CN115293056A - Modelica model-oriented multi-objective optimization algorithm - Google Patents
Modelica model-oriented multi-objective optimization algorithm Download PDFInfo
- Publication number
- CN115293056A CN115293056A CN202211230801.8A CN202211230801A CN115293056A CN 115293056 A CN115293056 A CN 115293056A CN 202211230801 A CN202211230801 A CN 202211230801A CN 115293056 A CN115293056 A CN 115293056A
- Authority
- CN
- China
- Prior art keywords
- population
- individuals
- parameter
- parameters
- fitness
- 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.)
- Granted
Links
Images
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Physiology (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Genetics & Genomics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention discloses a Modelica model-oriented multi-objective optimization algorithm, which comprises the following steps: acquiring various input parameters and various output parameters to be optimized in a Modelica model; generating an initial population formed by combining N values of each input parameter; substituting various group individuals in the current group into the Modelica model for simulation; executing a rapid non-dominated sorting algorithm on the current population to obtain a first non-dominated level and a first crowding degree of each population of individuals, and calculating a first fitness of each population of individuals; generating a new current population and N new value combinations of corresponding output parameters according to the first fitness and the priority of each input parameter, and returning to the calculation operation of the first non-dominated level and the first congestion degree until a set optimization termination condition is reached; and selecting at least one optimal value combination of each input parameter in the final current population. The embodiment is used for multi-objective simultaneous optimization.
Description
Technical Field
The embodiment of the invention relates to the technical field of multi-field system simulation optimization, in particular to a Modelica model-oriented multi-objective optimization algorithm.
Background
Modelica is an open equation-based language, is suitable for modeling large-scale complex heterogeneous physical systems, comprises mechanical, electronic, electric, hydraulic, thermal, control and process-oriented subsystem models, and is widely accepted in the simulation industry at present.
However, in the model simulation, especially in the large complex model, how to quickly obtain the expected target parameter value (i.e. output parameter value) is a difficult problem which puzzles the model builder.
The existing simulation software usually supports single target optimization (namely, optimization of a single output parameter), but in the practical application of the Modelica model, multiple targets (namely, multiple output parameters) all achieve a Pareto optimal solution to meet the reality; even if multiple single-target optimizations are used, there may be parameter conflict or target conflict issues. In some multi-objective optimization algorithms, the Modelica model needs to be modified, which is very limited in practical use.
Disclosure of Invention
The embodiment of the invention provides a Modelica model-oriented multi-objective optimization algorithm, which realizes multi-objective simultaneous optimization under the condition of not changing a model structure.
In a first aspect, an embodiment of the present invention provides a multiobjective optimization algorithm for a Modelica model, including:
acquiring input parameters to be optimized in a Modelica model and output parameters for evaluating the performance of the model;
generating an initial population formed by N value combinations of each input parameter as a current population, wherein each value combination is a population individual, and N is a natural number;
substituting various group individuals in the current group into the Modelica model for simulation to obtain N value combinations of various output parameters;
according to the N value combinations of the output parameters, a rapid non-domination sorting algorithm is executed on the current population to obtain a first non-domination level and a first crowding degree of each population individual;
calculating first fitness of each population of individuals according to the first non-dominated level and the first crowding degree;
generating a new current population and N new value combinations of corresponding output parameters according to the first fitness and the priority of each input parameter, and returning to the calculation operation of the first non-dominated level and the first crowding degree until a set optimization termination condition is reached;
and selecting at least one population individual from the final current population as at least one optimal value combination of each input parameter.
In a second aspect, an embodiment of the present invention further provides an electronic device, where the electronic device includes:
one or more processors;
a memory for storing one or more programs,
when the one or more programs are executed by the one or more processors, the one or more processors implement a Modelica model-oriented multi-objective optimization algorithm according to any embodiment.
In a third aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements a Modelica model-oriented multi-objective optimization algorithm according to any embodiment.
The embodiment of the invention provides a Modelica model-oriented multi-objective optimization algorithm, which realizes simultaneous optimization of multiple output parameters. Compared with the traditional Modelica model target optimization algorithm, the optimization effect can be achieved by selecting parameters, setting constraint conditions and the like without modifying the model. In the specific parameter optimization process, the fitness function is constructed through the non-dominated level and the crowdedness degree and is directly applied to the genetic algorithm, an additional fitness function does not need to be constructed, and the calculation efficiency is greatly improved. Meanwhile, the priorities of the input parameters and the output parameters are respectively set, so that the rapid convergence of the population towards the required direction is promoted.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of a multiobjective optimization algorithm oriented to a Modelica model according to an embodiment of the present invention.
Fig. 2 is a selection interface of a Modelica model according to an embodiment of the present invention.
Fig. 3 is a selection interface for inputting parameters according to an embodiment of the present invention.
Fig. 4 is a setting interface for inputting parameters according to an embodiment of the present invention.
Fig. 5 is a setting interface of an output parameter according to an embodiment of the present invention.
Fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described below. It is to be understood that the disclosed embodiments are merely exemplary of the invention, and are not intended to be exhaustive or exhaustive. 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.
In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, and are only for convenience of description and simplification of description, but do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
In the description of the present invention, it should also be noted that, unless otherwise explicitly specified or limited, the terms "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood in a specific case to those of ordinary skill in the art.
