H of the target division to be 1;
step S22: calculating a weight set according to the fraction of the target division by adopting a weight generation algorithm, and giving an Archimedes deviation function;
step S23: solving the Archimedes deviation function endowed with the weight according to a self-adaptive linear algorithm to obtain an optimized solution of each target under each weight group;
step S24: calculating a target deviation matrix according to the optimal solution and a target expected value;
step S25: standardizing the target deviation matrix to obtain a standardized deviation matrix;
step S26: performing correlation calculation according to the standardized deviation matrix to obtain a target correlation characteristic matrix;
step S27: calculating the standard deviation of the target correlation characteristic matrix, and outputting the target correlation characteristic matrix when the difference value of the two adjacent standard deviations is within a set error range; when the difference between the two adjacent standard deviations is not within the set error range, the number of copies H of the target division is increased, and the process returns to step S22.
Optionally, the performing, by using a hierarchical clustering method, target clustering analysis according to the target correlation characteristic matrix to obtain a final target clustering result specifically includes:
step S31: performing target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a plurality of groups of target clustering results;
step S32: judging whether each target priority is traversed or not; if not, returning to the step S13; if the target clustering results are traversed, grading the multiple groups of target clustering results to obtain target clustering results of different grading conditions, and executing the step S33;
step S33: judging whether a convergence condition is reached; if the convergence condition is reached, taking the target clustering results of different current grading conditions as final target clustering results; and if the convergence condition is not met, taking the target clustering result with the largest occurrence frequency in the target clustering results of different current grading conditions as the last target clustering result, and returning to the step S13.
Optionally, the optimization solution and the target expectation are usedCalculating a target deviation matrix by using the values; the target deviation matrix is D d =[d 1 ,…,d N ];
By usingCalculating the deviation d of the target i i (ii) a Wherein f is i Represents the target optimization solution, F i Indicating the expected value of the object i.
The invention also provides a multi-target self-adaptive clustering optimization system, which comprises:
the priority grading module is used for carrying out priority grading on the N targets according to different clusters;
the target correlation characteristic matrix generation module is used for generating a target correlation characteristic matrix according to the N targets with the divided priorities;
and the target clustering analysis module is used for carrying out target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a final target clustering result.
Optionally, the priority module specifically includes:
the multi-target description problem construction unit is used for constructing a multi-target description problem based on multiple targets;
the compromise decision problem construction unit is used for constructing a compromise decision problem according to the multi-target description problem;
and the priority grading unit is used for grading the priorities of the N targets according to different clusters by utilizing a priority deviation function in the compromise decision problem according to the last target clustering result.
Optionally, the target correlation characteristic matrix generating module specifically includes:
an initialization unit configured to initialize, and set a score H of the target division to 1;
the assignment unit is used for calculating a weight set according to the scores divided by the targets by adopting a weight generation algorithm and endowing an Archimedes deviation function;
the solving unit is used for solving the Archimedes deviation function endowed with the weight according to a self-adaptive linear algorithm to obtain an optimized solution of each target under each weight group;
the target deviation matrix calculation unit is used for calculating a target deviation matrix according to the optimal solution and a target expected value;
the standardization processing unit is used for standardizing the target deviation matrix to obtain a standardized deviation matrix;
the target correlation characteristic matrix calculation unit is used for carrying out correlation calculation according to the standardized deviation matrix to obtain a target correlation characteristic matrix;
the first judgment unit is used for calculating the standard deviation of the target correlation characteristic matrix, and outputting the target correlation characteristic matrix when the difference value of the two adjacent standard deviations is within a set error range; and when the difference value of the standard deviations of two adjacent times is not in the set error range, increasing the number of copies H of the target partition, and returning to the assignment unit.
Optionally, the target cluster analysis module specifically includes:
the target clustering analysis unit is used for carrying out target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a plurality of groups of target clustering results;
the second judgment unit is used for judging whether each target priority is traversed or not; if not, returning to a priority level unit; if traversing, grading the multiple groups of target clustering results to obtain target clustering results of different grading conditions, and executing a 'third judging unit';
a third judging unit for judging whether a convergence condition is reached; if the convergence condition is met, taking the target clustering results of different grading conditions as final target clustering results; and if the convergence condition is not met, taking the target clustering result with the largest occurrence frequency in the target clustering results of different current grading conditions as the last target clustering result, and returning to the priority grading unit.
Optionally, the calculating a target deviation matrix according to the optimal solution and a target expected value; the target moment of deviationArray is D d =[d 1 ,…,d N ];
By usingCalculating the deviation d of the target i i (ii) a Wherein f is i Represents the target optimization solution, F i Indicating the expected value of the target i.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention can distinguish the relation of each target based on the final clustering result under the condition of not knowing the related subject and related field knowledge, and the sensitivity of each target weight to the result. After knowing the interrelationships between the targets, the designer can reselect the design, and the relationships between the targets have three cases: strong correlation, weak correlation, no correlation; for the target cluster with strong correlation, the sensitivities of all targets to variables are almost consistent, and redundant targets with little influence on results can be deleted in the process of optimization design so as to reduce the design complexity; for the target cluster with weak correlation, setting the target cluster with weak correlation as the same priority during optimization design, and regarding the target cluster as the same type of target; for the non-correlation target, the target needing priority guarantee is found during design, a higher priority is set to obtain a better scheme, and the method can be applied to a plurality of technical fields, so that multi-field universality can be realized.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
FIG. 1 is a flow chart of a multi-objective adaptive cluster optimization method according to an embodiment 1 of the present invention;
FIG. 2 is a specific flowchart of a multi-objective adaptive cluster optimization method according to embodiment 1 of the present invention;
FIG. 3 is a normalized difference vector distribution diagram according to example 1 of the present invention;
FIG. 4 is a diagram of a multi-objective adaptive clustering optimization system in accordance with embodiment 2 of the present invention;
FIG. 5 is a schematic of a heat exchanger Rankine cycle in accordance with embodiment 3 of the present invention;
FIG. 6 is a graph showing the variation of temperature T and entropy S of the Rankine cycle in embodiment 3 of the invention;
FIG. 7 is a diagram illustrating example 3 of raising target 3 by 20%.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a multi-target self-adaptive clustering optimization method and a multi-target self-adaptive clustering optimization system so as to improve the universality and reusability.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Example 1
As shown in fig. 1-2, the present invention provides a multi-objective adaptive cluster optimization method, which includes:
step S1: and carrying out priority classification on the N targets according to different clusters.
