Power electronic transformer optimization method

Abstract

The invention discloses a power electronic transformer optimization method, and relates to the field of power electronic transformer optimization. Comprising a pre-processing module, a multi-objective optimization module and a post-processing module, the pre-processing module screens out candidate target values meeting design conditions through a free parameter scanning method, the candidate target values serve as an initial population of the multi-objective optimization module, and the post-processing module further processes a Pareto solution set obtained by the multi-objective optimization module to obtain a unique optimal solution. The objective guidance of the free parameter scanning method, the efficient calculation capability of the non-dominated sorting genetic algorithm and the weight determination mechanism of the entropy weight method are fused, so that the problems of randomness, multiplicity, subjectivity and the like in the optimization process are effectively solved, and the effective scientific guidance on the optimization design of the power electronic transformer is realized.

Description

Technical Field

[0001] The invention belongs to the field of power electronic transformer optimization, and in particular relates to a power electronic transformer optimization method. Background Art

[0002] With the continuous growth of global energy demand and increasing concern about environmental impact, power electronics technology is increasingly used in power systems. As a key component in this field, power electronic transformers are gradually replacing traditional mechanical transformers and becoming a new solution for power conversion and distribution. The application of power electronic transformers in renewable energy systems, electric vehicle charging stations, smart grids, and distributed generation provides technical support for the efficient use of energy and the widespread promotion of clean energy. They can better adapt to the fluctuations of the power grid, improve the reliability and stability of the system, and reduce energy loss and operating costs. However, the design and application of power electronic transformers still face many challenges, including thermal management of devices, electromagnetic compatibility, system stability analysis, and development of control strategies. To overcome these challenges, researchers and engineers are working on developing new materials, new topologies, and advanced optimization design algorithms to achieve more efficient, reliable, and intelligent power electronic transformers.

[0003] The design of power electronic transformers needs to comprehensively consider factors such as work efficiency, manufacturing cost, volume, temperature rise, etc. Therefore, the optimization of power electronic transformers is a typical multi-objective optimization problem. Many researchers have optimized the design of power electronic transformers by improving multi-objective optimization algorithms. Multi-objective optimization algorithm is an algorithm used to solve optimization problems with two or more conflicting objectives. In such problems, there is usually no single optimal solution, but a Pareto optimal solution set of multiple solutions. The objective function value corresponding to the Pareto optimal solution set is called the Pareto preface. The solution in the Pareto optimal solution set weighs the benefits between multiple objectives, making multiple objectives in a relatively optimal state. Multi-objective optimization algorithms are diverse, such as genetic algorithms, particle swarm optimization, simulated annealing, ant colony algorithms, etc. Among the multi-objective optimization algorithms, the best performance so far is the non-dominated sorting genetic algorithm (NSGA-II) proposed by Kalyanmoy Deb and other scholars in 2002. This algorithm is developed for multi-objective optimization problems. It adopts an elite retention strategy to obtain good individuals and improves the overall evolution level of the population.

[0004] In order to adapt the NSGA-II algorithm to the optimization of power electronic transformers, researchers have improved the algorithm internally to achieve higher computational efficiency and search accuracy. Patent No. CN 115310353 A improves the NSGA-II algorithm by introducing expert experience to reduce the algorithm's computational complexity and improve convergence performance. Patent No. CN 110517874 B establishes a mathematical model of the medium-frequency transformer and introduces the idea of ​​free parameter scanning into the NSGA-II algorithm, and uses the NSGA-II algorithm to replace the global free scanning process to reduce the amount of optimization calculations.

[0005] In the optimization of power electronic transformers, the NSGA-II algorithm has been widely studied due to its application in multi-objective optimization problems. However, the initial population generation of the NSGA-II algorithm in the prior art relies on randomness, resulting in instability in the efficiency and accuracy of the Pareto frontier search. In addition, the prior art lacks a further post-processing mechanism when processing the Pareto solution set, relies on the subjective choice of decision makers, and lacks objectivity. At the same time, the existing algorithm improvement methods are difficult to adapt to the needs of diverse application scenarios, and lack effective pre-processing technology in the optimization process, resulting in limited scientificity and objectivity of the optimization results. How to construct a scientific and reasonable full-process optimization method is an important technical issue in the optimization technology of power electronic transformers. Summary of the invention

[0006] In view of the shortcomings of the above-mentioned technology, the object of the present invention is to provide a power electronic transformer optimization method to solve the problems of initial population randomness and lack of post-processing technology in the existing multi-objective optimization NSGA-II algorithm.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] A power electronic transformer optimization method comprises: a pre-processing module M1, a multi-objective solution module M2, and a post-processing module M3.