Fig. 1 is a flowchart of a Modelica model-oriented multi-objective optimization algorithm according to an embodiment of the present invention. The method is suitable for optimizing the input parameter value of a specific Modelica model so as to achieve the condition of expected output parameters and is executed by electronic equipment. As shown in fig. 1, the method specifically includes:
s110, obtaining input parameters to be optimized in the Modelica model and output parameters for evaluating the performance of the model.
Specifically, in the field of automobiles, the modecia model may be large or small, and may be a modeica model of one component, or may be a system model composed of modeica models of a plurality of components, and the present embodiment is not limited thereto. The input parameters and the output parameters refer to specific parameter names, not parameter values. For example, for a model ica model of a tire, the input parameters to be optimized include tire thickness, radius, etc., and the output parameters for evaluating the performance of the tire model include tire pressure, gravity, etc. In this embodiment, the values of the input parameters are optimized, so that the values of the output parameters meet the performance requirements.
Optionally, the Modelica model, the input parameters and the output parameters are obtained through a visual user interface. Fig. 2 is a selection interface of a Modelica model according to an embodiment of the present invention. And responding to the selection operation of the user on the interface, calling the simulation model needing to be optimized from the model library, and setting information such as simulation starting time, stopping time, simulation step length and the like. After the Modelica model is selected, the selected Modelica model is analyzed by ANTLR4, boClass of the model is obtained (an entity class structure is formed according to a Modelica grammar structure), then BoVariable S of all hierarchies is obtained step by step in a recursive manner, each BoVariable is converted into a TreeList style node to form a tree structure, and the tree structure is displayed in an interface shown in FIG. 3. Each node comprises a variable name, a variable initial value, a variable unit, variable description and the like. Referring to FIG. 3, PID _controllerrepresents a module in the Modelica model, driveAngle, PI, etc. represent the primary variables in PID _ Controller, xi _ start, xd _ start, etc. represent the secondary variables, and so on. The tree structure display method is convenient for users to clearly know all variables and the dependency relationship among the variables of the model, and is convenient for developers to display only input parameters on an optimized parameter interface. And determining parameters needing to be optimized in response to the operation of checking one or more variables on the tree structure by the user. And after the input parameters are selected, acquiring target parameters for evaluating the quality degree of the model. And similarly, establishing a target parameter tree structure which can be selected by a user, and determining target parameters after selection. The analysis logic is the same as the input parameters, and the difference is that only the output parameters are displayed on the selection interface.
And S120, generating an initial population formed by N value combinations of the input parameters as a current population. Wherein each value combination is a population individual, and N is a natural number.
Specifically, various values of each input parameter are determined according to the value range of each input parameter. And different values are arranged and combined to obtain a plurality of value combinations. For example, one combination of tire thickness =10cm and radius =20cm, and another combination of tire thickness =15cm and radius =25 cm. And taking each value combination as a population individual to form an initial population as the input of a genetic algorithm.
Optionally, the value range of each input parameter is set by a user. FIG. 4 is a setting interface for input parameters presented to a user. And responding to the setting operation of a user, and acquiring the value range, data precision, priority, display sequencing and the like of each input parameter to be optimized. The value range refers to the value range of the input parameters to be optimized in the optimization algorithm, and comprises a minimum value and a maximum value. Data precision refers to the maximum decimal place of the input parameter value. The priority can be embodied in the form of weight, which refers to the importance degree among input parameters, and the more important parameters, the higher the priority. The display sorting is set according to the quality standard of the input parameters, the larger the more excellent the input parameters can be set to be in a descending order, and the smaller the more excellent the input parameters can be set to be in an ascending order; the parameter values listed above will be displayed preferentially in the final presentation interface so that the user can select these values preferentially.
In a specific embodiment, the values of the input parameters are stored in the population individuals in a binary code form, and the generation of the initial population includes the following steps:
step one, determining the binary coding length of each input parameter according to the value range and the precision of each input parameter. In genetic algorithms, a chromosome (i.e., a population of individuals) is represented by a binary sequence, and thus input parameters need to be binary-coded. For any input parameter, firstly, calculating the binary coding length required by the parameter according to the data precision (namely the number of bits after the decimal point) required by the parameter; then, considering the equality among the parameters, adding normalization operation into the calculation formula of each parameter to obtain a final calculation formula: ceiling (math.log ((div.max-div.min) × math.pow (10, div.decilmalnum) + 1, 2)). Wherein, dist represents the input parameter, dist.max represents the maximum value of the input parameter, dist.min represents the minimum value of the input parameter, and dist.decilmalnum represents the data precision of the input parameter; (disv.max-disv.min)' math.pow (10, disv.decilmalnum) represents the range of the entire solution space of the input parameters; log ((div.max-div.min) × math.pow (10, div.decilmalnum) + 1, 2) the binary code of how many bits to solve can represent the whole solution space range; ceiling () represents rounding up because the coding length of the binary must be an integer. Because the value range and the data precision of each input parameter are different, the binary code length corresponding to each input parameter is different.
And step two, generating a Harrington sequence consisting of N decimal parts according to a low-difference initial strategy. In order to avoid the situation that local optimization is involved in the Modelica model simulation process, the step ensures that the distribution of the initial population is as uniform as possible to cover the whole situation. Specifically, a low-variance sequence technique is adopted to create a Harton serialization, which consists of a plurality of decimal fractions x uniformly distributed in the range of 0-1 1 ,... x N Is composed of, and x 1 ,... x N Neither excessively dense nor excessively loose in any sub-region.