Step S2: and generating a target correlation characteristic matrix according to the N prioritized targets.
And step S3: and performing target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a final target clustering result.
The individual steps are discussed in detail below:
step S1: the method for carrying out priority classification on N targets according to different clusters specifically comprises the following steps:
step S11: constructing a multi-target description problem based on multiple targets; the multi-objective description problem comprises a decision variable set, an objective function, boundary conditions, constraints and each objective value.
The set of decision variables is: x = (x) 1 ,x 2 ,...,x n );x n Is the nth decision variable; x is a set of decision variables
The objective function f (x) = { f min (x),f max (x)};
wherein f is min (x) To minimize the target, f max (x) To maximize the goal, f N (x 1 ,x 2 ,...,x n ) Is the maximum target N, f k (x 1 ,x 2 ,...,x n ) To minimize the target n.
Boundary x = (x) 1 ,x 2 ,...,x n )∈Ω。
The constraint is a constraint function of a multi-target description problem, namely, the multi-target can only carry out value taking within the range that each variable meets the constraint function; the constraint function is:wherein h (x) is a constraint function, g (x) is a constraint function, x belongs to omega, and omega is a variable value range.
The target desired value is:
f={F min ,F max }={(F 1 ,F 2 ,..,F k ),(F k+1 ,F k+2 ...F N )};
wherein, f is the target expectation value,F min is a minimum target expectation set, F max Is a minimum target expectation set, F N Is the Nth target expectation value, and N is the total number of targets.
Step S12: and constructing a compromise decision problem according to the multi-target description problem.
The compromise decision problem comprises: giving a condition, searching, meeting condition description, a boundary condition and a minimum optimization function;
1. given the conditions: number of system targets N, objective function f (x) = { f min (x),f max (x) The system constraint function g (x) ≦ 0 and h (x) =0, the bias function Z = (d) - ,d + )。
2. Searching: system design variable x = (x) 1 ,x 2 ,...,x n ) Deviation variable d i 。
3. And satisfying the following condition:
when F is present min At 0, the objective formula for pursuit of minimization is revised as:
wherein f is i (x) For the purpose of the (i) th object,respectively, the lower deviation and the upper deviation of the target i.
when F is present max At 0, the objective equation for pursuit maximization is revised as:
4. boundary conditions: x = ∈ Ω; d is a radical of - ·d + =0,0≤d - ,d + ≤1。
5. Priority bias function:
Z(d - ,d + )=[f 1 (d - ,d + ),...,f k (d - ,d + )]k denotes the divided target hierarchy, f k (d - ,d + ) Objects representing the kth object level, Z 1 (d - ,d + ) Is a priority bias function.
The compromise decision problem is a design specification model, can be regarded as a combination of mathematical programming and target programming, and has two modeling solving modes based on Archimedes form modeling and priority form modeling.
Step S13: and according to the last target clustering result, carrying out priority classification on the N targets according to different clusters by using a priority deviation function in the compromise decision problem.
Initially, all targets are set to the same level, that is, the priorities of the N targets are all the same, and a target cluster is initialized to 1.
Step S2: generating a target correlation characteristic matrix according to the prioritized N targets, specifically comprising:
step S21: initialization, the score H of the target partition is set to 1.
Step S22: calculating a corresponding weight vector value given to each level target according to the score of target division by adopting a weight generation algorithm, and further determining an Archimedes deviation function, wherein the weight generation algorithm specifically comprises the following steps:
1. for N targets, dividing each target into H shares, then m sets of weights are generated, the specific formula is:
2. Let the jth group weight point be s j =(s 1j ,s 2j ,...,s Nj ) Is provided withWherein s is ij Is the weight vector of target i under jth group of weights,S=[s 1 ,…,s m ] T ,s m each target weight matrix being an mth set of weights.
4. For each x ∈ x, the operation is such thatWherein j is more than or equal to 1 and less than or equal to m, and i is more than or equal to 1 and less than or equal to N.
5. Obtaining coordinate values of each objective function:obtain a set of weights S, S ij The weight value of the target i under the jth group of weights.
6. Determining an archimedes deviation function:
wherein,andrespectively an upper weight and a lower weight, Z, of the object i 2 (d - ,d + ) Is an archimedean deviation function.
Step S23: and solving the Archimedes deviation function endowed with the weight according to a self-adaptive linear algorithm to obtain an optimized solution of each target under each weight group.
Step S24: calculating a target deviation matrix according to the optimal solution and a target expected value; the target deviation matrix is D d =[d 1 ,…,d N ]。
Specifically, utilizeCalculating the deviation d of the target i i (ii) a Wherein f is i Represents the target optimization solution, F i Indicating the expected value of the target i.
Step S25: target deviation matrix D d Carrying out standardization processing to obtain a standardized deviation matrix D' d The normalized deviation matrix is D' d =[d′ 1 ,d′ 2 ,...,d′ N ]。
Specifically, utilizeCalculating a normalized deviation d 'of target i' i (ii) a Wherein the deviation d of the target i i =[d i1 ,d i2 ,...,d im ] T ,d im Deviation corresponding to target i under mth set of weights, d ij The deviation corresponding to the target i under the jth group weight is m, and the total weight group number is m.
