Method, device and medium for designing magnetic components of equipment with dynamic congestion optimization

Abstract

The present invention relates to a design method, device and medium for magnetic components with optimized dynamic congestion degree, including: generating an initial population through random coding based on a constraint range; calculating the core loss and transmitted magnetic energy for each individual in the initial population; performing non-dominated sorting on the initial population according to the calculation results, dividing the individuals into non-dominated levels, calculating the dynamic congestion degree of each individual in the non-dominated levels to obtain a sorting result; based on the sorting result, iteratively updating the initial population, and when the maximum number of iterations is reached, outputting the non-dominated solution set on the Pareto front, setting a weight vector, calculating the weighted Euclidean distance between the non-dominated solution and the ideal point, and selecting the solution with the smallest weighted Euclidean distance as the optimal magnetic component design scheme. By introducing dynamic congestion degree calculation and adaptive sorting strategies, the diversity of population distribution and the high quality of optimization results are realized, so as to provide an optimal magnetic component design scheme applicable to different scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-objective optimization design, and specifically relates to a design method, device and medium for magnetic components of equipment optimized by dynamic crowding degree. Background Technique

[0002] In the fields of electronic devices, transformers and power systems, magnetic components (such as magnetic cores, inductors and transformers) are key components, and their performance directly affects the efficiency, power transmission capacity and energy loss of the equipment. Traditional magnetic component design mainly relies on empirical methods and trial-and-error methods. Design parameters such as temperature, frequency, waveform, peak magnetic flux density and the combination level of magnetic core materials often lack systematicness and optimization, resulting in high magnetic core losses and low efficiency of transmitting magnetic energy, and it is difficult to meet the design requirements of high-efficiency energy transmission and low loss.

[0003] In recent years, with the development of optimization algorithms, designs based on multi-objective optimization methods have gradually been applied to the engineering field. However, although existing multi-objective optimization algorithms (such as the non-dominated sorting genetic algorithm NSGA-II) can effectively handle the conflict relationship between multiple objectives, there are still some deficiencies, such as:

[0004] Limitations in crowding degree calculation: When the crowding distances of multiple individuals are equal, it is impossible to accurately judge the advantages and disadvantages of individuals, which affects the diversity of the population and the optimization effect;

[0005] Uneven population distribution: When solving the Pareto front solution set, the individual distribution may be too concentrated in a certain area, resulting in insufficient coverage of the solution set in the objective space;

[0006] Insufficient convergence: Traditional algorithms are prone to falling into local optima in high-dimensional complex problems, which affects the quality of the Pareto front.

[0007] Therefore, how to construct an optimized design model for magnetic components that can balance the minimization of magnetic core loss and the maximization of transmitted magnetic energy, and improve the existing algorithms to improve the optimization efficiency, solution set distribution uniformity and convergence performance has become an urgent technical problem to be solved. Summary of the Invention

[0008] The present invention provides a design method, device and medium for magnetic components of equipment optimized by dynamic crowding degree, the purpose of which is to construct an optimized design model for magnetic components that can balance the minimization of magnetic core loss and the maximization of transmitted magnetic energy, and improve the existing algorithms to improve the optimization efficiency, solution set distribution uniformity and convergence performance.

[0009] To achieve the above object, the first aspect of the present invention provides a design method for magnetic components of equipment optimized by dynamic crowding degree, including the following steps:

[0010] Generate an initial population through random coding based on the constraint ranges of temperature, frequency, peak magnetic flux density, waveform, and magnetic core material;

[0011] Calculate the magnetic core loss and transmitted magnetic energy for each individual in the initial population, obtain the magnetic core loss value and transmitted magnetic energy value, and record them as the objective function values;

[0012] Perform fast non-dominated sorting on the initial population according to the objective function values, divide the individuals in the initial population into multiple non-dominated levels, calculate the dynamic crowding degree of each individual in the non-dominated level, and obtain the crowding degree sorting result;

[0013] Based on the crowding degree sorting result, iteratively update the initial population until the maximum number of iterations is reached;

[0014] When the maximum number of iterations is reached, output the non-dominated solution set on the Pareto front;

[0015] Set the weight vector according to the requirements, and calculate the weighted Euclidean distance between each solution in the non-dominated solution set and the ideal point;

[0016] Select the solution with the smallest weighted Euclidean distance as the optimal magnetic component design scheme, and output the five-factor parameter combination of temperature, frequency, waveform, peak magnetic flux density, and magnetic core material.

[0017] Furthermore, it also includes: performing polynomial interpolation fitting on the non-dominated solution set on the Pareto front, calculating the curvature based on the second derivative of the fitting curve, identifying the inflection point with the maximum curvature, and taking the five-factor parameter combination corresponding to the inflection point as the optimal magnetic component design scheme in the loss mutation sensitive scenario.

[0018] Furthermore, the method for calculating the magnetic core loss and transmitted magnetic energy for each individual in the initial population and recording the objective function value of the individual includes:

[0019] Decode the decision variables of each individual in the initial population to obtain specific temperature, frequency, waveform, peak magnetic flux density, and magnetic core material parameters:

[0020] Based on the decoded parameters, calculate the magnetic core loss value based on the following magnetic core loss model:

[0021]

[0022]

[0023] Calculate the transmitted magnetic energy value based on the following transmitted magnetic energy formula:

[0024]

[0025] Record the core loss value and the transferred magnetic energy value as two sub-goals of the objective function respectively;

[0026] Among them, represents the core loss, is the decision variable, represents the temperature, represents the operating frequency of the magnetic component, represents the waveform, represents the peak magnetic flux density, represents the core material, represents the transferred magnetic energy.

[0027] Furthermore, the fast non-dominated sorting includes the following methods:

[0028] Sort all individuals according to the objective function value, and identify the individuals that are not dominated by other individuals to form the first non-dominated layer;

[0029] Repeatedly identify non-dominated layers from the remaining individuals until all individuals are assigned.