And step three, solving the values of all input parameters corresponding to any decimal, and converting all the values into binary codes with corresponding lengths. Each decimal in the harrington sequence corresponds to a value combination of each input parameter. For example, assuming that the tire width W is in the range of [10,19] and the radius R is in the range of [20,29], the decimal 0.5 corresponds to W =15, R =25, the decimal 0.1 corresponds to W =11, R =21, and the like. And after the value combination of the input parameters is obtained, converting the values of the parameters into binary codes with corresponding lengths solved in the first step. For example, assuming that the binary code length corresponding to the tire width is L1 and the binary code length corresponding to the radius is L2, W =15 is converted into the binary code B1 having the length L1, and R =25 is converted into the binary code B2 having the length L2. It should be noted that the parameter value range and the corresponding relationship between the decimal and the parameter value are only used to illustrate the generation process of the binary code, and do not represent the actual value.
And step four, cascading the binary codes according to the priority of each input parameter to form a population individual corresponding to the decimal. Specifically, the binary code of the input parameter with high priority is arranged in front, and the binary code of the input parameter with low priority is arranged in back. Assuming that the priority of the tire width is higher than the tire radius, B1 is arranged in front of B2, and population units with decimal 0.5 are formed by cascading B1 and B2.
Because the genetic algorithm usually finds the local optimal solution near the individuals of the initial population, compared with a pseudorandom sequence, the initial population generated according to the Hardgon sequence has higher convergence efficiency, and meanwhile, the population individuals can be ensured to meet uniform distribution under different population sizes N, thereby being beneficial to avoiding falling into the local optimal solution to finish the training cycle in advance.
And S130, substituting various group individuals in the current group into the Modelica model for simulation to obtain N value combinations of each output parameter. Each population individual corresponds to one value combination of each output parameter.
The values of the input parameters in each population individual and basic simulation information (shown as the setting in fig. 2) are input into a Modelica model together, and one-time simulation is executed to obtain the values of the output parameters output by the current simulation, so that a value combination is formed.
It is worth mentioning that the implementation of the step depends on the existing Modelica simulation software, but the operation sequence inside the Modelica kernel is optimized instead of simply calling the outermost interface operated by the Modelica model directly. Specifically, in the whole genetic algorithm, hundreds of thousands of times of simulation (corresponding to hundreds of thousands of population individuals) are involved, and if the outermost layer interface operated by the Modelica model is directly called, the steps of opening the model, loading the class library, checking the model, compiling the model, setting parameters, calculating the simulation, unloading the class library, ending the simulation and the like are sequentially executed in each simulation, so that the program is complicated and the consumed time is long. Therefore, as the same model is involved in hundreds of times of simulation operation, the steps of class library loading, model checking, compiling and the like of the Modelica model are executed only before the first simulation operation, then a solver of the Modelica model is called for many times to execute the simulation operation, each simulation operation takes the value of each input parameter in a group of individuals as input, and takes a group of values of each output parameter as output; and after the multiple times of simulation operation are completed, unloading the class library of the Modelica model. Therefore, the simulation time can be greatly shortened, and the optimization efficiency is improved. Further, the process is realized through secondary development of Modelica simulation software, for example, a stopping program is added between a compiler and a solver, and a multi-cycle body or a single inlet is added on the outer layer of the solver after stopping.
And S140, according to the N value combinations of the output parameters, executing a rapid non-dominance sorting algorithm on the current population to obtain a first non-dominance level and a first crowding degree of each population individual.
And evaluating the advantages and disadvantages of various group individuals by a quick non-dominated sorting algorithm according to the non-dominated level and the crowding degree of the group individuals. Wherein, the calculation of the non-dominant hierarchy comprises the following steps:
step one, determining an evaluation function corresponding to each output parameter according to the constraint condition of each output parameter. Specifically, after the output parameters are obtained in S110, the output parameters may also be presented to the user through the setting interface in fig. 5; in response to a setting operation by a user, an evaluation value, a constraint condition, an expected value, a tolerance, a priority, and the like of each output parameter are acquired. Optionally, the priority of the output parameters is also embodied in the form of weight (as shown in fig. 5), which represents the importance degree between the output parameters. It should be noted that the input parameters and the output parameters are not in one-to-one correspondence, and a plurality of input parameters may affect a plurality of output parameters; therefore, the weights of the input parameters and the weights of the output parameters are independent of each other, and are not necessarily related. The estimation refers to which time or which operation output parameter value is taken from a time sequence obtained by single simulation, and the system supports 5 estimation methods, namely Maximum (taking the Maximum value in a simulation interval), minimum (taking the Minimum value in the simulation interval), average (taking the Average value in the simulation interval), initialValue (taking the value at the start time of simulation) and FinalValue (taking the value at the stop time of simulation). The constraint condition refers to a measure of the simulation result, and the system supports 5 constraint conditions, namely Maximize (hope for the output parameter to be maximized), minimize (hope for the output parameter to be minimized), equal to (hope for the output parameter to be equal to the expected value), greaterthann (hope for the output parameter to be greater than or equal to the expected value), and LessThen (hope for the output parameter to be less than or equal to the expected value). The expected value depends on the constraint and needs to be set only if the constraint is equal to, greaterThan, lessThen. Tolerance refers to the convergence error in the iteration of the optimization algorithm, and when the relative error between the output parameter value and the expected value is within the tolerance range, the corresponding input parameter value is reserved only when the optimization goal is achieved.