Step S26: according to the standardized deviation matrix D' d Performing correlation calculation to obtain a target correlation characteristic matrix D Cor 。
Taking the target a and the target b as an example, a rectangular coordinate system is constructed by taking two targets as x and y axes, and the deviation values under the ownership recombination are put on a rectangular coordinate system, wherein each target is a target bThe point coordinate is (d' a ,d′ b )={(d′ a1 ,d′ b1 ),(d′ a2 ,d′ b2 ),...,(d′ am ,d′ bm ) H, the distribution of m points, see in particular fig. 3, the closer the two target values are, the closer the distribution of normalized deviation points is to the diagonal, thus calculating the target correlation characteristic matrix D Cor The concrete formula of (1) is as follows:
wherein, d' aj Is the normalized deviation, d ', of the target a under the weight of the j-th group' bj The normalized deviation of the target b under the jth group weight is m, which is the total group number of weights.
As shown in fig. 3, the size of the straight line II' is the value of the distance method of the objects a =1 and b =2 under the ith group of weights, and sin θ corresponds to the value of the sine method; the smaller the correlation characteristic value is, the higher the target correlation is; the correlation characteristic values of any two other targets can be calculated in sequence to obtain a correlation characteristic matrix.
Step S27: calculating the standard deviation of the target correlation characteristic matrix, and when the difference value of the two adjacent standard deviations is within a set error range, indicating that the weight quantity is proper, and outputting the target correlation characteristic matrix; when the difference between the two adjacent standard deviations is not within the set error range, it indicates that the weight number is not appropriate, the number of copies H of the target partition is increased, and the process returns to step S22 to generate more sets of weight numbers.
And step S3: performing target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a final target clustering result, which specifically comprises the following steps:
step S31: and (4) performing target clustering analysis by adopting a hierarchical clustering method according to the target correlation characteristic matrix to obtain target clustering results of different current grading conditions.
Step S32: judging whether each target priority is traversed or not; if not, returning to the step S13; if so, step S33 is executed.
Step S33: judging whether a convergence condition is reached; if the convergence condition is reached, taking the target clustering results of different current grading conditions as final target clustering results; and if the convergence condition is not met, taking the target clustering result with the largest occurrence frequency in the target clustering results of different grading situations as the last target clustering result, and returning to the step S13.
In this embodiment, the convergence condition includes two conditions: one is that the number of iterations is greater than the set maximum number of cycles; the second is that the difference value of two adjacent target clustering results is in a set range or the standard deviation of two adjacent target clustering results reaches a set required value.
Example 2
As shown in fig. 4, the present invention further provides a multi-objective adaptive cluster optimization system, which includes:
and the priority grading module 1 is used for carrying out priority grading on the N targets according to different clusters.
And the target correlation characteristic matrix generating module 2 is used for generating a target correlation characteristic matrix according to the N prioritized targets.
And the target clustering analysis module 3 is used for carrying out target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a final target clustering result.
As an embodiment, the priority ranking module 1 of the present invention specifically includes:
and the multi-target description problem construction unit is used for constructing the multi-target description problem based on multiple targets.
And the compromise decision problem construction unit is used for constructing a compromise decision problem according to the multi-target description problem.
And the priority grading unit is used for grading the priorities of the N targets according to different clusters by utilizing a priority deviation function in the compromise decision problem according to the last target clustering result.
As an embodiment, the target correlation characteristic matrix generating module 2 of the present invention specifically includes:
and the initialization unit is used for initializing and setting the fraction H of the target division to be 1.
And the assignment unit is used for calculating a weight set according to the fraction of the target division by adopting a weight generation algorithm and endowing an Archimedes deviation function.
And the solving unit is used for solving the Archimedes deviation function endowed with the weight according to the adaptive linear algorithm to obtain an optimized solution of each target under each weight group.
And the target deviation matrix calculation unit is used for calculating a target deviation matrix according to the optimal solution and the target expected value.
And the standardization processing unit is used for carrying out standardization processing on the target deviation matrix to obtain a standardized deviation matrix.
And the target correlation characteristic matrix calculation unit is used for performing correlation calculation according to the standardized deviation matrix to obtain a target correlation characteristic matrix.
The first judgment unit is used for calculating the standard deviation of the target correlation characteristic matrix, and outputting the target correlation characteristic matrix when the difference value of the two adjacent standard deviations is within a set error range; and when the difference value of the standard deviations of two adjacent times is not in the set error range, increasing the number of copies H of the target partition, and returning to the assignment unit.
As an implementation manner, the target cluster analysis module 3 of the present invention specifically includes:
and the target clustering analysis unit is used for carrying out target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a plurality of groups of target clustering results.
The second judgment unit is used for judging whether each target priority is traversed or not; if not, returning to the priority level unit; and if the target clustering results are traversed, grading the multiple groups of target clustering results to obtain target clustering results of different grading conditions at present, and executing a third judgment unit.
A third judging unit for judging whether a convergence condition is reached; if the convergence condition is reached, taking the target clustering results of different current grading conditions as final target clustering results; and if the convergence condition is not met, taking the target clustering result with the largest occurrence frequency in the target clustering results of different current grading conditions as the last target clustering result, and returning to the priority grading unit.
Example 3
Small "power plant" systems may have a variety of applications, and these systems use small generators to generate electricity or use the generated electricity directly mechanically. Building systems around a rankine cycle is a common approach given the available heat sources. A rankine cycle is a mathematical representation of a heat engine that converts heat into mechanical work when undergoing a phase change. A schematic of a rankine cycle, as shown in fig. 5. The main components of the system are a power generating turbine, a pump to pressurize the flow to the turbine and two heat exchangers, a condenser and a heater.
An ideal rankine cycle involves 4 processes, as shown in fig. 6, with two adiabatic isentropic processes (constant entropy) and two isobaric processes (constant pressure). (1) - (2) adiabatic from PMIN to PMAX, (2) - (4) adding isobaric heat to TMAX in a heat exchanger, (4) - (5) adiabatic expansion in the turbine, expansion from PMAX to PMIN, generating power, and possibly wet steam exiting the turbine, and (5) - (1) isobaric heat loss in the condenser.