[0030] Furthermore, the calculation of the dynamic crowding degree includes the following methods:

[0031] Sort the individuals in the population, and arrange the non-boundary individuals in sequence according to the objective function value;

[0032] Initialize the initial crowding degree of each individual, assign the maximum crowding degree to the boundary individuals, and initially set the left and right crowding degrees of non-boundary individuals to zero;

[0033] Calculate the left crowding degree of non-boundary individuals, and the left crowding degree is determined based on the difference in the objective function values between the current individual and its adjacent individual on the left;

[0034] Calculate the right crowding degree of non-boundary individuals, and the right crowding degree is determined based on the difference in the objective function values between the current individual and its adjacent individual on the right;

[0035] Accumulate the left crowding degree and the right crowding degree to obtain the dynamic crowding degree of the current individual;

[0036] When an individual needs to be deleted from the population, identify the individual with the smallest crowding degree and delete it;

[0037] Recalculate the dynamic crowding degree of the left and right adjacent individuals of the deleted individual, and update the dynamic crowding degree;

[0038] Based on the updated dynamic crowding degree, re-sort the remaining individuals;

[0039] Repeat the above steps until the requirement for the number of solution sets is met.

[0040] Further, based on the dynamic crowding degree sorting result, perform the following operations until the maximum number of iterations is reached:

[0041] Select individuals with higher non-dominated level priority and larger dynamic crowding degree to form the parent population;

[0042] Perform crossover and mutation operations on the parent population to generate the offspring population;

[0043] Merge the parent population and the offspring population to obtain a new population;

[0044] Perform fast non-dominated sorting on the new population and sort according to the dynamic crowding degree, and select the optimal individuals from it to form the next generation population;

[0045] Increase the number of iterations. If the number is greater than the maximum number of iterations, terminate the algorithm and output the optimal solution set; otherwise, return to the above steps.

[0046] Further, the method for calculating the weighted Euclidean distance includes:

[0047] Standardize the core loss and transmission magnetic energy of each solution in the non-dominated solution set to obtain the standardized values and , where the standardization process is as follows:

[0048]

[0049] Among them, and are the standardized values, corresponding to the standardization results of the core loss and transmission magnetic energy respectively, represents the core loss value corresponding to the alternative magnetic component design scheme, represents the transmission magnetic energy value corresponding to the alternative magnetic component design scheme, is the value set of all alternative magnetic component design schemes on the core loss index, is the value set of all alternative magnetic component design schemes on the transmission magnetic energy index, is the minimum value among all core loss values, that is, the optimal core loss, is the maximum value among all core loss values, that is, the worst core loss, is the minimum value among all transmission magnetic energy values, that is, the worst transmission magnetic energy, is the maximum value among all transmission magnetic energy values, that is, the optimal transmission magnetic energy;

[0050] Calculate the weighted Euclidean distance of the ideal point based on the standardized values and the weight vector:

[0051]

[0052] Among them, is the weighted Euclidean distance, is the weight vector, is the standardized target value, that is, and , is the ideal point.

[0053] Furthermore, the method for determining the optimal magnetic component design scheme in the loss mutation sensitive scenario includes:

[0054] Obtain the non-dominated solution set on the Pareto front, and sort it according to the two objectives of core loss and transmitted magnetic energy to obtain the discrete points in the solution set;

[0055] Perform polynomial interpolation fitting on the discrete points on the Pareto front to obtain a fitting curve;

[0056] Calculate the second derivative of the fitting curve to obtain the curvature of each discrete point;

[0057] Output the five-factor parameter combination corresponding to the inflection point with the maximum curvature as the optimal magnetic component design scheme.

[0058] To achieve the above object, the second aspect of the present invention provides an electronic device, including a processor and a memory, and the processor is used to implement the steps of the method for designing an equipped magnetic component with dynamic congestion degree optimization when executing the computer program stored in the memory.

[0059] To achieve the above object, the third aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored, and the computer program executes the steps of the method for designing an equipped magnetic component with dynamic congestion degree optimization when run by a processor.

[0060] The beneficial effects of the present invention:

[0061] Compared with the prior art, a method, device and medium for designing a magnetic component with optimized dynamic congestion degree provided by the present invention constructs a multi-objective optimization design model aiming at minimizing core loss and maximizing transmitted magnetic energy, combines the constraint conditions of five factors including temperature, frequency, waveform, peak magnetic flux density and core material, improves the traditional non-dominated sorting genetic algorithm (NSGA-II), and proposes a dynamic congestion degree and adaptive sorting strategy. This technical solution effectively solves the problem of unreasonable selection of individuals when the congestion distance is the same by dynamically updating the congestion degree of individuals; introduces a dynamic congestion degree adaptive sorting method to improve the uniformity and diversity of the population distribution; enhances the convergence and coverage of the solution set through improved fast non-dominated sorting and elite retention strategies, so as to obtain a set of non-dominated solution sets with excellent performance on the Pareto front. Finally, by introducing the weighted improved ideal point method and the curvature inflection point method, combined with the requirements of different application scenarios, the optimal magnetic component design scheme is selected from the Pareto front solution set, realizing the optimization design requirements of magnetic components in multiple scenarios such as loss-averse type, loss-tolerant type, loss-accepting type and loss-mutation sensitive type. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for description in the embodiments.

[0063] Figure 1 It is a flowchart of a method for designing a magnetic component with optimized dynamic congestion degree disclosed in an embodiment of the present invention.

[0064] Figure 2 It is a diagram of the non-dominated solution set of the DNSGA-II algorithm disclosed in an embodiment of the present invention.

[0065] Figure 3 It is a relationship diagram between core loss and transmitted magnetic energy disclosed in an embodiment of the present invention.

[0066] Figure 4 It is a distribution diagram of core loss and transmitted magnetic energy under different temperature conditions disclosed in an embodiment of the present invention.

[0067] Figure 5 It is an influence diagram of frequency on core loss and transmitted magnetic energy disclosed in an embodiment of the present invention.