And step two, determining a corresponding evaluation function according to the constraint condition of the output parameter. Optionally, when the constraint condition is Maximize, the corresponding evaluation function is the normalized output parameter (the normalized maximum value is 1). When the constraint condition is Minimize, the corresponding evaluation function is: the normalized output parameter is subtracted from 1, or 1 is divided by the normalized output parameter. When the constraint condition is Equalto, the corresponding evaluation function satisfies the following conditions: the value of the output parameter is subtracted from the desired value, the smaller the absolute value of the difference, the better. When the constraint condition is greaterthann, the corresponding evaluation function satisfies: the value of the output parameter is differed from the expected value, and the individual difference value larger than 0 is better than the individual difference value smaller than 0; for individuals with a difference greater than 0, the smaller the difference, the better; for individuals with a difference less than 0, the larger the difference the better. When the constraint condition is LessThen, the corresponding evaluation function satisfies the following conditions: the value of the output parameter is differed from the expected value, and the individual difference value smaller than 0 is better than the individual difference value larger than 0; for individuals with differences greater than 0, the smaller the difference, the better; for individuals whose difference values are all less than 0, the larger the difference value, the better.
And step three, calculating the dominance factor of each population individual in the current population according to the value of each output parameter and the corresponding evaluation function. In the fast non-dominated sorting algorithm, the dominating factor is used to judge the quality of an individual. Specifically, in the present embodiment, a corresponding dominance factor algorithm is automatically generated in response to the setting of the output parameters and the constraint conditions by the user. For example, each time an output parameter is added, the evaluation function corresponding to the output parameter is added to the polynomial for calculating the dominance factor; every time one output parameter is reduced, the evaluation function corresponding to the output parameter is deleted from the polynomial for calculating the dominance factor.
And step four, according to the dominance factors of various groups of individuals, executing a rapid non-dominance ordering algorithm on the individual groups of individuals to obtain first non-dominance levels of the various groups of individuals. Specifically, according to a rapid non-domination sorting method, domination relations among individuals are compared according to domination factors, then populations are layered according to the domination relations and domination sizes, and non-domination hierarchies of various populations and individuals are obtained. The lower the non-dominant level, the better the population of individuals.
After obtaining the non-domination levels of various groups of individuals, the calculation of the crowding degree comprises the following steps:
and step one, calculating the crowdedness of various groups of individuals in the same non-dominant level under each evaluation function according to the evaluation function corresponding to each output parameter. The crowding degree is used for comparing population individuals with the same non-dominant level. The higher the crowding degree is, the better the population individuals are, so that the individuals in the quasi Pareto domain can be expanded to the whole Pareto domain and uniformly distributed, and the diversity of the population is kept. The congestion operator is actually a function of the calculated congestion distance. In this step, population individuals are sorted based on an evaluation function of a certain output parameter, and then the congestion degrees of two (maximum and minimum) individuals at a boundary are set to infinity, the congestion degree I [ I ] of the ith individual is calculated by the formula I [ I ] = I [ I ] + (Im [ I + 1] -Im [ I-1 ])/(fmax-fmin), where Im [ ] represents the evaluation function corresponding to the output parameter, im [ I + 1] and Im [ I-1] respectively represent evaluation function values corresponding to (I + 1) and (I-1) th populations, and fmax and fmin represent the maximum evaluation function value and the minimum evaluation function value in the populations, and normalization processing is performed for next comparison. Finally, each population individual corresponds to one crowding degree under each output parameter.
And step two, fusing the crowdedness degrees of the individuals in the same population according to the priority of each output parameter to obtain a first crowdedness degree of the individuals in the population. Specifically, for any population of individuals, the crowdedness of the individual in all output parameters is weighted and accumulated according to the weight of the output parameters, and the first crowdedness of the population of individuals is obtained. The step emphasizes more intentional optimization targets (namely, output parameters) through the weight of the output parameters, and is particularly suitable for the condition that the quality trends of the output parameters are inconsistent, namely when one output parameter is more optimal, the other output parameter is deteriorated. In addition, the priority allocation effect is only effective among population individuals in the same non-dominant level, the individual advantages and disadvantages determined by the non-dominant level are fully respected, and the interference caused by excessive human factors is avoided.
S150, calculating first fitness of each population of individuals according to the first non-dominated level and the first crowding degree.
In the embodiment, for the non-dominant hierarchy and the crowding degree, an adaptive fitness function is constructed for quantitatively representing the quality degree of population individuals. Specifically, through fast non-dominated sorting and congestion degree calculation, each population individual obtains two attributes: non-dominant hierarchy rank n and congestion degree d _ n. By using these two attributes, the dominant relationship between any two individuals in the population can be distinguished. Individual i dominates individual j if and only if rank _ i < rank _ j, or rank _ i = rank _ j and d _ i > d _ j, i.e. individual i is superior to individual j. Therefore, according to the maximum value of all non-dominant levels and the maximum value of all crowdedness, a fitness function of each population individual is constructed:
F=(rank_max-rank_n)×d_max+d_n
wherein F represents fitness, rank _ manx represents the maximum value of non-dominant levels of all population individuals, d _ max represents the maximum value of congestion of all population individuals, rank _ n represents the non-dominant levels of the current population individuals, and d _ n represents the congestion of the current population individuals.