The P1-P51 related system variable parameters are given as follows:
p1: CARNOT is Carnot cycle efficiency (%);
p2: CPEE is the specific heat value of the input line in the exchanger;
p3: CPRE is the specific heat value (J/(kg · K)) of the Rankine (output) feed line in the exchanger;
P4-P5: lower temperature limit of each fluid (K) in Rankine cycle (DBTMNR) and heat exchanger (DBTMNE): DBTMNR/DBTMNE =273.16;
P6-P7: an upper temperature limit DBTMXR/DBTMXE =2000.0 for each fluid (K) in the Rankine cycle (DBTMXR) or heat exchanger (DBTMXE);
p8: DELTLM is the logarithmic principal temperature difference (K);
P9-P13: dens, i =1,2,. 5 is density (1) - (5) (kg/m 3), respectively;
P14-P15: EDIA/ELEN is the heat exchanger diameter/length (m), respectively;
P16-P20: ENTHi, i =1,2,. 5, specific enthalpies in (1) - (5) (J/kg), respectively;
P21-P22: ENTHMX/ENTHMN is the enthalpy of TMINE/TMAXE in the exchanger;
p23: FLOWR/FLOWE is the mass flow (kg/s) of the Rankine cycle/exchanger respectively;
p24: FRMXR is the upper limit (kg/s) of Rankine cycle mass flow;
P25:f 6 heat transfer efficiency (%);
p26: PPUMP/PTURB is pump/turbine power (W), respectively;
P27-P31: PRESi, i =1,2,. 5, pressures (1) - (5) (kPa), respectively;
p32: QINR is heat transfer (W) in the heat exchanger;
p33: QUTE is heat transfer (W) of a heat exchanger;
P34-P38: QUALi, i =1,2,. 5, water flow mass (%) at (1) - (5), respectively;
P39:f 2 rankine cycle efficiency (%);
P40:f 1 rankine cycle moisture (%) in the turbine;
p41: REQPOW is the power required by the Rankine cycle (kW);
p42: RFEEDL is a calculated value of the rankine cycle length required for a given diameter (m);
p43: SAREAE is the surface area of the heat exchanger (m 2);
P44:f 4 system efficiency 1 (%);
P45:f 5 system efficiency 2 (%);
p46: TDELE =10, requirement of minimum temperature variation in heat exchanger (K);
p47: TDELC is the minimum temperature difference between the lowest temperature of the heat exchanger and the temperature of (2);
P48:f 3 heat exchanger efficiency (%);
p49: TEMPi, i =1,2, …, temperatures at (1) - (5) (K), respectively;
p50: TMINE is the exchanger minimum temperature (K);
p51: UHTC is the total heat transfer coefficient (W/(m 2. K)).
Functional relationships between the parameters and the system variables F1-F14:
F5:FLOWE·(ENTHMX-ENTHMN)=FLOWR·(ENTH4-ENTH2);
F8:PPUMP=(ENTH2-ENTH1)·FLOWR;
F9:PPUMB=(ENTH4-ENTH5)·FLOWR;
F10:QINR=FLOWR·CPRE·(TEMP4-TEMP2);
F11:QOUTE=FLOWE·CPEE(x 4 -TMINE);
F13:SAREAE=π·EDIA·ELEN;
constructing a multi-target description problem; the multi-objective description problem comprises four decision variables, one linear constraint, ten nonlinear inequality constraints and six objectives.
1. Four decision variables:
x 1 -maximum pressure in the rankine cycle variable;
x 2 -rankine cycle minimum pressure;
x 3 -rankine cycle maximum temperature;
x 4 -the maximum temperature of the heating liquid in the exchanger;
2. six objective functions:
Goal 6, maximizing heat transfer efficiency of heat exchanger —)maxf 6 (x 1 ,x 2 ,x 3 ,x 4 )。
3. Boundary conditions:
500≤x 1 ≤5000(kPa);
350≤x 2 ≤850(K);
350≤x 3 ≤850(K)。
4. ten constraint functions:
The mass of the constraint 5 and 4 is superheated steam-QUAL 4 ≥ 1.0 → g 5 (x)≤0;
Constraint 6, heating temperature differential limitation in exchanger-TMAXE-TMINE ≧ TDELE → g 6 (x)≤0;
Constraint 7, lowest temperature limit of heating liquid in exchanger-TMINE-TMEP 2 ≧ TDELC → g 7 (x)≤0;
Constraint 8, ideal Carnot cycle efficiency is greater than system efficiency 1-CARNOT is not less than f 4 →g 8 (x)≤0;
Constraint 9, ideal Carnot cycle efficiency greater than System efficiency 2-CARNOT ≧ f 5 →g 9 (x)≤0;
5. Target desired value:
f={F min ,F max }={(0),(1,1,1,1,1)}。
and (3) constructing compromise decision modeling, wherein the modeling result is as follows:
1. given conditions
Parameters including related system variables P1-P51Number, and system constraint g 1 (x)-g 10 (x)。
2. Searching
a. System variable x = (x) 1 ,x 2 ,x 3 ,x 4 );
3. And satisfying the condition specification: system object
4. boundary condition
a. System design variable boundary conditions:
500≤PMAX≤5000(kPa);
350≤TMAX≤850(K);
350≤TMAXE≤850(K)。
b. deviation variable d
5. Minimizing an optimization function
a. Archimedes deviation function:
b. priority bias function:
Z(d - ,d + )=[f 1 (d - ,d + ),...,f k (d - ,d + )]and k denotes a divided target hierarchy.
All targets are classified into the same level, namely, the priority of the six targets is the same.
According to the weight generation algorithm, the target number N =6 and H =1 in this case. Then generateAnd (4) point. The resulting points are (1,0,0,0,0,0), (0,1,0,0,0,0), (0,0,1,0,0,0), (0,0,0,1,0,0), (0,0,0,0,1,0), (0,0,0,0,0,1), respectively.