[0068] Figure 6 It is an influence diagram of different waveforms on core loss and transmitted magnetic energy disclosed in an embodiment of the present invention.

[0069] Figure 7 It is an influence diagram of the peak magnetic flux density on core loss and transmitted magnetic energy disclosed in an embodiment of the present invention.

[0070] Figure 8It is a diagram showing the influence of different magnetic core materials on magnetic core loss and transmitted magnetic energy disclosed in an embodiment of the present invention.

[0071] Figure 9 It is a comparison diagram of the Pareto front solution sets of the DNSGA-II algorithm and the NSGA-II algorithm disclosed in an embodiment of the present invention.

[0072] Figure 10 It is a radar chart comparing the indicators of the DNSGA-II algorithm and the NSGA-II algorithm disclosed in an embodiment of the present invention.

[0073] Figure 11 It is a diagram for selecting the optimal magnetic component design scheme for loss-averse, loss-tolerant, and loss-accepting types disclosed in an embodiment of the present invention.

[0074] Figure 12 It is a diagram for selecting the optimal magnetic component design scheme for loss mutation-sensitive types disclosed in an embodiment of the present invention. Detailed implementation manners

[0075] To enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0076] According to the embodiments of the present invention, it should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although the logical order is shown in the following manufacturing method, in some cases, the steps shown or described can be executed in a different order than here.

[0077] As Figure 1 shown, the present invention provides a method for designing magnetic components of equipment with optimized dynamic crowding degree, including the following steps:

[0078] Step S100: Generate an initial population through random coding based on the constraint ranges of temperature, frequency, peak magnetic flux density, waveform, and magnetic core material;

[0079] Step S200: Calculate the magnetic core loss and the transmitted magnetic energy for each individual in the initial population, obtain the magnetic core loss value and the transmitted magnetic energy value, and record them as the objective function values;

[0080] Step S300: Perform fast non-dominated sorting on the initial population according to the objective function value, divide the individuals in the initial population into multiple non-dominated levels, calculate the dynamic crowding degree of each individual in the non-dominated level, and obtain the crowding degree sorting result;

[0081] Step S400: Based on the crowding degree sorting result, iteratively update the initial population until the maximum number of iterations is reached;

[0082] Step S500: When the maximum number of iterations is reached, output the non-dominated solution set on the Pareto front.

[0083] Step S600: Set the weight vector according to the requirements, and calculate the weighted Euclidean distance between each solution in the non-dominated solution set and the ideal point;

[0084] Step S700: Select the solution with the smallest weighted Euclidean distance as the optimal magnetic component design scheme, and output the five-factor parameter combination of temperature, frequency, waveform, peak magnetic flux density, and magnetic core material.

[0085] In this embodiment, as described in step S100 above, based on the constraint ranges of temperature, frequency, peak magnetic flux density, waveform, and magnetic core material, an initial population is generated by random coding, including the following: First, encode each design variable according to the preset constraint range, and implement the encoding rule using the mapping relationship between continuous floating-point numbers and decision vectors, where the temperature has a value range of , the frequency has a value range of , the peak magnetic flux density has a value range of , the waveform has a value range of the discrete set {1, 2, 3}, and the magnetic core material has a value range of the discrete set {1, 2, 3, 4}. Then, based on the random floating-point number generation method, generate random coding values for each individual for , , , and respectively, and map the floating-point numbers to the corresponding physical parameter values through the decoding formula to ensure that the generated parameter combinations meet the constraint conditions. Finally, an initial population with a population size of is generated to provide basic data for the subsequent optimization process.

[0086] Gene coding , where are all continuous floating-point numbers in (0, 1), and the mapping relationship with the elements in the decision vector is as follows:

[0087]

[0088] Among them, represents the temperature of the magnetic component, and are respectively the maximum and minimum values of the temperature in the experimental data, represents the coded value of the temperature within the range of [0, 1]. Through mapping, the actual value of the temperature can be obtained; represents the frequency, and are respectively the maximum and minimum values of the frequency in the experimental data, represents the coded value of the frequency within the range of [0, 1]. Through mapping with this value, the actual value of the frequency can be obtained; represents the peak magnetic flux density, and are respectively the maximum and minimum values of the peak magnetic flux density in the experimental data, represents the coded value of the peak magnetic flux density within the range of [0, 1]; represents the waveform, which can be one of three waveforms: sine wave, triangular wave, or trapezoidal wave, represents the coded value of the waveform; is the type of the magnetic core material, represents the coded value of the magnetic core material.

[0089] In this embodiment, as described in the above step S200, the core loss calculation and the transmitted magnetic energy calculation are performed on each individual in the initial population, and the objective function values of the individuals are recorded. Specifically, it includes the following content: First, decode the decision variables of the individuals in the initial population, to obtain the actual parameter values of the temperature , frequency , peak magnetic flux density , waveform and magnetic core material . Second, based on the core loss model , substitute the decoded parameters into the core loss calculation formula to calculate the core loss value of the individual. Then, according to the transmitted magnetic energy formula , substitute the frequency and the peak magnetic flux density to calculate the transmitted magnetic energy value

[0090] of the individual. Finally, record the core loss value and the transmitted magnetic energy value of each individual respectively as the two sub-objectives of the objective function, providing basic data for subsequent non-dominated sorting and dynamic crowding degree calculation.In this embodiment, as described in the above step S300, for the optimization problem of two objectives such as the optimal design model of magnetic components, some multi-objective evolutionary algorithms can be used to find multiple Pareto optimal solutions, such as the nondominated sorting geneticalgorithm (NSGA). Aiming at the three problems existing in NSGA, namely, the excessively high time complexity of nondominated sorting, the lack of elitist-preserving strategy, and the need to artificially specify sharing parameters to ensure population diversity, NSGA-I has three key improvements: fast nondominated sorting, elitist-preserving strategy, and parameterless niche operator, and has better performance in terms of solution quality and convergence.

[0091] The NSGA-I algorithm stratifies the population through the dominance relationship between individuals. First, it identifies the set of nondominated individuals as the first layer. Subsequently, it further screens out nondominated individuals from the remaining individuals to form the second layer, and this process continues until all individuals are stratified. The set of individuals in each layer represents the solution set that is not dominated by other individuals in the current iteration. Within each layer, the algorithm further calculates the crowding distance of individuals, which is an index to measure the distribution density of individuals in the objective space. When performing the selection operation, NSGA-II first compares according to the layer where the individuals are located and the crowding distance to determine the optimal individual set to form a new parental population. Next, the algorithm uses the crossover and mutation operations of the genetic algorithm to generate a new offspring population. After merging the newly generated offspring population with the existing parental population, genetic operations, including selection, crossover, and mutation, are performed to generate a new generation of population. This process is continuously iterated until the preset population size or other termination conditions are met. This calculation method of stratification and crowding distance, as well as the selection mechanism based on these indexes, is the key for the NSGA-II algorithm to effectively handle multi-objective optimization problems. In this way, the algorithm can consider both the superiority and diversity of solutions, so as to explore a set of balanced Pareto optimal solution sets in the solution space.

[0092] To optimize the diversity of the population, this embodiment innovates the calculation method of crowding degree and introduces the concept of dynamic crowding degree. This method solves the problem that it is difficult to determine which individual should be deleted when the crowding degrees are the same. Specifically, when an individual needs to be removed from the population, first identify the solution with the lowest crowding degree and delete it. Subsequently, update the crowding degrees of neighboring individuals to ensure that the distribution of the remaining individuals in the population is more uniform, thereby maintaining the diversity of the population.

[0093] During the application process of the NSGA-II algorithm, when two individuals have the same crowding distance, the traditional crowding degree measurement method cannot clearly indicate which individual should be given priority when performing the individual deletion operation. In this case, the algorithm faces a selection difficulty because there is no clear criterion to guide the decision-making of individual elimination. Take the 6 individuals in Table 1 below as an example:

[0094] Table 1 Crowding Distance Table

[0095]

[0096] Table 1 shows the crowding distance values calculated for 6 individuals under a single objective. According to the criteria of the NSGA-I algorithm, the crowding degree is calculated as follows:

[0097]

[0098] where, represents the crowding degree of individual ; and respectively represent the individuals adjacent to individual in the objective space, represents the individual, represents the objective space. This formula determines the priority of individuals by calculating the crowding degree of each individual. Individuals with a smaller crowding degree will be retained, while individuals with a larger crowding degree may be eliminated to enhance the diversity of solutions. When faced with the need to delete individuals according to the crowding degree and multiple individuals have the same crowding distance, the algorithm usually cannot determine which individual should be deleted. In this case, the NSGA-II algorithm may randomly select an individual for deletion.

[0099] To avoid potential problems caused by this randomness, this embodiment introduces a new left and right crowding degree index. This method first initializes the crowding degree of each individual to ensure that the crowding degree of boundary nodes is given the maximum value to ensure that they can be retained in the next generation. For non-boundary nodes, the concepts of left crowding degree and right crowding degree are introduced and initially set to 0.

[0100] After sorting the sub-objective function values, the left crowding degree and right crowding degree of each non-boundary node are calculated in turn. By this method, the crowding degree of each individual can be evaluated more precisely, so as to provide a more reasonable individual selection mechanism when facing the situation of equal crowding distances. Finally, based on the calculated left crowding degree and right crowding degree, the crowding degree of each node can be determined, so as to make a more informed decision in the deletion operation. This improvement not only improves the selection efficiency of the algorithm when dealing with individuals with the same crowding degree, but also enhances the protection of the population diversity by dynamically updating the crowding degree, making the final obtained solution set more dispersed in the objective space, which helps to explore a more extensive solution space.

[0101] In the NSGA-II algorithm, the Pareto ranking and crowding degree of individuals are two key indicators for selecting and eliminating individuals. Within the same Pareto ranking, the crowding degree is particularly important because it determines whether an individual is retained or eliminated. However, experimental analysis shows that there are certain limitations in using the fixed crowding degree sorting method. When a large number of individuals with low crowding degrees gather in a certain area, the fixed crowding degree sorting strategy may cause all individuals in this area to be eliminated at one time, thus reducing the diversity of the finally obtained Pareto solution set.

[0102] In practical applications, when an individual is eliminated, the crowding degree of its adjacent individuals will also change. The fixed crowding degree sorting method ignores this change, which may lead to a further reduction in the diversity of the solution set.

[0103] To solve this deficiency of the fixed crowding degree sorting method, this embodiment improves the NSGA-I algorithm, introduces a crowding degree adaptive sorting strategy, and names the improved NSGA-I algorithm as the DNSGA-II algorithm. When the algorithm needs to eliminate the individual with the smallest crowding degree from the Fn-th layer of the population, the DNSGA-II algorithm will adopt the following steps:

[0104] S301. Use the crowding degree adaptive sorting strategy to eliminate the individual with the smallest crowding degree;

[0105] S302. Record the position of the eliminated solution .

[0106] S303. Recalculate the crowding degrees of the solutions located at and .

[0107] S304. Re-sort according to the updated crowding degrees and continue to eliminate the solution with the smallest crowding degree until the number of solutions in the solution set meets the requirements.

[0108] Through this dynamic crowding degree sorting strategy, the DNSGA-I algorithm can more effectively maintain the diversity of the solution set, ensure that the finally obtained Pareto solution set is more evenly distributed in the objective space, and thus improve the optimization effect of the algorithm.

[0109] In this embodiment, as described in the above step S400, the process of the DNSGA-I algorithm is as follows:

[0110] S401. Initialize the population. Randomly generate an initial population of size according to the constraint conditions . Perform function evaluation on each individual, and perform fast non-dominated sorting and dynamic crowding degree calculation on the individuals in the population in turn.

[0111] S402. Select the parent population. Compare the order and crowding distance of the layers where the individuals in are located, and select the optimal individuals from them to form a new parent population.

[0112] S403. Generate the offspring population. Perform crossover and mutation on the parent population to generate a new offspring population .

[0113] S404. Mix the parent population and the offspring population to form a new population .

[0114] S405. Selection. Select excellent individuals to join the population according to the number of layers of fast non-dominated sorting of the intermediate population and the dynamic crowding degree sorting method used in this article .

[0115] S406. Termination condition. Let , if is greater than the maximum number of iterations MaxGen, the algorithm terminates, and at the same time let the non-dominated solution set in be used as the Pareto optimal solution set; if is less than the maximum number of iterations MaxGen, return to step S402.

[0116] In this embodiment, as described in the above step S500 - step S700, the DNSGA-II algorithm is used to solve the magnetic component optimization design model, and the parameter settings are as follows: the population size is 100, the maximum number of iterations MaxGen is 1000, the crossover probability Cross is 0.8, and the mutation probability MU is 0.3. When the maximum number of iterations is reached, the number of non-dominated solutions on the Pareto front is 100. Then, the non-dominated solutions that do not conform to the actual situation on the Pareto front, that is, the non-dominated solutions with core loss less than 0, are removed, and the number of non-dominated solutions is 93. The non-dominated solution set of the DNSGA-II algorithm is asFigure 2 , it can be noted that since the two sub-goals in the optimal design model of magnetic components conflict with each other and cannot reach the optimum simultaneously, it can only be continuously optimized towards the optimum direction by reasonable allocation. The optimization effects of core loss and transferred magnetic energy on the non-dominated solution set are plotted as Figures 3 - 8 , where Figure 3 shows the relationship between core loss and transferred magnetic energy, demonstrating the Pareto front between the two, that is, the trend of mutual restriction during the optimization process; Figure 4 shows the distribution of core loss and transferred magnetic energy under different temperature conditions. The color represents the temperature value, and the influence of temperature on the optimization solution can be observed; Figure 5 reflects the influence of frequency on core loss and transferred magnetic energy. The lighter the color, the higher the frequency, revealing that high frequency may increase the transferred magnetic energy but also increase the core loss; Figure 6 shows the influence of different waveforms on core loss and transferred magnetic energy. The color represents the waveform category, and the influence trend of different waveforms on the optimization goal can be analyzed; Figure 7 shows the influence of the peak value of magnetic flux density on core loss and transferred magnetic energy. The color represents the magnetic flux density, indicating that the increase in magnetic flux density may help increase the transferred magnetic energy; Figure 8 shows the influence of different core materials on core loss and transferred magnetic energy. The color represents different core materials, and the performance differences of different materials in the optimization goal can be observed.

[0117] The multi-objective optimization algorithm obtains the optimal solution by continuously converging the Pareto solution set and maintains strong diversity. To verify the advantages of the DNSGA-II algorithm proposed in this embodiment, this embodiment is compared with the NSGA-I algorithm. The test parameters used are the same as those of the DNSGA-I algorithm proposed in this embodiment, and the comparison of the Pareto front solution sets of the DNSGA-II algorithm and the NSGA-II algorithm is as Figure 9 .

[0118] Five indicators, namely the Inverted Generational Distance (IGD), Hypervolume (HV), Set Coverage (SC), Spacing (SM), and Schottky Spread (SSPread), are selected to evaluate the solution effects of the DNSGA-I algorithm and the NSGA-II algorithm respectively. The specific introductions of these 5 types of indicators are as follows:

[0119] (1) Inverted Generational Distance

[0120] The Inverted Generational Distance (IGD) is an important tool for measuring the performance of an algorithm. IGD evaluates the convergence and distribution of an algorithm by calculating the sum of the shortest distances between the individuals on the Pareto surface and the Pareto optimal solution set generated by the algorithm. As follows:

[0121]

[0122] Among them, is the set of individuals distributed on the Pareto surface, is the number of individuals distributed on the Pareto surface, is the Pareto optimal solution set, is the Pareto surface The individual to the Pareto optimal solution set The minimum Euclidean distance. Generally, the smaller the IGD, the better the convergence and distribution efficiency of the algorithm.

[0123] (2) Hypervolume indicator

[0124] The hypervolume indicator (Hypervolume, abbreviated as HV) is a performance evaluation indicator widely used in multi-objective optimization. It measures the volume of the hypercube enclosed by the non-dominated solution set generated by the algorithm and a reference point in the objective space. This indicator can evaluate the convergence and diversity of the solution set at the same time, that is, how close the solution set is to the true Pareto front and the distribution of the solutions. The calculation formula is as follows:

[0125]

[0126] Among them, is the non-dominated solution set, is the hypervolume element, Essentially, it calculates the sum of the volumes of the hypercubes formed by the points in all Pareto solution sets and the reference point. In this embodiment, the reference point is selected as , where represents the maximum value of the TME index in the solution set obtained by the DNSGA-II algorithm.

[0127] (3) Set coverage indicator

[0128] The set coverage indicator (SetCoverage, abbreviated as SC) is a binary indicator used to compare the coverage relationship between two solution sets A and B. It measures the degree to which the solutions in one solution set are dominated by the solutions in the other solution set. Specifically, SC(A, B) represents the proportion of the solutions in B that are dominated by at least one solution in A among the number of solutions in B. If the value of SC(A, B) is 1, it means that all solutions in B are dominated by at least one solution in A; if the value is 0, it means that no solution in B is dominated by the solutions in A. The SC indicator can reflect the coverage ability of the solution set B generated by the algorithm relative to the reference solution set A. The smaller the value, the better the solution set B. In this embodiment, B is the solution set generated by the corresponding algorithm (NSGA-II or DNSGA-II), and A is the solution set generated by the DNSGA-II algorithm.

[0129] (4)Spacing Metric

[0130] The Spacing Metric (SM for short) is an index used to evaluate the uniformity of the distribution of the non-dominated solution set generated by multi-objective optimization algorithms. It measures the distribution of the solution set in the objective space by calculating the distance between each solution in the solution set and its nearest neighbor solution. If these distances are all equal, it means that the solution set is evenly distributed on the Pareto front, and the value of the Spacing Metric will be smaller, indicating that the algorithm performs well in terms of diversity. On the contrary, if the distance differences are large, it means that the distribution in the solution set is uneven, and the value of the Spacing Metric will be larger, indicating that the algorithm has deficiencies in covering the entire Pareto front. It should be noted that the SM is particularly suitable for optimization problems with two objectives and may not be effective enough in the case of more than two objectives. The calculation formula is as follows:

[0131]

[0132] where, represents the Euclidean distance between two consecutive Pareto fronts, represents the arithmetic mean of represents the Euclidean distance between the actual Pareto front and the ideal Pareto front, represents the number of solutions on the actual Pareto front, represents the number of solutions in the non-dominated solution set. Generally, the smaller the SM, the more evenly distributed the Pareto front obtained by the algorithm is.

[0133] (5)Schott's Spread

[0134] Schott's Spread (SSpread for short) is an index used to measure the uniformity of the distribution of the non-dominated solution set generated by multi-objective optimization algorithms. It is achieved by calculating the standard deviation of the distances between each solution in the solution set and its nearest neighbor solution. The smaller the value of Schott's Spread (i.e., SS pread ), the more evenly distributed the solution set is. The calculation formula is as follows:

[0135]

[0136] where, is the total number of solutions in the solution set, the th solution to the distance to its nearest neighbor solution in is the average of all .

[0137] The results of these five evaluation indicators are shown in Table 2 and Figure 10 as follows.

[0138] Table 2 Comparison of indicators between NSGA-I and DNSGA-II algorithms

[0139]

[0140] *To avoid inconvenient display due to too large values, the HV, SC, and SS pread indicators are normalized.

[0141] All of the above five indicators are better when they are smaller. It can be found that, compared with the NSGA-I algorithm, the DNSGA-II algorithm is significantly better in four indicators: the inverse generational distance evaluation indicator (IGD), the hypervolume indicator (HV), the set coverage indicator (SC), and the spacing indicator (SM), and is only worse in the short distribution indicator (SSPread), which further verifies that the DNSGA-I algorithm has stronger convergence, diversity, uniformity, and robustness.

[0142] To achieve the two sub-goals of minimizing the core loss and maximizing the transmitted magnetic energy, it is necessary to select the optimal conditions from the Pareto solution set according to the requirements of different schemes. In this embodiment, based on the improved ideal point method, multiple Pareto solutions are retained for different preference scenarios, and the corresponding optimal magnetic component design schemes are given respectively, further expanding the multi-objective optimization method into a multi-modal multi-objective optimization method. Its significance lies in: (1) It can meet the magnetic energy transmission requirements under preset different preference scenarios; (2) It is beneficial to reveal some potential characteristics in the process of magnetic energy transmission applications; (3) It is convenient to quickly match and correct the optimal magnetic component design scheme for new scenarios to solve the dynamic adaptive optimization problem; (4) It can increase the possibility of finding an optimal magnetic component design scheme with stronger robustness.

[0143] Suppose the obtained Pareto front is , where is the final population size (obtained previously ), and the core loss and transmitted magnetic energy corresponding to the alternative magnetic component design scheme are and respectively, and there are .

[0144] First, standardize the two indicators to make them have the same magnitude. Among them, the core loss is a cost-type indicator, and the smaller it is, the better. The standardization process is as follows. The transmitted magnetic energy is a benefit-type indicator, and the larger it is, the better. The standardization process is as follows:

[0145]

[0146] Among them, and are the standardized values, corresponding to the standardized results of core loss and transferred magnetic energy respectively, represents the core loss value corresponding to the alternative magnetic component design scheme, represents the transferred magnetic energy value corresponding to the alternative magnetic component design scheme, is the value set of all alternative magnetic component design schemes for the core loss index, is the value set of all alternative magnetic component design schemes for the transferred magnetic energy index, is the minimum value among all core loss values, that is, the optimal core loss, is the maximum value among all core loss values, that is, the worst core loss, is the minimum value among all transferred magnetic energy values, that is, the worst transferred magnetic energy, is the maximum value among all transferred magnetic energy values, that is, the optimal transferred magnetic energy;

[0147] Let the ideal point be the set of idealized values of all indexes. In the general ideal point method, according to the Euclidean distance formula, the distance between the alternative magnetic component design scheme and the ideal point is calculated as follows:

[0148]

[0149] Among them, the two indexes of the standardized core loss and transferred magnetic energy are both cost-type indexes where the lower the better, so . Finally, all alternative magnetic component design schemes are sorted according to the calculated ideal distance, and the alternative magnetic component design scheme farthest from the ideal point is the most ideal choice.

[0150] In this embodiment, a weight vector is added based on different preference scenarios, satisfying and where and represent the preference weight coefficients of core loss and transferred magnetic energy respectively. Then, in the improved ideal point method, the distance between the alternative magnetic component design scheme and the ideal point is as follows:

[0151]

[0152] Among them, is the weighted Euclidean distance, is the weight vector, is the standardized target value, that is, and , is the ideal point;

[0153] According to the differences in the preference degrees of minimizing core loss and maximizing transmitted magnetic energy in different scenarios, different weight coefficients and regularization parameters are assigned in the ideal point method measurement. In this section, the following three application scenarios are first considered:

[0154] (1) Loss-averse scenario. In applications with high requirements for energy efficiency, such as LED lighting power supplies or chargers for mobile devices, minimizing core loss is the primary consideration in the design. Because these devices usually have strict requirements for thermal management and efficiency, excessive core loss will lead to energy waste and increase the burden on the heat dissipation system. At this time, the weight vector .

[0155] (2) Loss-tolerant scenario. For high-power transmission applications, such as inverters for electric vehicles or solar inverters, the priority of transmitting magnetic energy may be higher. These applications need to process and transmit a large amount of energy in a short time, so the core material needs to be able to support high-power magnetic energy transfer, even if this may mean slightly increasing some core losses. At this time, the weight vector .

[0156] (3) Loss-acceptable scenario. For applications with a short expected usage time or intermittent usage, such as handheld power tools or temporary power supplies, a certain degree of core loss may be acceptable. In these cases, users may be more concerned about the magnetic energy transmission efficiency of the device to reduce energy cost expenditures rather than the long-term energy efficiency performance. At this time, the weight vector .

[0157] In addition, in practical applications, if the existing magnetic energy transmission efficiency can already well meet the current needs, but it is desired to slightly increase the transmitted magnetic energy without causing an obvious mutation in core loss, this embodiment names this application scenario as a loss mutation-sensitive scenario.

[0158] For the loss mutation sensitive scenario, in essence, it is hoped to find the design scheme of the magnetic component corresponding to the point with the largest bending degree (i.e., the point with the largest curvature) on the Pareto front of "core loss - transferred magnetic energy". This point shows a sudden change in trend on the curve, such as changing from rapid growth to slow growth, or vice versa. For continuous sequences, the curvature can be calculated by various methods, including the angle method, Menger curvature, or exponentially weighted moving average (EWMA). For the uneven discrete sequence such as the Pareto front of "core loss - transferred magnetic energy", first, polynomial interpolation method is needed for fitting. Then, the points on the fitted curve are evenly sampled according to the core loss to obtain an equally spaced uniform discrete sequence. Finally, the curvature of each point on the Pareto front is calculated based on the second derivative calculation formula of the uniform discrete sequence, and the point with the largest curvature is found as the inflection point.

[0159]

[0160] Among them, represents a finite difference approximation form of the second derivative (i.e., curvature) of the function at the point , reflecting the curvature at the discrete point , that is, the change trend at this point; represents the function value of the previous point; represents the function value of the next point; represents the function value of the current point; it can be understood that the calculation result of this formula represents the curvature, used to measure the trend change at :

[0161] If , it means that this point is a local minimum point (concave, the curve bends upward);

[0162] If , it means that this point is a local maximum point (convex, the curve bends downward);

[0163] If the absolute value is large, it means that the change trend at this point is steep, that is, the curvature is large.

[0164] Based on the above methods, the optimal magnetic component target achievement effects in the four scenarios of loss aversion type, loss tolerance type, loss acceptance type, and loss mutation sensitivity type are obtained as shown in Table 3 and Figure 11 , Figure 12 as shown.

[0165] Table 3 Optimal magnetic component target achievement effects in different scenarios

[0166]

[0167] From Figure 11It can be noted that as the attention weights of the three scenarios of "loss aversion - loss tolerance - loss acceptance" for core loss continuously decrease and the attention weights for transmitted magnetic energy continuously increase, the corresponding optimal magnetic component target achievement effect moves towards the upper right corner, and both the core loss and the transmitted magnetic energy increase continuously;

[0168] At this time, the optimal magnetic component design schemes under the four scenarios of loss aversion, loss tolerance, loss acceptance, and loss mutation sensitivity are obtained, as shown in Table 4:

[0169] Table 4 Optimal magnetic component design schemes for different scenarios

[0170]

[0171] In summary, the method of this embodiment establishes an optimization design model for magnetic components, and gives the 0-1 continuous coding rules for five conditions of temperature, frequency, waveform, peak magnetic flux density, and core material; then, on the basis of the NSGA-II multi-objective optimization algorithm, the concepts of proximity crowding degree and dynamic crowding degree are introduced, the improved DNSGA-II algorithm is proposed, and the solution process of this algorithm is given; then, the DNSGA-II algorithm is run under the parameter conditions of population size N = 100, maximum number of iterations MaxGen = 1000, crossover probability Cross = 0.8, and mutation probability MU = 0.3. After simple positive value filtering, a set (93) of non-dominated solution sets on the Pareto front is obtained, and compared with the NSGA-II algorithm, it is found that the DNSGA-II algorithm is significantly superior to the NSGA-II algorithm in four indicators: the inverted generational distance evaluation index (IGD), hypervolume index (HV), set coverage index (SC), and spacing index (SM), further verifying that the DNSGA-II algorithm has stronger convergence, diversity, uniformity, and robustness; finally, based on the improved ideal point method and inflection point method with weights, the optimal magnetic component design schemes under the four scenarios of loss aversion, loss tolerance, loss acceptance, and loss mutation sensitivity are obtained. In the order of loss mutation sensitivity, loss aversion, loss tolerance, and loss acceptance, the core loss and transmitted magnetic energy of their respective corresponding optimal magnetic component design schemes increase in turn.

[0172] According to another aspect of the embodiments of the present application, an electronic device is further provided, including a processor and a memory, and the processor is used to implement the steps of the method when executing the computer program stored in the memory.

[0173] In the above embodiments of the present invention, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0174] In several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. For example, the division of the units can be a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of units or modules can be in an electrical or other form.

[0175] In addition, each functional unit in various embodiments of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0176] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs and other various media that can store program codes.

[0177] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can still be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for designing equipment magnetic components with dynamic crowding optimization, characterized in that: The steps include: Generate the initial population by random coding based on the constraints of temperature, frequency, magnetic flux density peak, waveform, and core material; Calculate the core loss and transmission magnetic energy of each individual in the initial population, obtain the core loss value and transmission magnetic energy value, and record them as the objective function value; According to the objective function value, the initial population is quickly non-dominated sorted, the individuals in the initial population are divided into multiple non-dominated layers, the dynamic crowding degree of each individual in the non-dominated layer is calculated, and the crowding degree sorting result is obtained; Based on the crowding ranking results, the initial population is iteratively updated until the maximum number of iterations is reached; When the maximum number of iterations is reached, the non-dominated solution set on the Pareto front is output; A weight vector is set according to the requirement, and a weighted Euclidean distance between each solution in the non-dominated solution set and the ideal point is calculated; The solution with the smallest weighted Euclidean distance is selected as the optimal magnetic component design solution, and the five-factor parameter combination of temperature, frequency, waveform, magnetic flux density peak and core material is output.