Then, the rank _ n and d _ n of each population of individuals are substituted into the fitness function, and the fitness F of each population of individuals is calculated. Through the fitness function, two indexes (non-dominated levels and crowding degree) for evaluating the quality of the population individuals are ingeniously fused into one index (fitness), so that storage space and computing resources are greatly saved no matter the data structure is simplified or the following optimal individual group is searched, and the efficiency of the whole optimization algorithm is improved.
And S160, generating a new current population and N new value combinations of corresponding output parameters according to the first fitness and the priority of each input parameter, and returning to the calculation operation of the first non-dominated level and the first crowding degree until a set optimization termination condition is reached.
The embodiment continuously optimizes the value combination of each input parameter based on the genetic algorithm, and greatly improves the efficiency of population iteration by using the fitness function constructed in S150 and the priority of the input parameters in the population iteration. Specifically, S160 includes the following steps:
the method comprises the steps of firstly, selecting M population individuals from a current population according to first fitness, and generating a filial generation population based on the M population individuals, wherein M is less than or equal to N. Alternatively, first, M population individuals are selected from N population individuals using a roulette selection method, and chromosomes having a higher first fitness value are more likely to be selected in roulette. Specifically, in the roulette selecting method, the fitness of all population individuals in the population calculated in the previous step is summarized to obtain the total fitness, then the probability of each population being selected is calculated by dividing the fitness of each population by the total fitness, and then the cumulative probability P of each population is calculated i . For example, the cumulative probability of the nth population individual = the sum of the probabilities of the first n population individuals. Finally, N random numbers in the range of 0-1 are generated through roulette selection, and a random number satisfying P is selected from the population M-1 Less than random number less than or equal to P M To generate N new population individuals. It is worth mentioning that since the fitness function is directly constructed by directly utilizing the non-dominated hierarchy and the crowding degree in S150, the result of the rapid non-dominated sorting can be directly applied to the gene selection of the genetic algorithm without constructing an additional fitness function, thereby improving the overall efficiency of the algorithm. Combining and exchanging two random groups of the N population individuals for a certain binary code of the N population individualsThese fragments become new individuals. For each new individual, one or more binary codes are randomly selected, the corresponding codes are inverted (0 is changed into 1,1 is changed into 0), and all finally obtained population individuals form a child population.
And step two, combining the current population and the offspring population, and selecting X population individuals with the optimal input parameters, wherein X is greater than N. Because each population individual is formed by arranging the values of all input parameters according to the priority from high to low, firstly, a union set of the current population and the offspring population is obtained, and then the input parameters arranged at the top in all the population individuals of the union set are read; and sequencing various population individuals according to the input parameters, and selecting X population individuals with the optimal input parameters. If the input parameters of the two population individuals are the same, continuously reading the input parameters which are arranged at the forefront and are not read in the two population individuals, comparing the advantages and the disadvantages of the two population individuals through the input parameters, and circulating until a result of comparing the advantages and the disadvantages is obtained, or the selection of the X population individuals is finished. Therefore, the parameter with the highest priority can be selected preferentially, and the number of population individuals needing to be processed in the subsequent steps is reduced. Particularly, the priority of the input parameters is embodied in the binary arrangement sequence of the population individuals, so that the input parameters can be compared only by comparing binary values bit by bit without a complex addressing traversal process, and the calculation efficiency is further improved. The value of X may be determined according to actual needs, and this embodiment is not limited.
And step three, substituting Y population individuals which are not simulated in the X population individuals into the Modelica model for simulation to obtain Y value combinations of each output parameter. Specifically, the X population individuals are all binary codes, wherein the individuals reserved from the current population are simulated by a Modelica model. For other individuals, decoding the other individuals into corresponding decimal input parameters according to the binary coding method in the S120, substituting the decimal input parameters into the Modelica model for simulation, and obtaining a value combination of each output parameter for each population individual.
And fourthly, according to the value combination of the output parameters corresponding to the X population individuals, re-executing the rapid non-dominated sorting algorithm on the X population individuals to obtain a second non-dominated level and a second crowding degree of various population individuals. The calculation process of the dominance level and the congestion degree is the same as the process described in S140, except that the sort target in S140 is the current population, and the sort target here is the X population individuals selected from the current population and the child population. For convenience of distinction and description, the non-dominated hierarchy and the congestion degree obtained in S140 are referred to as a first non-dominated hierarchy and a first congestion degree, respectively, and the non-dominated hierarchy and the congestion degree obtained here are referred to as a second non-dominated hierarchy and a second congestion degree, respectively.
And step five, calculating second fitness of each population individual according to the second non-dominated level and the second crowding degree. Similarly, the process of calculating the fitness here corresponds to the process described in S150. For the purpose of differentiation, the fitness in S150 is referred to as a first fitness, and the fitness here is referred to as a second fitness.
And step six, selecting N population individuals with the optimal second fitness as a new current population. Optionally, when selecting the individual with the optimal fitness, preferentially selecting the population individual with the better fitness; and preferentially selecting the population individuals with better input parameters for various population individuals with the same fitness. The quality of the input parameters can be set by a user, and corresponds to descending and ascending orders in fig. 3, wherein the descending order is larger and the ascending order is smaller and the ascending order is better.