And solving the Archimedes deviation function endowed with the weight by using a DESIDS solver to obtain a group of solutions.
From the above solution results, a deviation matrix is calculated as shown in table 1. For any object I, set f i Represents the target calculation optimization result value, F i Represents an ideal target value ofObjective i is to pursue the maximum; orThe objective i is to pursue a minimum.
TABLE 1 deviation matrix
Namely:
the deviation matrix is standardized first, and then the correlation between the targets is calculated to obtain a correlation characteristic matrix.
Currently, only one set of weight values cannot be compared to determine whether the standard deviation changes, so the number of target partitions H is increased, let H =2. Returning to step s22. At this time, generateThere are 21 sets of weight values.
And repeating the steps from s22 to s27 again, and calculating the standard deviation of the target correlation characteristic matrix, wherein the difference value of the two adjacent standard deviations is in the error range, which indicates that the weight quantity requirement is met at the moment.
And (3) performing cluster analysis by using a hierarchical clustering method and taking the target correlation characteristic matrix as an input in the cluster analysis to obtain a clustering result, wherein the six targets can be divided into three types of [1,6], [2,4] and [3,5 ].
And judging whether the priority is traversed or not, wherein each target priority in the first round is 1, and therefore, the first-time traversal is executed.
The first round only produces one set of clustering results. It is clear that the convergence condition is not met, i.e. the convergence condition is to set the maximum number of cycles to 10. At present, only one group of clustering results cannot judge whether the clustering results converge, so that the clustering results of the targets in the previous round are updated, the targets are divided into three types at the moment, and the step s13 is returned to carry out target grading. Here, [1,6] is set as the first level, [2,4] is set as the second level, and [3,5] is set as the third level.
Steps s13-s31 are executed until step s32, the priority level is not traversed and the sorting is again done at the new level. Set to [3,5] first level, [2,4] second level, [1,6] third level, return to step s13. And the process is circulated until all levels are traversed.
The second iteration cases obtained after the priority traversal are grouped as [1], [2,4], [3,5,6]. And step S33 is executed again, convergence conditions are judged, the clustering results of the two times are not converged, the target clustering result with the largest occurrence frequency in the target clustering results of the current different grading conditions is used as the last target clustering result, the step S13 is returned, and the iteration is carried out again in a circulating mode.
After the third iteration, the target clustering results are [1], [2,4], [3,5,6], a convergence condition is achieved, and the final target clustering result is output.
Table 2 shows the results of clustering
It can be found by cluster analysis that targets 2 and 4 are always in one class, targets 3 and 5 are in one class, and the clustering of targets 1 and 6 is iteratively changed. This means that the objects 2,4 are highly correlated, as are the objects 3, 5. The invention was validated by the relevant discipline knowledge, see table 3.
TABLE 3 parameter Classification Table
With knowledge of the relevant disciplines, it is true that the system efficiency indication will improve the Rankine cycle efficiency, while the heat exchange efficiency represented by targets 3,5, during system operation, the temperature exchanger and heat transfer work have a synergistic effect. Target 6 is sometimes clustered with targets 3,5 as they both relate to the efficiency of the heat exchanger. The targets 3,5 are temperature dependent and the target 6 is liquid flow dependent. On the other hand, the targets 2 and 4 are both rankine cycle efficiency indicators, and it can be determined from the formulas of the two that the effects of the two on the changes of the respective variables are almost the same.
For multi-objective parallel design problems, the objectives may represent the performance of various subsystems in the design. An improvement in one target may result in a greater sacrifice of another target or targets. Referring to FIG. 7, the designer separately increases the achievement of goal 3 by setting priorities and weights without knowing the interrelationships between the goals so that the goal bias resultsTo Target 3 efficiency increased by 20%, but target 1 and target 2 deteriorated by 80%. This may be desirable in some situations, but designers often prefer to avoid this situation because the overall performance of the design cannot be substantially improved by improving one subsystem.
By the cluster analysis method, the relation between the targets can still be found in a data analysis mode under the condition that the related subject knowledge of the heat exchanger is not cleared. The six targets are divided into three categories according to the clustering analysis result: [1],[2,4],[3,5,6]. And verified through knowledge of the relevant disciplines. After the target clustering condition is known, optimization calculation can be carried out, and a designer is helped to make a decision, and the method comprises the following three aspects:
1. and reasonably distributing the target priority and the weight according to the clustering condition. The targets in the same group have strong correlation and are given the same priority in design. In the case where a certain target achievement degree is not specified in advance, [3,5,6] may be set as a first priority, [2,4] may be set as a second priority, and [1] may be set as a third priority. Therefore, more targets can be ensured to reach the optimal condition as far as possible, and the comprehensive performance of the system is integrally improved.
2. And (4) carrying out target reduction, and deleting targets with small influence on the result, so as to better and more quickly find a design scheme. In this case, it can be found through the correlation characteristic matrix that the distance between the correlation characteristics of the target 2 and the target 4 is 0, which indicates that the two targets are highly correlated, the effect of variable change on the result is almost consistent, one of the targets can be regarded as a redundant target, and the target 4 can be deleted during optimization calculation, so that the six-target problem is reduced to five targets, and the design complexity is reduced; in addition, it is found that there is some correlation between object 3 and object 5, and the priorities of the two objects are kept consistent when designing, and there is no correlation between object 1 and object 6. Designers can give an optimal scheme by considering the target priority and the weight according to the target clustering condition; when considering the priority, in order to satisfy as many targets as possible, the designer may set the cluster containing a larger number of targets as a higher priority.