[0009] Furthermore, the pre-processing module M1 adopts a free parameter scanning method, and different parameters can be selected for the following parameters of different calculation models, including the following steps:

[0010] S1.1, determine the optimization target, the optimization target is not less than 2;

[0011] S1.2, input fixed parameters, which refer to parameters that cannot be changed in the transformer system, including but not limited to working state parameters, winding parameters, core parameters, insulation distance, etc.;

[0012] S1.3, select free parameters, where the free parameters refer to variable parameters in the transformer system, including but not limited to winding parameters and core parameters;

[0013] S1.4, set the scanning range and step size of free parameters;

[0014] S1.5, create scanning schemes, including but not limited to linear mode, random mode, and programmable mode;

[0015] S1.6, setting constraints, the number of which can be any;

[0016] S1.7, scan all combinations and record the obtained target values.

[0017] Furthermore, the multi-objective solution module M2 adopts a non-dominated sorting genetic algorithm (NSGA-II), which includes the following steps:

[0018] S2.1, taking the target value obtained in step S1.6 of the pre-processing module M1 as the initial population of the NSGA-II algorithm;

[0019] S2.2, get better first generation parents by fast non-dominated sort;

[0020] S2.3, merge the parent and offspring generations into a new population;

[0021] S2.4, performing a selection operation on the merged new population. Preferably, the selection operation adopts a strategy including but not limited to a bidding competition strategy, giving priority to individuals with a lower non-dominated sorting level. If two individuals are at the same level, the individual with a larger crowding degree is selected;

[0022] S2.5, performing a crossover operation on the merged new population, preferably, the crossover operation adopts, including but not limited to, a simulated binary crossover operator;

[0023] S2.6, performing a mutation operation on the merged new population, preferably, the mutation operation uses, including but not limited to, a polynomial mutation operator;

[0024] S2.7, calculating the crowding degree. Preferably, the crowding degree indicates the sparseness of individuals around the individual. An individual with higher diversity has a greater probability of being retained and entering the next generation.

[0025] S2.8, using the elite retention strategy to retain the best individuals as the parents of the new population;

[0026] S2.9, determine whether the termination condition is met, if it is met, output the optimal solution set (i.e., the Pareto solution set), if not, return to step S2.3 and repeat steps S2.3 to S2.9.

[0027] Furthermore, the post-processing module M3 adopts an entropy weight method, comprising the following steps:

[0028] S3.1, select the Pareto solution set output by step S2.9 in the multi-objective solution module M2 as the object to be solved by the entropy weight method. Preferably, the Pareto solution set contains n objects, m indicators, x ij is the jth index value of the i-th object (i=1,2,...,n; j=1,2,...,m), z ij is the jth indicator value of the i-th object after standardization;

[0029] S3.2, determine whether the indicator orientation is consistent. If not, normalize the indicator. Preferably, the indicator orientation is usually a positive indicator. The normalization formula is:

[0030]

[0031] S3.3, calculate the proportion p of the index value of the i-th item under the j-th index ij , preferably, the calculation formula of the specific gravity is:

[0032]

[0033] S3.4, calculate the entropy value e of the jth indicator j Preferably, the entropy value is calculated as follows:

[0034]

[0035] S3.5, calculate the entropy weight ω of the jth indicator j Preferably, the calculation formula of the entropy weight is:

[0036]

[0037] S3.6, calculate the comprehensive score of each solution, the calculation formula of the comprehensive score method is:

[0038]

[0039] Optionally, if the decision maker has a preference for the target to be optimized, the preference weight α can be designed j To obtain the comprehensive weight β j Then use the comprehensive weight β j Instead of entropy weight ω j Calculate the comprehensive score of each solution, and the calculation formula of the comprehensive weight is:

[0040]

[0041] S3.7, obtain a set of optimal solutions based on the comprehensive scores calculated in S3.6.

[0042] The beneficial effects of this application are at least:

[0043] 1) Using the results of free parameter scanning as the initial population of the NSGA-II algorithm avoids the randomness of the initial population and can effectively improve the algorithm's search efficiency and accuracy for the Pareto frontier.

[0044] 2) The entropy weight method is used to post-process the Pareto solution set, which avoids the influence of non-objective factors selected by decision makers from the Pareto solution set, and can obtain a more objective and unique optimal solution.

[0045] 3) In addition to obtaining a relatively objective and unique optimal solution, this application also provides a method for decision makers to optimize according to their preferences to meet the needs of different situations. At the same time, this method avoids the impact of introducing decision maker preferences into the NSGA-II algorithm.

[0046] 4) By incorporating pre-processing modules and post-processing modules, the present application ensures that the optimization process before, during and after is not affected by uncontrollable factors such as randomness and subjectivity, thereby greatly improving the accuracy and objectivity of the optimization algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 Flowchart of the optimization method for power electronic transformers.

[0048] Figure 2 This is a flow chart of a power electronic transformer optimization method according to an embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention, so as to help those skilled in the art understand the content of the present invention. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without paying creative labor.

[0050] See also Figure 2 , which shows the process of the power electronic transformer optimization method according to an embodiment of the present application.

[0051] The present application embodiment takes a power high-frequency transformer as an example, and the design steps of the embodiment are shown in FIG\ref{fig: embodiment flow chart}. The specific embodiment design steps are as follows:

[0052] (1) Pre-processing module

[0053] S1.1, the initial structure of the high-frequency transformer has been determined, and the loss and efficiency are established as the optimization goals, and the temperature rise is used as the constraint condition.

[0054] S1.2, determine the fixed parameters, which refer to the parameters that cannot be changed in the transformer system, including: temperature rise coefficient, rated power, apparent power, primary winding voltage, secondary winding voltage, duty cycle, operating frequency, magnetic loss empirical coefficient, winding turn length, wire resistivity, core cross-sectional area, window area, current density, waveform factor, and current proportional density.

[0055] S1.3, determine free parameters, which refer to variable parameters in the transformer system, including: number of turns of the primary winding, number of turns of the secondary winding, maximum magnetic induction intensity, core cross-sectional area, primary winding wire diameter, and secondary winding wire diameter.

[0056] S1.4, sets the step length of the scannable free parameters.

[0057] S1.5, select the linear scanning mode, which refers to the linear change of all selected parameters between the minimum value and the maximum value.

[0058] S1.6, calculate the temperature rise and determine whether the temperature rise meets the limit temperature rise. The temperature rise is calculated using the engineering transformer temperature rise calculation method:

[0059]

[0060] In the formula, K s is the temperature rise coefficient, which depends on the magnetic core, P is the total loss of the transformer, A p Value of the area product method.

[0061] S1.7, calculate the losses and efficiencies when the temperature rise constraints are met, including:

[0062] 1) The loss calculation formula is:

[0063]

[0064] In the formula, R i ,I i , U i , i , L i They are the AC resistance coefficient of the transformer primary and secondary windings, the effective value of the winding current, the rated voltage, the resistivity, and the average winding turn length; K m , α, β are the core loss characteristic parameters, B m is the peak value of the working magnetic flux density, f is the working excitation frequency, A e is the effective cross-sectional area of ​​the core.

[0065] 2) The efficiency calculation formula is:

[0066]

[0067] In the formula, P, P N They are total loss and rated power respectively.

[0068] S1.8, scan all combinations and record all loss and efficiency values.

[0069] (2) Multi-objective solution module

[0070] S2.1, the target value obtained in step S1.8 of the pre-processing module M1 is used as the initial population of the NSGA-II algorithm.

[0071] S2.2, get better first-generation parents through fast non-dominated sort.

[0072] S2.3, merge the parent and offspring generations into a new population.

[0073] S2.4, performing a selection operation on the merged new population. Preferably, the selection operation adopts a bidding strategy, giving priority to individuals with lower non-dominated sorting levels. If two individuals are at the same level, the individual with a larger crowding degree is selected.

[0074] S2.5, performing a crossover operation on the merged new population. Preferably, the crossover operation uses a simulated binary crossover operator.

[0075] S2.6, performing a mutation operation on the merged new population. Preferably, the mutation operation uses a polynomial mutation operator.

[0076] S2.7, calculate the crowding degree. Preferably, the greater the crowding degree, the sparser the individuals around the individual, the higher the diversity, and the greater the probability of being retained and entering the next generation.

[0077] S2.8, adopt the elite retention strategy to retain the best individuals as the parents of the new population.

[0078] S2.9, determine whether the termination condition is met, if it is met, output the optimal solution set (i.e., the Pareto solution set), if not, return to step S2.3 and repeat steps S2.3 to S2.9.

[0079] (3) Post-processing module

[0080] S3.1, select the Pareto solution set output by step S2.9 in the multi-objective solution module M2 as the object to be solved by the entropy weight method. Preferably, the Pareto solution set contains n objects, m indicators, x ijis the jth index value of the i-th object (i=1,2,...,n; j=1,2,...,m), z ij is the jth indicator value of the oth object after standardization.

[0081] S3.2, determine whether the indicator orientation is consistent. If not, normalize the indicator. Preferably, the indicator orientation is usually a positive indicator. The normalization formula is:

[0082]

[0083] S3.3, calculate the proportion p of the index value of the i-th item under the j-th index ij , preferably, the calculation formula of the specific gravity is:

[0084]

[0085] S3.4, calculate the entropy value e of the jth indicator j Preferably, the entropy value is calculated as follows:

[0086]

[0087] S3.5, calculate the entropy weight ω of the jth indicator j Preferably, the calculation formula of the entropy weight is:

[0088]

[0089] S3.6, calculate the comprehensive score of each solution, the calculation formula of the comprehensive score method is:

[0090]

[0091] Optionally, if the decision maker has a preference for the target to be optimized, the preference weight α can be designed j To obtain the comprehensive weight β j Then use the comprehensive weight β j Instead of entropy weight ω j Calculate the comprehensive score of each solution, and the calculation formula of the comprehensive weight is:

[0092]

[0093] S3.7, based on the comprehensive score calculated in S3.6, obtain a set of solutions corresponding to the free parameters under the conditions of optimal efficiency and volume.

[0094] It should be noted that the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or apparatus.

[0095] The above is only a specific implementation of the present application, not all embodiments, so the protection scope of the present application is not limited thereto. All other embodiments obtained by ordinary technicians in this field without creative work, any modifications, equivalent substitutions and improvements within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for optimizing a power electronic transformer, characterized in that: include: Pre-processing module, multi-objective solution module, and post-processing module.

2. A power electronic transformer optimization method according to claim 1, characterized in that: The pre-processing module M1 adopts the free parameter scanning method; the multi-objective solution module M2 adopts the non-dominated sorting genetic algorithm (NSGA-II); and the post-processing module M3 adopts the entropy weight method.

3. The free parameter scanning method according to claim 2, characterized in that: include: (1) The number of optimization objectives shall be no less than 2; (2) Fixed parameters refer to parameters that cannot be changed in the transformer system; (2) Free parameters refer to the variable parameters in the transformer system; (3) The scanning schemes are diverse, including but not limited to linear mode, random mode, and programmable mode; (4) The number of constraints can be arbitrary.

4. The NSGA-II algorithm according to claim 2, characterized in that: include: (1) Using the target value obtained in claim 3 as the initial population of the non-dominated sorting genetic algorithm; (2) The selection operation adopts a strategy including but not limited to a bidding competition, giving priority to individuals with lower non-dominated sorting levels. If two individuals are at the same level, the one with greater crowding degree is selected; (3) the crossover operation uses, including but not limited to, a simulated binary crossover operator; (4) the mutation operation uses, including but not limited to, polynomial mutation operators; (5) Crowding indicates the sparseness of individuals around the individual. An individual with higher diversity has a greater probability of being retained and entering the next generation.

5. The entropy weight method according to claim 2, characterized in that: include: (1) Using the Pareto solution set calculated in claim 4 as the object to be solved by the entropy weight method; (2) Two methods are provided: objective comprehensive scoring and decision maker preference weighting and scoring.

6. The decision maker preference weighting and scoring according to claim 5, characterized in that: The formula for calculating the comprehensive weight of decision maker preference weighting and scoring is: In the formula, α j is the preference weight; β j is the comprehensive weight; ω j is the entropy weight; ij is the jth indicator value of the ith object after standardization.

Images