And step seven, after a new current population is obtained, returning to the operation of S140-S160, and repeating the steps until a set optimization termination condition is reached. Optionally, the optimizing the termination condition includes: the number of cycles to reach the setting, or all the output parameters satisfy the constraint condition, etc., and the embodiment is not particularly limited.
S170, selecting at least one population individual from the final current population as at least one optimal value combination of each input parameter.
Optionally, the value combinations of the input parameter values within the tolerance range are stored in the result set, and are displayed according to the sorting rule. And when the optimal maximum combination number is exceeded, sequencing the data in the result set according to the weight and the tolerance, and only keeping the optimal maximum combination number.
The embodiment provides a Modelica model-oriented multi-objective optimization algorithm, and simultaneous optimization of multiple output parameters is realized. Compared with the traditional Modelica model target optimization algorithm, the optimization effect can be achieved by selecting parameters, setting constraint conditions and the like without modifying the model. In a specific parameter optimization process, a fitness function is constructed through the non-dominated hierarchy and the crowding degree and is directly applied to a genetic algorithm, an additional fitness function does not need to be constructed, and the calculation efficiency is greatly improved. Meanwhile, the priorities of the input parameters and the output parameters are respectively set, and the priority of the input parameters influences the cross variation object in the genetic algorithm, so that the input parameters with high priority can be better valued; the calculation of the crowdedness is influenced by the priority of the output parameters, so that the output parameters with high priority are easier to be selected as excellent individuals, and the rapid convergence of the population towards a required direction is promoted.
Fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, as shown in fig. 6, the electronic device includes a processor 60, a memory 61, an input device 62, and an output device 63; the number of processors 60 in the device may be one or more, and one processor 60 is taken as an example in fig. 6; the processor 60, the memory 61, the input device 62 and the output device 63 in the apparatus may be connected by a bus or other means, as exemplified by the bus connection in fig. 6.
The memory 61 is used as a computer readable storage medium for storing software programs, computer executable programs, and modules, such as program instructions/modules corresponding to a multiobjective optimization algorithm oriented to a Modelica model in the embodiments of the present invention. The processor 60 executes various functional applications and data processing of the device by running software programs, instructions and modules stored in the memory 61, namely, a multiobjective optimization algorithm oriented to the Modelica model is realized.
The memory 61 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required for at least one function; the storage data area may store data created according to the use of the terminal, and the like. Further, the memory 61 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some examples, the memory 61 may further include memory located remotely from the processor 60, which may be connected to the device over a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input device 62 may be used to receive entered numeric or character information and to generate key signal inputs relating to user settings and function controls of the apparatus. The output device 63 may include a display device such as a display screen.
Embodiments of the present invention further provide a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements a Modelica model-oriented multi-objective optimization algorithm according to any one of the embodiments.
Computer storage media for embodiments of the invention may employ any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, smalltalk, C + +, or the like, as well as conventional procedural programming languages, such as the C language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the technical solutions of the embodiments of the present invention.
Claims (10)
1. A Modelica model-oriented multi-objective optimization algorithm is characterized by comprising the following steps:
acquiring input parameters to be optimized in a Modelica model and output parameters for evaluating the performance of the model;
generating an initial population formed by N value combinations of each input parameter as a current population, wherein each value combination is a population individual, and N is a natural number;
substituting various group individuals in the current group into the Modelica model for simulation to obtain N value combinations of various output parameters;
according to the N value combinations of the output parameters, a rapid non-domination sorting algorithm is executed on the current population to obtain a first non-domination level and a first crowding degree of each population;
calculating a first fitness of each population of individuals according to the first non-dominated level and the first crowding degree;
generating a new current population and N new value combinations of corresponding output parameters according to the first fitness and the priority of each input parameter, and returning to the calculation operation of the first non-dominated level and the first crowding degree until a set optimization termination condition is reached;
and selecting at least one population individual from the final current population as at least one optimal value combination of each input parameter.
2. The multi-objective optimization algorithm of claim 1, wherein the generating an initial population composed of N combinations of values of each input parameter comprises:
determining the binary coding length of each input parameter according to the value range and the precision of each input parameter;
generating a Harrington sequence consisting of N decimal parts according to a low-difference initial strategy;
solving the values of each input parameter corresponding to any decimal in the Harton sequence, and converting each value into a binary code with corresponding length;
and cascading the binary codes according to the priority of each input parameter to form population individuals corresponding to the decimal.
3. The multi-objective optimization algorithm according to claim 1, wherein the substituting various group individuals in the current group into the Modelica model for simulation to obtain N value combinations of each output parameter includes:
executing class library loading, model checking and compiling of the Modelica model;
calling a solver of the Modelica model for multiple times to execute simulation operation, wherein each simulation operation takes a group of individuals as input and one value combination of each output parameter as output;
and after the multiple times of simulation operation are completed, unloading the class library of the Modelica model.
4. The multi-objective optimization algorithm according to claim 1, wherein the performing a fast non-dominated sorting algorithm on the current population according to the N value combinations of the output parameters to obtain the first non-dominated level and the first crowdedness of the various population individuals comprises:
determining an evaluation function corresponding to each output parameter according to the constraint condition of each output parameter;
calculating a dominance factor of each population individual in the current population according to the value of each output parameter and the corresponding evaluation function;
and according to the dominance factors of various groups of individuals, performing a rapid non-dominance sorting algorithm on the various groups of individuals to obtain first non-dominance hierarchies of the various groups of individuals.
5. The multi-objective optimization algorithm according to claim 1, wherein the performing a fast non-dominated sorting algorithm on the current population according to N value combinations of each output parameter to obtain a first non-dominated level and a first congestion degree of each population individual comprises:
after the first non-dominant level of each group individual is obtained, calculating the crowdedness of each group individual in the same non-dominant level under each evaluation function according to the evaluation function corresponding to each output parameter;
and fusing the crowdedness degrees of individuals in the same population according to the priority of each output parameter to obtain a first crowdedness degree of the individuals in the population.
6. The multi-objective optimization algorithm of claim 1, wherein the calculating a first fitness for each population of individuals based on the first non-dominated hierarchy and a first crowdedness comprises:
constructing a first fitness function of each population individual according to the maximum values of all the first non-dominated levels and the maximum values of all the first crowdedness;
and substituting the first non-dominated level and the first crowding degree of each group of individuals into the first fitness function to calculate the first fitness of each group of individuals.
7. The multi-objective optimization algorithm according to claim 1, wherein the generating a new current population and corresponding N new value combinations of each output parameter according to the first fitness and the priority of each input parameter comprises:
selecting M population individuals from the current population according to the first fitness, and generating a progeny population based on the M population individuals, wherein M is not more than N;
combining the current population and the offspring population, and selecting X population individuals with the optimal input parameters, wherein X is greater than N;
substituting Y population individuals which are not simulated in the X population individuals into the Modelica model for simulation to obtain Y value combinations of each output parameter;
according to the value combination of the output parameters corresponding to the X population individuals, re-executing the rapid non-dominated sorting algorithm on the X population individuals to obtain a second non-dominated level and a second crowding degree of various population individuals;
and calculating second fitness of each population individual according to the second non-dominated level and the second crowding degree, and taking N population individuals with the optimal second fitness as a new population.
8. The multi-objective optimization algorithm according to claim 7, wherein each population individual is formed by arranging values of input parameters in order of priority from high to low;
combining the current population and the offspring population, and selecting X population individuals with optimal input parameters from the current population and the offspring population, wherein the method comprises the following steps:
obtaining a union of the current population and the offspring population;
reading the input parameters arranged at the top in the various groups of the union;
and sequencing various population individuals according to the input parameters, and selecting X population individuals with the optimal input parameters, wherein if the input parameters of the two population individuals are the same, the input parameters which are ranked at the forefront and are not read in the two population individuals are continuously read, the advantages and the disadvantages of the two population individuals are compared through the input parameters, and the operation is repeated in a circulating way until a result of the comparison of the advantages and the disadvantages is obtained, or the selection of the X population individuals is finished.
9. An electronic device, comprising:
one or more processors;
a memory for storing one or more programs,
when executed by the one or more processors, cause the one or more processors to implement a Modelica model-oriented multi-objective optimization algorithm of any one of claims 1-8.
10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, implements a modecia model-oriented multi-objective optimization algorithm according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211230801.8A CN115293056B (en) | 2022-10-10 | 2022-10-10 | Modelica model-oriented multi-objective optimization method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211230801.8A CN115293056B (en) | 2022-10-10 | 2022-10-10 | Modelica model-oriented multi-objective optimization method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115293056A true CN115293056A (en) | 2022-11-04 |
| CN115293056B CN115293056B (en) | 2023-01-10 |
Family
ID=83819414
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211230801.8A Active CN115293056B (en) | 2022-10-10 | 2022-10-10 | Modelica model-oriented multi-objective optimization method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115293056B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117744410A (en) * | 2024-02-19 | 2024-03-22 | 中汽研汽车检验中心（天津）有限公司 | Interval multi-objective optimization method, equipment and medium based on vehicle stability requirement |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102817088A (en) * | 2012-08-09 | 2012-12-12 | 东华大学 | Fusant conveying process of fusant directly-spun terylene filament |
| CN103593986A (en) * | 2013-11-25 | 2014-02-19 | 东南大学 | Main line green wave coordination control signal time method for optimizing exhaust gas emission |
| CN106611287A (en) * | 2016-08-18 | 2017-05-03 | 四川用联信息技术有限公司 | Multi-target resource no-preference optimal configuration algorithm in cloud manufacturing |
| CN109522602A (en) * | 2018-10-18 | 2019-03-26 | 北京航空航天大学 | A kind of Modelica Model Parameter Optimization method based on agent model |
| CN109598092A (en) * | 2018-12-28 | 2019-04-09 | 浙江工业大学 | Merge the air source heat pump multi-objective optimization design of power method of BP neural network and more parent genetic algorithms |
| CN111985730A (en) * | 2020-09-08 | 2020-11-24 | 南方电网数字电网研究院有限公司 | Logistics path planning method and device, computer equipment and storage medium |
| CN113225370A (en) * | 2021-03-08 | 2021-08-06 | 河北工业大学 | Block chain multi-objective optimization method based on Internet of things |
| CN113626976A (en) * | 2021-06-21 | 2021-11-09 | 江苏省镔鑫钢铁集团有限公司 | NSGA-II algorithm-based sintering material multi-objective optimization method, device and equipment |
| CN114243791A (en) * | 2022-01-19 | 2022-03-25 | 中国电力科学研究院有限公司 | Multi-objective optimization configuration method, system and storage medium for wind-solar-hydrogen storage system |
| US11390306B1 (en) * | 2021-08-11 | 2022-07-19 | East China Jiaotong University | Method and system for multi-objective optimization of urban train operation |
-
2022
- 2022-10-10 CN CN202211230801.8A patent/CN115293056B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102817088A (en) * | 2012-08-09 | 2012-12-12 | 东华大学 | Fusant conveying process of fusant directly-spun terylene filament |
| CN103593986A (en) * | 2013-11-25 | 2014-02-19 | 东南大学 | Main line green wave coordination control signal time method for optimizing exhaust gas emission |
| CN106611287A (en) * | 2016-08-18 | 2017-05-03 | 四川用联信息技术有限公司 | Multi-target resource no-preference optimal configuration algorithm in cloud manufacturing |
| CN109522602A (en) * | 2018-10-18 | 2019-03-26 | 北京航空航天大学 | A kind of Modelica Model Parameter Optimization method based on agent model |
| CN109598092A (en) * | 2018-12-28 | 2019-04-09 | 浙江工业大学 | Merge the air source heat pump multi-objective optimization design of power method of BP neural network and more parent genetic algorithms |
| CN111985730A (en) * | 2020-09-08 | 2020-11-24 | 南方电网数字电网研究院有限公司 | Logistics path planning method and device, computer equipment and storage medium |
| CN113225370A (en) * | 2021-03-08 | 2021-08-06 | 河北工业大学 | Block chain multi-objective optimization method based on Internet of things |
| CN113626976A (en) * | 2021-06-21 | 2021-11-09 | 江苏省镔鑫钢铁集团有限公司 | NSGA-II algorithm-based sintering material multi-objective optimization method, device and equipment |
| US11390306B1 (en) * | 2021-08-11 | 2022-07-19 | East China Jiaotong University | Method and system for multi-objective optimization of urban train operation |
| CN114243791A (en) * | 2022-01-19 | 2022-03-25 | 中国电力科学研究院有限公司 | Multi-objective optimization configuration method, system and storage medium for wind-solar-hydrogen storage system |
Non-Patent Citations (3)
| Title |
|---|
| 小菜鸡本菜: "非支配排序遗传算法NSGA-II学习笔记", 《HTTPS://BLOG.CSDN.NET/BESTZY6/ARTICLE/DETAILS/123834350》 * |
| 辜小花等: "基于演化模型偏好多目标优化的智能采油辅助决策支持", 《机械工程学报》 * |
| 陈廷伟等: "面向交通信号优化改进快速非支配排序遗传算法研究", 《计算机应用研究》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117744410A (en) * | 2024-02-19 | 2024-03-22 | 中汽研汽车检验中心（天津）有限公司 | Interval multi-objective optimization method, equipment and medium based on vehicle stability requirement |
| CN117744410B (en) * | 2024-02-19 | 2024-05-07 | 中汽研汽车检验中心（天津）有限公司 | Interval multi-objective optimization method, equipment and medium based on vehicle stability requirement |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115293056B (en) | 2023-01-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113411369B (en) | Cloud service resource collaborative optimization scheduling method, system, medium and equipment | |
| Erbas et al. | Multiobjective optimization and evolutionary algorithms for the application mapping problem in multiprocessor system-on-chip design | |
| US20080307399A1 (en) | Gene expression programming based on hidden markov models | |
| US20190303196A1 (en) | Method and apparatus for automatically scheduling jobs in computer numerical control machines using machine learning approaches | |
| CN109165081B (en) | Web application self-adaptive resource allocation method based on machine learning | |
| CN116629189B (en) | Layout generation method and device of integrated circuit | |
| CN115293056B (en) | Modelica model-oriented multi-objective optimization method | |
| Liu et al. | Configuration space evolutionary algorithm for multi-objective unequal-area facility layout problems with flexible bays | |
| Tao et al. | Construction project scheduling under workspace interference | |
| CN115578023A (en) | Assembly workshop scheduling method, device, equipment and storage medium | |
| CN114461386A (en) | Task allocation method and task allocation device | |
| CN114444383A (en) | Nuclear reactor design scheme optimization method, apparatus, computer device and medium | |
| CN117271101B (en) | Operator fusion method and device, electronic equipment and storage medium | |
| Lin et al. | Design space exploration of memory controller placement in throughput processors with deep learning | |
| CN117060408B (en) | New energy power generation prediction method and system | |
| CN109377111A (en) | Job scheduling method and device based on modified-immune algorithm | |
| CN113220437B (en) | Workflow multi-target scheduling method and device | |
| CN114691370A (en) | Task allocation method and device based on genetic algorithm | |
| CN114969148A (en) | System access amount prediction method, medium and equipment based on deep learning | |
| CN113642226A (en) | Training method of fair machine learning model based on multi-objective evolutionary algorithm | |
| CN114492251A (en) | Low-speed flow field divergence processing method, device, equipment and medium in supercomputing environment | |
| CN110689320A (en) | Large-scale multi-target project scheduling method based on co-evolution algorithm | |
| CN111310993A (en) | Project configuration method, device and system based on multi-objective evolutionary algorithm | |
| CN112988372B (en) | Method and device for determining allocation mode of hardware operation platform | |
| US20240232058A9 (en) | Method, device, and computer program product for generating test case |
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 |