3. And the computational complexity is reduced. For the equalization of all six targets, the digital equalization scenario is theoretically 6! =720. By the method of the invention, the targets are graded based on the clustering result, the clustering result is 3 types, and the number of homogenization schemes is reduced to 3! And (6). Within each level, the targets are combined using the weight vectors. The stop criterion of weight vector generation (step S27) helps to prevent the designer from using too many unnecessary weight vectors. Using the angle-based correlation method to compute the correlation property matrix, 142 design scenarios were explored and converged in three iterations, and therefore the number of designs was reduced from 720 to 142.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (8)
1. A multi-objective adaptive cluster optimization method is characterized by comprising the following steps:
step S1: carrying out priority classification on the N targets according to different clusters;
step S2: generating a target correlation characteristic matrix according to the N targets with the divided priorities;
and step S3: performing target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a final target clustering result;
the method for carrying out priority classification on the N targets according to different clusters specifically comprises the following steps:
step S11: constructing a multi-target description problem based on multiple targets;
step S12: constructing a compromise decision problem according to the multi-target description problem;
step S13: according to the last target clustering result, carrying out priority classification on N targets according to different clusters by using a priority deviation function in the compromise decision problem;
the multi-target description problem comprises four decision variables, one linear constraint, ten nonlinear inequality constraints and six targets;
four decision variables:
x 1 : maximum pressure in rankine cycle variables;
x 2 : rankine cycle minimum pressure;
x 3 : rankine cycle maximum temperature;
x 4 : the maximum temperature of the heating liquid in the exchanger;
the target function is six:
goal 1, minimizing steam of the turbine: minf 1 (x 1 ,x 2 ,x 3 ,x 4 );
Goal 2, maximizing rankine cycle efficiency: maxf 2 (x 1 ,x 2 ,x 3 ,x 4 );
Goal 3, maximizing heat exchanger efficiency: maxf 3 (x 1 ,x 2 ,x 3 ,x 4 );
Goal 4, maximize system efficiency index 1: maxf 4 (x 1 ,x 2 ,x 3 ,x 4 );
Goal 5, maximize system efficiency index 2: maxf 5 (x 1 ,x 2 ,x 3 ,x 4 );
Goal 6, maximizing the heat transfer efficiency of the heat exchanger: maxf 6 (x 1 ,x 2 ,x 3 ,x 4 );
Boundary conditions:
500≤x 1 ≤5000(kPa);
350≤x 2 ≤850(K);
350≤x 3 ≤850(K);
ten constraint functions:
constraint 1, upper temperature increase limit: TMAXE-TMAX ≧ DELTLM → g 1 (x)≤0;
Constraint 2, moisture in the turbine is less than the upper limit: RCMIT ≦ TMXL → g 2 (x)≤0;
Constraint 3, rankine cycle mass flow less than upper limit: FLOWR ≦ FRMXR → g 3 (x)≤0;
Constraint 4, temperature between heat exchanger and turbine is equal to or greater than at heat exchanger: TEMP4 ≧ TEMP3 → g 4 (x)≤0;
Constraint 5, the mass between the heat exchanger and the turbine is superheated steam: QUAL4 is more than or equal to 1.0 → g5 (x) is less than or equal to 0;
constraint 6, heating temperature differential limit in exchanger: TMAXE-TMINE ≥ TDELE → g 6 (x)≤0;
Constraint 7, exchangeHeating liquid lowest temperature limit in the vessel: TMINE-TMEP 2. Gtoreq.TDELC → g 7 (x)≤0;
Constraint 8, ideal carnot cycle efficiency greater than system efficiency 1: CARNOT is more than or equal to f 4 →g 8 (x)≤0;
Constraint 9, ideal carnot cycle efficiency greater than system efficiency 2: CARNOT is more than or equal to f 5 →g 9 (x)≤0;
Constraint 10, the effective temperature of the fluid is within the range: DBTMXE ≧ TMAXE → g 10 (x)≤0;
Wherein DELTLM is the logarithmic principal temperature difference; FLOWR is the mass flow of the Rankine cycle; FRMXR is the upper limit of Rankine cycle mass flow; TEMP4 is the temperature at which the pressure drops through the turbine; TEMP3 is the temperature at which the temperature across the heat exchanger increases; QUAL4 is the water flow quality through the turbine pressure drop; TMINE is heat exchanger minimum temperature; TMEP2 is the temperature at which the pump pressure increases; TDELC is the minimum temperature difference between the lowest temperature of the heat exchanger and the temperature at which the pump pressure is increased; CARNOT is Carnot cycle efficiency; DBTMXE is a heat exchanger; g is a radical of formula 1 (x)-g 10 (x) Is a system constraint; f. of 4 System efficiency 1 (%); f. of 5 The system efficiency was 2 (%).
2. The multi-objective adaptive cluster optimization method according to claim 1, wherein the generating of the objective correlation property matrix according to the prioritized N objectives specifically comprises:
step S21: initializing, and setting the fraction H of the target division to be 1;
step S22: calculating a weight set according to the fraction of the target division by adopting a weight generation algorithm, and giving an Archimedes deviation function;
step S23: solving the Archimedes deviation function endowed with the weight according to a self-adaptive linear algorithm to obtain an optimized solution of each target under each weight group;
step S24: calculating a target deviation matrix according to the optimal solution and a target expected value;
step S25: standardizing the target deviation matrix to obtain a standardized deviation matrix;
step S26: performing correlation calculation according to the standardized deviation matrix to obtain a target correlation characteristic matrix;
step S27: calculating the standard deviation of the target correlation characteristic matrix, and outputting the target correlation characteristic matrix when the difference value of the two adjacent standard deviations is within a set error range; when the difference between the two adjacent standard deviations is not within the set error range, the number of copies H of the target division is increased, and the process returns to step S22.
3. The multi-objective adaptive cluster optimization method according to claim 1, wherein the target clustering analysis is performed according to the target correlation characteristic matrix by using a hierarchical clustering method to obtain a final target clustering result, and the method specifically comprises:
step S31: performing target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a plurality of groups of target clustering results;
step S32: judging whether each target priority is traversed or not; if not, returning to the step S13; if the target clustering result is traversed, grading the multiple groups of target clustering results to obtain target clustering results of different grading conditions, and executing the step S33;
step S33: judging whether a convergence condition is reached; if the convergence condition is reached, taking the target clustering results of different current grading conditions as final target clustering results; and if the convergence condition is not met, taking the target clustering result with the largest occurrence frequency in the target clustering results of different current grading conditions as the last target clustering result, and returning to the step S13.