2. The method for designing magnetic components for equipment with dynamic congestion optimization according to claim 1, characterized in that: Also includes: Polynomial interpolation fitting is performed on the non-dominated solution set on the Pareto front, and the curvature is calculated based on the second-order derivative of the fitting curve. The inflection point with the largest curvature is identified, and the five-factor parameter combination corresponding to the inflection point is used as the optimal magnetic component design scheme for loss mutation sensitive scenarios.

3. The method for designing magnetic components for equipment with dynamic congestion optimization according to claim 1, characterized in that: The method of calculating the core loss and the transmitted magnetic energy for each individual in the initial population, obtaining the core loss value and the transmitted magnetic energy value, and recording them as the objective function value includes: The decision variables of each individual in the initial population are decoded to obtain the specific temperature, frequency, waveform, peak value of magnetic flux density and core material parameters: According to the decoded parameters, the core loss value is calculated based on the following core loss model: The transmission magnetic energy value is calculated based on the following transmission magnetic energy formula: The core loss value and the transmitted magnetic energy value are recorded as two sub-objectives of the objective function value; in, represents the core loss, is the mathematical model used to calculate the core loss, which represents the core loss and decision variables The relationship between is the decision variable, Indicates temperature, represents the operating frequency of the magnetic component, Represents the waveform, represents the peak value of magnetic flux density, Indicates the core material, Indicates the transmission of magnetic energy.

4. The method for designing magnetic components for equipment with dynamic congestion optimization according to claim 1, characterized in that: Fast non-dominated sorting includes the following methods: Sort all individuals by the objective function value, identify individuals that are not dominated by other individuals to form the first non-dominated layer; Repeatedly identify the non-dominated layer from the remaining individuals until all individuals are assigned.

5. The method for designing magnetic components for equipment with dynamic congestion optimization according to claim 1, characterized in that: The calculation of dynamic congestion includes the following methods: Sort the individuals in the population and arrange the non-boundary individuals in order according to the objective function value; Initialize the initial crowding of each individual, assign the maximum crowding to the boundary individual, and set the left and right crowding of the non-boundary individual to zero initially; Calculate the left crowding degree of the non-boundary individual, where the left crowding degree is determined based on the difference in objective function values ​​between the current individual and its left adjacent individual; Calculate the right crowding degree of the non-boundary individual, where the right crowding degree is determined based on the difference in objective function values ​​between the current individual and its right adjacent individual; The left crowding degree and the right crowding degree are accumulated to obtain the dynamic crowding degree of the current individual; When individuals need to be deleted from the population, the individuals with the smallest dynamic crowding are identified and deleted; Recalculate the dynamic crowding degree of the left and right adjacent individuals of the deleted individual and update the dynamic crowding degree; Based on the updated dynamic crowding degree, the remaining individuals are re-sorted; Repeat the above steps until the required number of solutions is met.

6. The method for designing magnetic components for equipment with dynamic congestion optimization according to claim 1, characterized in that: Based on the crowding sorting results, the following operations are performed until the maximum number of iterations is reached: Individuals with a priority on non-dominated hierarchy and a large dynamic crowding degree are selected to form the parent population; Perform crossover and mutation operations on the parent population to generate the offspring population; Merge the parent population and the child population to get a new population; Perform fast non-dominated sorting on the new population and sort it according to the dynamic crowding degree, and select the best individuals from them to form the next generation population; Increase the number of iterations. If the number is greater than the maximum number of iterations, terminate the algorithm and output the optimal solution set; otherwise, return to the above steps.

7. The method for designing magnetic components for equipment with dynamic congestion optimization according to claim 1, characterized in that: Methods for calculating weighted Euclidean distance include: The core loss and transmitted magnetic energy of each solution in the non-dominated solution set are normalized to obtain the standardized values and , where the standardization process is as follows: in, and are the standardized values, corresponding to the standardized results of core loss and transmitted magnetic energy, respectively. Indicates the core loss value corresponding to the alternative magnetic component design scheme, It represents the transmission magnetic energy value corresponding to the alternative magnetic component design scheme, It is the set of values ​​of all alternative magnetic component design schemes in terms of core loss index. It is the set of values ​​of all alternative magnetic component design schemes in terms of transmission magnetic energy index. is the minimum value among all core loss values, that is, the optimal core loss. It is the maximum value of all core loss values, that is, the worst core loss. It is the minimum value among all the transmission magnetic energy values, that is, the worst transmission magnetic energy. It is the maximum value among all the transmission magnetic energy values, that is, the optimal transmission magnetic energy; The weighted Euclidean distance between each solution in the non-dominated solution set and the ideal point is calculated based on the standardized value and the weight vector: in, is the weighted Euclidean distance, is the weight vector, is the standardized target value, that is and , For the ideal point.

8. The method for designing magnetic components for equipment with dynamic congestion optimization according to claim 2, characterized in that: Methods for determining the optimal magnetic component design for loss-susceptible scenarios include: Obtain the non-dominated solution set on the Pareto front, sort it according to the two objectives of core loss and transmitted magnetic energy, and obtain the discrete points in the solution set; Perform polynomial interpolation fitting on the discrete points on the Pareto front to obtain a fitting curve; Calculate the second derivative of the fitting curve to obtain the curvature of each discrete point; The five-factor parameter combination corresponding to the inflection point with the maximum output curvature is taken as the optimal magnetic component design solution.

9. An electronic device, characterized in that: The invention comprises a processor and a memory, wherein the processor is used to implement the steps of the method for designing magnetic components of equipment with dynamic crowding optimization as claimed in any one of claims 1 to 8 when executing a computer program stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for designing magnetic components for equipment with dynamic crowding optimization according to any one of claims 1 to 8 are executed.

Images