4. The multi-objective adaptive cluster optimization method according to claim 2, wherein the target deviation matrix is calculated according to the optimization solution and a target expected value; the target deviation matrix is D d =[d 1 ,…,d N ];
5. A multi-objective adaptive cluster optimization system, the system comprising:
the priority grading module is used for carrying out priority grading on the N targets according to different clusters;
the target correlation characteristic matrix generation module is used for generating a target correlation characteristic matrix according to the N targets with the divided priorities;
the target clustering analysis module is used for carrying out target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a final target clustering result;
the priority ranking module specifically includes:
the multi-target description problem construction unit is used for constructing a multi-target description problem based on multiple targets;
the compromise decision problem construction unit is used for constructing a compromise decision problem according to the multi-target description problem;
the priority grading unit is used for carrying out priority grading on the N targets according to different clusters by utilizing a priority deviation function in the compromise decision problem according to the last target clustering result;
the multi-target description problem comprises four decision variables, one linear constraint, ten nonlinear inequality constraints and six targets;
four decision variables:
x 1 : maximum pressure in rankine cycle variables;
x 2 : rankine cycle minimum pressure;
x 3 : rankine cycle maximum temperature;
x 4 : the maximum temperature of the heating liquid in the exchanger;
six objective functions:
object 1, minimizing steam of a turbomachine:minf 1 (x 1 ,x 2 ,x 3 ,x 4 );
Goal 2, maximizing rankine cycle efficiency: maxf 2 (x 1 ,x 2 ,x 3 ,x 4 );
Goal 3, maximizing heat exchanger efficiency: maxf 3 (x 1 ,x 2 ,x 3 ,x 4 );
Goal 4, maximize system efficiency index 1: maxf 4 (x 1 ,x 2 ,x 3 ,x 4 );
Goal 5, maximize system efficiency index 2: maxf 5 (x 1 ,x 2 ,x 3 ,x 4 );
Goal 6, maximizing heat transfer efficiency of the heat exchanger: maxf 6 (x 1 ,x 2 ,x 3 ,x 4 );
Boundary conditions:
500≤x 1 ≤5000(kPa);
350≤x 2 ≤850(K);
350≤x 3 ≤850(K);
ten constraint functions:
constraint 1, upper temperature increase limit: TMAXE-TMAX ≥ DELTLM → g 1 (x)≤0;
Constraint 2, moisture in the turbine is less than upper limit: RCMIT ≦ TMXL → g 2 (x)≤0;
Constraint 3, rankine cycle mass flow less than upper limit: FLOWR ≦ FRMXR → g 3 (x)≤0;
Constraint 4, temperature between heat exchanger and turbine is greater than or equal to heat exchanger: TEMP4 ≧ TEMP3 → g 4 (x)≤0;
Constraint 5, the mass between the heat exchanger and the turbine is superheated steam: QUAL4 is more than or equal to 1.0 → g5 (x) is less than or equal to 0;
constraint 6, heating temperature differential limit in exchanger: TMAXE-TMINE ≧ TDELE → g 6 (x)≤0;
Constraint 7, heating liquid minimum temperature limit in exchanger: TMINE-TMEP 2. Gtoreq.TDELC → g 7 (x)≤0;
Constraint 8, ideal carnot cycle efficiency greater than system efficiency 1: CARNOT is more than or equal to f 4 →g 8 (x)≤0;
Constraint 9, ideal carnot cycle efficiency greater than system efficiency 2: CARNOT is more than or equal to f 5 →g 9 (x)≤0;
Constraint 10, the effective temperature of the fluid is within the range: DBTMXE ≧ TMAXE → g 10 (x)≤0;
Wherein DELTLM is the logarithmic principal temperature difference; FLOWR is the mass flow rate of the Rankine cycle; FRMXR is the upper limit of Rankine cycle mass flow; TEMP4 is the temperature at which the pressure drops through the turbine; TEMP3 is the temperature at which the temperature across the heat exchanger increases; QUAL4 is the water flow quality through the turbine pressure drop; TMINE is the heat exchanger minimum temperature; TMEP2 is the temperature at which the pump pressure increases; TDELC is the minimum temperature difference between the lowest temperature of the heat exchanger and the temperature at which the pump pressure is increased; CARNOT is Carnot cycle efficiency; DBTMXE is a heat exchanger; g 1 (x)-g 10 (x) Is a system constraint; f. of 4 1 (%) for system efficiency; f. of 5 The system efficiency was 2 (%).
6. The multi-objective adaptive cluster optimization system according to claim 5, wherein the objective correlation characteristic matrix generation module specifically comprises:
the initialization unit is used for initializing and setting the fraction H of the target division to be 1;
the assignment unit is used for calculating a weight set according to the fraction of the target division by adopting a weight generation algorithm and endowing an Archimedes deviation function;
the solving unit is used for solving the Archimedes deviation function endowed with the weight according to a self-adaptive linear algorithm to obtain an optimized solution of each target under each weight group;
the target deviation matrix calculation unit is used for calculating a target deviation matrix according to the optimal solution and a target expected value;
the standardization processing unit is used for standardizing the target deviation matrix to obtain a standardized deviation matrix;
the target correlation characteristic matrix calculation unit is used for carrying out correlation calculation according to the standardized deviation matrix to obtain a target correlation characteristic matrix;
the first judgment unit is used for calculating the standard deviation of the target correlation characteristic matrix, and outputting the target correlation characteristic matrix when the difference value of the two adjacent standard deviations is within a set error range; and when the difference value of the standard deviations of two adjacent times is not in the set error range, increasing the number of copies H of the target partition, and returning to the assignment unit.
7. The multi-objective adaptive cluster optimization system according to claim 5, wherein the objective cluster analysis module specifically comprises:
the target clustering analysis unit is used for carrying out target clustering analysis according to the target correlation characteristic matrix by adopting a hierarchical clustering method to obtain a plurality of groups of target clustering results;
the second judgment unit is used for judging whether each target priority is traversed or not; if not, returning to the priority level unit; if the target clustering results are traversed, grading the multiple groups of target clustering results to obtain target clustering results of different grading conditions at present, and executing a third judgment unit;
a third judging unit for judging whether a convergence condition is reached; if the convergence condition is reached, taking the target clustering results of different current grading conditions as final target clustering results; and if the convergence condition is not met, taking the target clustering result with the largest occurrence frequency in the target clustering results of different current grading conditions as the last target clustering result, and returning to the priority grading unit.
8. The multi-objective adaptive cluster optimization system of claim 6, wherein the target deviation matrix is calculated from the optimization solution and a target expectation; the target deviation matrix is D d =[d 1 ,…,d N ];
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110207836.9A CN112948997B (en) | 2021-02-24 | 2021-02-24 | Multi-objective adaptive clustering optimization method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110207836.9A CN112948997B (en) | 2021-02-24 | 2021-02-24 | Multi-objective adaptive clustering optimization method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112948997A CN112948997A (en) | 2021-06-11 |
CN112948997B true CN112948997B (en) | 2022-11-08 |
Family
ID=76246020
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110207836.9A Active CN112948997B (en) | 2021-02-24 | 2021-02-24 | Multi-objective adaptive clustering optimization method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112948997B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117811226A (en) * | 2023-11-16 | 2024-04-02 | 中国科学院空天信息创新研究院 | Method, device and medium for identifying characteristics of coupling mechanism of wireless power transmission system |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6850891B1 (en) * | 1999-07-23 | 2005-02-01 | Ernest H. Forman | Method and system of converting data and judgements to values or priorities |
CN107194609A (en) * | 2017-06-15 | 2017-09-22 | 北京理工大学 | A kind of product design system and method |
CN109185018A (en) * | 2018-10-25 | 2019-01-11 | 湖南大学 | A kind of frequency conversion permanent magnet hydroelectric power system and its control method |
CN110659723A (en) * | 2019-09-03 | 2020-01-07 | 腾讯科技(深圳)有限公司 | Data processing method, device, medium and electronic equipment based on artificial intelligence |
-
2021
- 2021-02-24 CN CN202110207836.9A patent/CN112948997B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6850891B1 (en) * | 1999-07-23 | 2005-02-01 | Ernest H. Forman | Method and system of converting data and judgements to values or priorities |
CN107194609A (en) * | 2017-06-15 | 2017-09-22 | 北京理工大学 | A kind of product design system and method |
CN109185018A (en) * | 2018-10-25 | 2019-01-11 | 湖南大学 | A kind of frequency conversion permanent magnet hydroelectric power system and its control method |
CN110659723A (en) * | 2019-09-03 | 2020-01-07 | 腾讯科技(深圳)有限公司 | Data processing method, device, medium and electronic equipment based on artificial intelligence |
Non-Patent Citations (1)
Title |
---|
基于模糊聚类的机载多功能毫米波雷达动态资源调度;陈杰等;《火控雷达技术》;20110625(第02期);8-13+44 * |
Also Published As
Publication number | Publication date |
---|---|
CN112948997A (en) | 2021-06-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Bakirtzis et al. | Genetic algorithm solution to the economic dispatch problem | |
Larsen et al. | Definitions of generalized multi-performance weighted multi-state K-out-of-n system and its reliability evaluations | |
Kou et al. | Many-objective optimization for coordinated operation of integrated electricity and gas network | |
CN112948997B (en) | Multi-objective adaptive clustering optimization method and system | |
CN111369000A (en) | High-dimensional multi-target evolution method based on decomposition | |
CN114841055A (en) | Unmanned aerial vehicle cluster task pre-distribution method based on generation of countermeasure network | |
CN109066651A (en) | The calculation method of wind-powered electricity generation-load scenarios limit transmitted power | |
Yin et al. | A Flexibility-oriented robust transmission expansion planning approach under high renewable energy resource penetration | |
Ferrara et al. | Optimization algorithms supporting the cost optimal analysis: the behavior of PSO | |
Wright et al. | Evolutionary synthesis of HVAC system configurations: algorithm development (RP-1049) | |
Khandelwal et al. | Security constrained transmission network expansion planning using grey wolf optimization algorithm | |
CN102682336B (en) | Method for optimizing design on number of regulating stage nozzles of steam turbine based on improved genetic algorithm | |
CN112906250B (en) | Complex system module classification method | |
Roy et al. | Designing a turbine blade cooling system using a generalised regression genetic algorithm | |
Liao et al. | Multiobjective design of equivalent accelerated life testing plans | |
Mousa et al. | Hybrid multiobjective evolutionary algorithm based technique for economic emission load dispatch optimization problem | |
Shukla et al. | Voltage stability improvement of distribution networks by enhancing DGs impacts | |
Lotfi et al. | Robust multi-objective optimization of state feedback controllers for heat exchanger system with probabilistic uncertainty | |
Tabassum et al. | Modeling and optimizing oxygen production system using differential evolution. | |
Lvovich et al. | Development of a simulation model for the supply of spare parts for heating equipment | |
Hussein et al. | A new strategy to synthesis an optimum controllable HEN by using fuzzy analogical gates | |
Rao et al. | ABC and ANN based minimization of power loss for distribution system stability | |
Létal et al. | Nonlinear finite element analysis-based flow distribution model for engineering practice | |
CN113722989A (en) | CPS-DP model-based aircraft engine life prediction method | |
Lamamra et al. | Optimization of the structure of a neural networks by multi-objective genetic algorithms |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |