Automatic design method and device for circuit parameters, equipment and storage medium

Abstract

The invention provides an automatic design method and device for circuit parameters, equipment and a storage medium. The automatic design method comprises the steps that circuit design parameter specifications and performance parameter specifications are determined; creating a circuit design parameter lookup table; constructing a data set through circuit simulation software according to the circuit design parameters of the power converter and the performance parameters obtained through simulation; a PINN-AM model is constructed; wherein the PINN-AM model is obtained by introducing an attention mechanism into a PINN model; training the PINN-AM model to obtain a power converter performance prediction model; based on the converter performance prediction model, circuit parameters of the power converter are optimized by adopting an optimization algorithm, and all circuit parameters which meet the circuit design parameter specifications, the performance parameter specifications and the circuit design parameter lookup table are combined and output to serve as a target circuit parameter set; and with the cost as a target, selecting an optimal circuit parameter from the target circuit parameter set as a target design parameter.

Description

Technical Field

[0001] The present invention belongs to the field of circuit design, and in particular relates to a method and device, equipment and storage medium for automatically designing circuit parameters. Background Art

[0002] Power converters, such as two-level inverters and DC-DC converters, regulate the conversion of electrical energy and are widely used in industry and everyday life. These applications include energy storage, electric vehicles, and renewable energy systems. The conversion efficiency and output distortion of power converters, as important performance indicators, are closely related to the parameter design of their key components. Typically, to ensure excellent overall performance of a power converter, multiple design objectives (such as efficiency, distortion, and size) must be considered simultaneously. However, these different design objectives are often conflicting. Therefore, optimizing the key parameter design of power converters is both important and challenging.

[0003] Currently, there are three main types of optimization design methods for key parameters of power converters: traditional manual design (HD) methods, computer-aided optimization design (CAO-D) methods, and artificial intelligence-based design methods (AI-D). These optimization design methods generally include two processes: analysis and derivation and optimization. The analysis and derivation process aims to obtain mathematical expressions for the optimization objectives and design constraints. Then, without violating the design constraints, the key parameters for achieving optimal performance are determined during the optimization process. In traditional HD, engineers are responsible for the entire analysis, derivation, and optimization process. To simplify the analysis, many approximations are usually used when deriving the mathematical expressions for the optimization objectives and design constraints. Overall, the main disadvantages of traditional HD methods are heavy workload, time consumption, and low analysis and derivation accuracy due to approximations.

[0004] CAO-D typically automates the optimization process using optimization algorithms such as particle swarm optimization (PSO) and genetic algorithms (GA), reducing the complexity of power converter parameter optimization design. While CAO-D frees engineers from the repetitive trial-and-error process, it still requires human intervention during the analysis and derivation process. Obtaining mathematical expressions for the optimization objective and design constraints requires complex and time-consuming analysis, which also suffers from accuracy issues.

[0005] In recent years, AI-D has gradually emerged as a means to completely liberate the human resources required for the analysis, derivation, and optimization processes in converter parameter optimization design. Based on a large amount of historical data or simulation data, this method uses AI to derive the relationship between design parameters and performance indicators and generate a performance model for the converter. An optimization algorithm is then used to automatically adjust the design parameters to obtain a parameter combination that achieves the best overall converter performance. It is capable of handling complex nonlinear and multidimensional optimization problems with a high degree of automation, making it particularly suitable for systems with complex performance indicators and constraints. However, it is worth noting that AI-D is highly dependent on converter operating data. Once the data has problems such as noise and errors, the converter modeling accuracy will be reduced, which will directly affect the effectiveness of the design work. In addition, AI-D is still in its emerging stage, and there is very little related research literature.

[0006] Therefore, how to improve the automation and intelligence of power converter circuit parameter design methods and how to ensure that the optimal circuit parameter combination can be obtained have become urgent tasks. Summary of the Invention

[0007] The object of the present invention is to provide a method and device for automatic design of circuit parameters, so as to realize the automatic design of circuit parameters and ensure the availability and accuracy of the designed circuit parameters.

[0008] In a first aspect, a method for automatically designing circuit parameters is provided, comprising: Determine circuit design parameter specifications and performance parameter specifications; Creating a circuit design parameter lookup table for storing available circuit design parameters; Using circuit simulation software, a data set is constructed based on the circuit design parameters of the power converter and the performance parameters obtained by simulation; Construct a PINN-AM model; the PINN-AM model is obtained by introducing the attention mechanism into the PINN model; The PINN-AM model is trained to obtain a power converter performance prediction model; Based on the converter performance prediction model, an optimization algorithm is used to optimize the circuit parameters of the power converter. All circuit parameters that meet the circuit design parameter specifications, performance parameter specifications, and the circuit design parameter lookup table are combined and output as the target circuit parameter set. Taking cost as a target, the optimal circuit parameters are selected from the target circuit parameter set as target design parameters.

[0009] Optionally, the circuit design parameter lookup table includes: Inductance lookup table and capacitance lookup table; the inductance lookup table includes all available inductors arranged by inductance size ; The capacitance lookup table includes all available capacitors arranged by capacitance size .

[0010] Optionally, the step of constructing a data set based on circuit design parameters of the power converter and performance parameters obtained through simulation using circuit simulation software includes: Based on the circuit design parameter specification, generate multiple groups of initial circuit design parameter combinations; According to the initial circuit design parameter combination, the power converter is simulated using simulation software to obtain the initial performance parameter combination of the power converter; wherein the initial performance parameters include efficiency and total harmonic distortion Total harmonic distortion The efficiency is obtained from the simulation software. Calculated according to the following formula:

[0011]

[0012] in, represents the input power of the power converter, represents the power loss of the power converter, represents the power loss caused by the capacitance of the power converter, Indicates the power loss caused by inductance, Indicates the power loss caused by the switching tube.

[0013] Optionally, the loss function of the PINN-AM model is as follows:

[0014]

[0015]

[0016] in, represents the total loss, represents the loss of the data-driven part, represents the loss of the driving part of the physical model, , Indicates the The time point of the group input parameters, represents the parameters of the estimated PINN-AM model output, represents the parameters of the real PINN-AM model output, Indicates the input Group parameters, Indicates the number of input parameter combinations, represents the attention weight, represents the input power of the circuit, Indicates the output power of the circuit.

[0017] Optionally, during the training of the PINN-AM model, the prediction accuracy of the model is used as the optimization target to dynamically adjust the attention weights. .

[0018] Optionally, based on the converter performance prediction model, an optimization algorithm is used to optimize the circuit design parameters of the power converter, and all circuit parameter combinations that meet the circuit design parameter specification, the power converter performance parameter specification, and the circuit design parameter lookup table are output as a target circuit parameter set, including the following steps: Taking the performance parameters of the circuit as the optimization target, optimizing the circuit parameters using an optimization algorithm, and taking all the circuit parameter combinations that meet the performance parameter specifications as the first circuit parameter set; Replacing the circuit parameters in each circuit parameter combination in the first circuit parameter set according to the circuit design parameter lookup table to obtain a second circuit parameter set; The second circuit parameter set is screened using the circuit design parameter specification to obtain a target circuit parameter set.

[0019] Optionally, the step of replacing the circuit parameters in each circuit parameter combination in the first circuit parameter set according to the circuit parameter lookup table to obtain the second circuit parameter set includes: Based on the available inductance in the inductance lookup table and capacitor lookup table for available capacitors Inductance in circuit parameter combination and capacitors Replace; if , then select As ,like , then select As ,in, is a coefficient with a value range of 0.85~0.95; if , then select As .

[0020] In a second aspect, a device for automatically designing circuit parameters is provided, comprising: A determination module, used to determine circuit design parameter specifications and performance parameter specifications; A lookup table creation module, used for creating a circuit design parameter lookup table for storing available circuit design parameters; A data set construction module is used to construct a data set based on circuit design parameters of the power converter and performance parameters obtained by simulation using circuit simulation software; The model construction module is used to build the PINN-AM model; the PINN-AM model is obtained by introducing the attention mechanism into the PINN model; A training module is used to train the PINN-AM model to obtain a power converter performance prediction model; A first parameter determination module is configured to optimize the circuit parameters of the power converter using an optimization algorithm based on the converter performance prediction model, and output a combination of all circuit parameters that meet the circuit design parameter specification, the performance parameter specification, and the circuit design parameter lookup table as a target circuit parameter set; The second parameter determination module takes cost as a target and selects the optimal circuit parameters from the target circuit parameter set as target design parameters.

[0021] According to a third aspect, an electronic device is provided, comprising the automatic circuit parameter design device as described above.

[0022] In a fourth aspect, a computer-readable storage medium is provided, wherein at least one program code is stored in the computer-readable storage medium, and the program code is executed by a processor to implement the automatic design method of circuit parameters as described in any one of the above items.

[0023] The technical solution provided by the present invention has the following unexpected technical effects: This paper investigates an automated design method based on a physics-informed neural network-attentional mechanism (PINN-AM)-PSO (Physics-Informed Neural Network-attentional mechanism, or PINN-AM-PSO-D). By adding AM to the PINN physical loss function and data loss function, the attention weights are automatically adjusted based on the model's performance. The model automatically emphasizes the impact of physical loss when faced with uncertainty or data scarcity, or prioritizes data loss when data is abundant. Therefore, PINN-AM overcomes the challenges of low data quality that AI-D may face, enabling high-precision intelligent modeling of power converters. Using the meta-heuristic optimization algorithm PSO, circuit parameter combinations are automatically optimized. Ultimately, the power converter parameter design achieves precise optimal design results, achieving a highly automated and intelligent optimization process. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 A flow chart of an automatic circuit parameter design method provided by the present invention; Figure 2 A schematic diagram of an inductance lookup table provided by the present invention; Figure 3 A schematic diagram of a capacitance lookup table provided by the present invention; Figure 4 A flow chart of a particle swarm optimization algorithm provided by the present invention; Figure 5 A flow chart of another automatic circuit parameter design method provided by the present invention; Figure 6 A schematic diagram of a circuit structure provided by the present invention; Figure 7 A schematic diagram of a circuit design specification and performance specification provided by the present invention; Figure 8 A schematic diagram of another inductance lookup table provided by the present invention; Figure 9 A schematic diagram of another capacitance lookup table provided by the present invention; Figure 10 A schematic diagram of the PINN-AM model training process provided by the present invention; Figure 11 A schematic diagram of target design parameters provided by the present invention; Figure 12 A structural block diagram of an automatic circuit parameter design device provided by the present invention; Figure 13 This is a structural block diagram of an electronic device provided by the present invention.

[0026] The reference numerals are as follows: 11: Determination module; 12: Lookup table creation module; 13: Data set construction module; 14: Model construction module; 15: Training module; 16: First parameter determination module; 17: Second parameter determination module; 21: Processor; 22: Memory. DETAILED DESCRIPTION

[0027] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0028] Based on the above considerations, and to address the difficulties faced by existing power converter parameter design methods, this paper investigates an automated design method based on a physics-informed neural network-attentional mechanism (PINN-AM)-PSO (PINN-AM-PSO-D). By adding AM to the PINN physical loss function and data loss function, the attention weights are automatically adjusted based on the model's performance. The model automatically emphasizes the impact of physical losses when faced with uncertainty or data scarcity, or prioritizes data losses when data is abundant. Therefore, PINN-AM overcomes the difficulty of low data quality that AI-D may face, enabling high-precision intelligent modeling of power converters. Using the meta-heuristic optimization algorithm PSO, circuit parameter combinations are automatically selected. Ultimately, the power converter parameter design achieves precise optimal design results, achieving a highly automated and intelligent optimization process.

[0029] The main contributions of this paper include: First, a PINN-AM-based converter performance estimation model is proposed. After training with a limited number of design parameters and performance indicators, the performance of new design parameter combinations is effectively predicted, automating the analytical derivation process in parameter design. Second, an optimization algorithm is applied with converter performance as the optimization objective and cost as the design constraint to automate the design parameter combination. Third, a PINN-AM-PSO-based automatic power converter circuit parameter design method is proposed. This method maximizes the compromise between optimization objectives without breaking the design parameter range, and is characterized by high accuracy and ease of implementation. Fourth, the proposed PINN-AM-PSO-D method can be used for circuit parameter design of various power converters, regardless of the application context. With only some customization, the proposed PINN-AM-PSO-D method remains applicable, achieving a highly automated and intelligent parameter design optimization process.

[0030] Figure 1 This is a flow chart of the automatic design method of circuit parameters provided by the present invention. Figure 1 , the method steps include: S101. Determine circuit design parameter specifications and performance parameter specifications.

[0031] In one example, the circuit parameter design specification includes capacitance ( ) design specifications, inductance ( ) design specifications, switching frequency ( ) design specifications. Among them, , , .

[0032] In another example, the capacitance and inductance in the circuit parameter design specification can be converted into the design parameters of the components. For example, the inductance can be converted into the design parameters of the magnetic core (including the geometric characteristics and magnetic characteristics of the magnetic core).

[0033] In one example, performance parameter specifications include: efficiency and total harmonic distortion ( ).in, , .

[0034] S102: Create a circuit design parameter lookup table for storing available circuit design parameters.

[0035] In one example, the circuit design parameter lookup table includes: Inductance lookup table and capacitance lookup table; the inductance lookup table includes all available inductors arranged by inductance size ; The capacitance lookup table includes all available capacitors arranged by capacitance size .

[0036] In another example, the inductance lookup table and the capacitance lookup table may be replaced by a design parameter lookup table of components, which is not limited in this disclosure.

[0037] Figure 2 An example of an inductance lookup table provided by the present invention is Figure 2 In the case of a magnetic core, the size of the inductance is determined by the geometric characteristics of the magnetic core (the inner diameter of the toroidal core). , the height of the toroidal core ), and magnetic characteristics (window utilization coefficient , inductance factor ) determines the range of inductance that can be used, and lists the range of inductance in the inductance lookup table, that is, the minimum inductance value corresponding to the geometric characteristics and magnetic characteristics of different magnetic cores. and the maximum inductance value that can be used ,in , Indicates the maximum inductance value corresponding to different geometric characteristics and magnetic characteristics. In the inductance lookup table, inductors with different inductance ranges are arranged in order from top to bottom.

[0038] The inductor lookup table will be the inductor The value of is mapped to the geometric and magnetic characteristics of the selected core. Taking a toroidal core as an example, the process of creating an inductance lookup table is as follows: Figure 2 The inductance lookup table contains inductors for different core materials and geometric characteristics. After determining the inductor core filling material and window utilization factor, the core geometric characteristics and magnetic characteristics obtained from the core database are obtained.

[0039] Furthermore, the maximum number of turns that can be achieved for each core is determined according to the following formula: and maximum inductance :

[0040]

[0041] in, is the number of coil turns; is the window utilization coefficient, which indicates the filling efficiency of the winding in the window; is the inner diameter of the annular core; is the cross-sectional area of the conductor; The height of the toroidal core; is the inductance factor. These parameters are often obtained from a magnetic core database.

[0042] Afterwards, To avoid inductor overload, use 0.9 The choice of inductor core is divided. For example, if Less than or equal to 0.9 , select core 1. If Greater than 0.9 Select core 2. Here Refers to the design parameters determined later ,Since the capacitance in the capacitance lookup table has multiple ranges, it is necessary to ,further determine the capacitance selection conditions and further adjust and ,screen the circuit design parameters determined by the model.

[0043] It should be noted that in the inductance lookup table, the core The inductance is less than the magnetic inductance.

[0044] Figure 3 An example of a capacitance lookup table provided by the present invention is Figure 3 In the capacitance lookup table, different specifications and characteristics of capacitors are included. Then, sort the capacitors in ascending order based on capacitance value. According to the principle of upward selection, the selection of capacitors is divided according to the size of the capacitance value. For example, if ,choose At the same time, this article connects capacitors in series and in parallel to obtain the available values of capacitors. The process of creating a capacitor lookup table is as follows: Figure 3 shown.

[0045] S103 , using circuit simulation software, constructing a data set based on the circuit design parameters of the power converter and the performance parameters obtained by simulation.

[0046] In one example, step S103 includes: In the first step, multiple groups of initial circuit design parameter combinations are generated based on the circuit design parameter specifications.

[0047] Among them, the design parameters are limited to the range 、 and Divided into p, q, and n parts respectively, the generated combination The total number is .

[0048] It should be pointed out that although there are some ways to achieve the same as possible (series and parallel connection, changing the core and number of turns), 、 There are more useful values, but the generated combinations should not be divided too much. This is because too many combinations will generate too much meaningless automation simulation time in the subsequent simulation steps.

[0049] also, It is also not advisable to divide too much. This is because in engineering applications The choice is often an integer, too many Samples also bring meaningless simulation time.

[0050] The second step is to simulate the power converter using simulation software based on the initial circuit design parameter combination to obtain the initial performance parameter combination of the power converter; the initial performance parameters include efficiency and total harmonic distortion Total harmonic distortion The efficiency is obtained from the simulation software. Calculated according to the following formula:

[0051]

[0052] in, represents the input power of the power converter, represents the power loss of the power converter, represents the power loss caused by the capacitance of the power converter, Indicates the power loss caused by inductance, Indicates the power loss caused by the switching tube.

[0053] against Group design parameters Combined, establish the converter Python-Plecs joint automation simulation model (also known as the simulation software in the present invention), calculate the converter and performance and automatically record data samples.

[0054] In the present invention, the switching loss Using the power switch Plesc model provided by the manufacturer, Plesc can evaluate the switch tube loss, including switching loss and conduction loss, taking into account thermal effects.

[0055] For capacitance loss , considering high frequency applications Self-equivalent series resistance and Equivalent series inductance is generated. According to the capacitor lookup table, and It can be calculated by the following formulas:

[0056]

[0057] in, and is the loss tangent and The factors are obtained from the capacitor database provided by the manufacturer. and Also included is the capacitance lookup table established in .

[0058] For inductor losses , which is caused by copper loss and magnetic loss Copper loss is usually composed of two parts. It can be calculated by the following formula:

[0059] in, is the RMS value of the inductor peak current, is the DC resistance of the inductor. It is related to the number of coil turns and the copper wire diameter, and is obtained from the database provided by the manufacturer.

[0060] Magnetic loss It can be obtained by Ansys simulation. However, for complex converter optimization design problems, although Python-Ansys joint automated simulation can be implemented, it is very time-consuming to use Ansys simulation to calculate the magnetic loss of a large number of inductor samples. Therefore, the present invention uses the following formula to estimate magnetic loss. In theory, this method will introduce certain design errors, but it greatly reduces the calculation cost and method complexity:

[0061]

[0062]

[0063]

[0064] in, is the inductor peak current; is the frequency; is the inductor flux density; is the magnetic permeability of the core; is the magnetizing force; is the number of coil turns; is the effective magnetic circuit length; is the outer diameter of the core; is the inner diameter of the core; 、 、 is the core material constant.

[0065] It should be noted that different magnetic cores have different material properties and geometric characteristics. Therefore, the data related to the magnetic core, such as 、 、 、 、 、 , obtained from the inductance database provided by the manufacturer.

[0066] In this process, Group design parameters ,Through the appropriate interface of the Plesc simulation tool, the ,programming language Python is used to automatically adjust the parameters and ,automate the simulation running and data recording.

[0067] S104. Construct a PINN-AM model; wherein the PINN-AM model is obtained by introducing an attention mechanism into the PINN model.

[0068] To address the problem of converter parameter optimization design, this application adds an attention mechanism to the PINN-AM loss function. By automatically adjusting attention weights, the impact of physical losses is automatically enhanced, or data losses are prioritized when data is abundant. This makes the model more flexible and adaptable, thereby improving the modeling accuracy of the method in different application scenarios.

[0069] The loss function of the PINN-AM model is as follows:

[0070]

[0071]

[0072] in, represents the total loss, represents the loss of the data-driven part, represents the loss of the driving part of the physical model, , Indicates the The time point of the group input parameters, represents the parameters of the estimated PINN-AM model output, represents the parameters of the real PINN-AM model output, Indicates the input Group parameters, Indicates the number of input parameter combinations, represents the attention weight, represents the input power of the circuit, Indicates the output power of the circuit.

[0073] Therefore, based on the normalized m sets of design parameters and optimization objectives, they are divided into a training set (70%) and a test set (30%). Assume that the normalized design parameters and optimization objectives are . Then the training set, validation set, and test set can be expressed as: Training set:

[0074] Validation set:

[0075] Test set:

[0076] in, , Indicates the Group design parameters The efficiency and total harmonic distortion of the converter are also considered. and When mapping the performance model, is the data sample, is the model target.

[0077] During the training process, based on the training set, the attention mechanism is used to automatically try different option and cross-validate the PINN-AM model in new unknown design parameter combinations to obtain the minimum error , which is the required attention weight for the PINN-AM loss function. The test set is the attention weight 𝛼 selected for secondary validation. Due to the large data sample size, this paper further employs L1 regularization and dropout to prevent overfitting during training. Furthermore, the physical information constraint also prevents overfitting to a certain extent.

[0078] In this embodiment, during the training of the PINN-AM model, the prediction accuracy of the model is used as the optimization target, and the attention weight is dynamically adjusted. .

[0079] It should be noted that although only m sets of different design parameters are used and optimization goals Training converter and Performance mapping model, but the data driven nature can obtain all design parameters when the converter is working and performance estimation model. This means that any A collection of design parameter combinations , the trained converter equivalent models can estimate the converter’s and performance.

[0080] S105 : Train the PINN-AM model to obtain a power converter performance prediction model.

[0081] S106. Based on the converter performance prediction model, an optimization algorithm is used to optimize the circuit parameters of the power converter, and all circuit parameter combinations that meet the circuit design parameter specifications, performance parameter specifications, and the circuit design parameter lookup table are output as a target circuit parameter set.

[0082] In one example, step S106 includes: Step 1: Taking the performance parameters of the circuit as the optimization target, the circuit parameters are optimized using an optimization algorithm, and all circuit parameter combinations that meet the performance parameter specifications are obtained as the first circuit parameter set.

[0083] The optimization algorithm is a particle swarm optimization (PSO) algorithm, a genetic algorithm, or any other optimization algorithm.

[0084] See also Figure 4 In one example, when using the particle swarm optimization algorithm, step 1 includes: The first step is to randomly generate N particles, each particle represents a set of design parameters .

[0085] The second step is to build a converter based on the third stage and Estimation model, using the design parameters for each particle to simulate the converter, estimate the converter and performance.

[0086] The third step is to design the fitness function according to the constraints. The fitness function is the current converter and The distance between performance and constraints.

[0087] The calculation method is:

[0088] The fourth step is to update the individual and global optimal solutions. Traverse each particle, if the fitness of the current particle is better than , then update If the fitness of the current particle is better than the global optimal , then update .in, For particles Design parameters with optimal fitness. is the design parameter with the best fitness among all particles.

[0089] The fifth step is to update the design parameter combination represented by each particle.

[0090] Step 6: If the fitness change is less than the set threshold of 0.01%, the iteration is stopped.

[0091] Step 7: Output the design parameter combination that meets the conditions and its corresponding fitness value and .

[0092] Step 2: Replace the circuit parameters in each circuit parameter combination in the first circuit parameter set according to the circuit design parameter lookup table to obtain a second circuit parameter set.

[0093] In one example, step 2 includes: Based on the available inductance in the inductance lookup table and capacitor lookup table for available capacitors Inductance in circuit parameter combination and capacitors Replace; if , then select As ,like , then select As ,in, is a coefficient with a value range of 0.85~0.95; if , then select As .

[0094] Step 3: Screen the second circuit parameter set using the circuit design parameter specification to obtain a target circuit parameter set.

[0095] S107 , taking cost as a target, selecting optimal circuit parameters from the target circuit parameter set as target design parameters.

[0096] In one example, step S107 includes: Considering the cost of capacitors and inductors, and taking minimum cost as the constraint, select the optimal design parameter combination.

[0097]

[0098] in, represents the cost of the inductor, Represents the cost of the capacitor.

[0099] See Figure 5 , Figure 5 This is a flow chart of another automatic circuit parameter design method provided by the present invention.

[0100] See Figures 6 to 11 , the present invention provides specific embodiments: See also Figure 6 , in the user side energy storage converter, the following is often used Figure 6 In the present invention, the design process takes efficiency and total harmonic distortion as optimization targets and cost as design constraint to carry out the circuit parameter optimization design of the user-side two-level energy storage converter.

[0101] The specific optimization process is as follows: Phase 1: Determine design specifications and goals The design conditions and requirements for the circuit parameter design of the two-level energy storage converter on the user side are as follows: Figure 7 As shown in Figure 2, a SiC MOSFET from Rohm was chosen for the power switching device, given the given operating parameters. If other power switches were considered, the proposed PINN-AM-PSO-D method would still be applicable.

[0102] Phase 2: Creating Capacitance and Inductance Lookup Tables For the inductors, four toroidal inductors with Sendust core material (part numbers: NPS130075, NPS157075, NPS184060, and NPS226060) produced by POCO were used. A lookup table for the inductors was created, such as Figure 8 shown. Figure 8 Information is provided on the actual geometrical and magnetic characteristics of the selected cores.

[0103] For capacitors, this design uses electrolytic capacitors produced by Rubycon. Research all electrolytic capacitors in the range of [10uF 100uF] and [200uF 2000uF] with a rated voltage of 500V from Rubycon and create the following Figure 9 A lookup table for capacitors is shown.

[0104] Phase 3: Establishing a converter performance estimation model Step 1: Choose the right 、 、 、 combination.

[0105] Based on reservation 、 、 、 Range 、 、 and , dividing it into 20, 10, 10 and 20 parts respectively, thus generating a total of 20×10×10×20=40000 combinations of design parameters.

[0106] Step 2: Expand the converter and Performance Automation Simulation Based on all the design parameter combinations, a joint automated simulation model is established in Python-Plecs. The converter generated by the design parameter combination is simulated and evaluated. and performance, and automatically record all data samples . Then, it is used as the training data of the PINN-AM model. It should be pointed out that the automated simulation calculation The switch tube PLECS loss model provided by the manufacturer was adopted.

[0107] Step 3: Build data-driven and Estimation model First, normalize the 40,000 recorded groups Then, it is divided into training set (70%) and test set (30%). and Estimation model. The model training process is as follows Figure 10 It should be noted that although only 40,000 sets of design samples were used, the data-driven nature enabled us to obtain all possible combinations of converters within the limited range of design parameters. and Performance estimation model. This means that the trained converter equivalent model can estimate the converter performance at this time. and performance.

[0108] exist Figure 10 In the figure, we can see that for any combination of design parameters, PINN-AM can well estimate the converter performance ( and The training accuracy of the model increases rapidly with the number of iterations, approaching 0.98 after about 50 iterations, and finally reaching 0.9981 at 256 iterations, demonstrating high accuracy. The training loss shows a rapid downward trend, gradually stabilizing after 50 iterations, and finally reaching a low of 0.0306 at 256 iterations, indicating good model convergence. In addition, in the later iterations, the training accuracy and loss values change little, indicating that the model training process has strong stability. Overall, the model performs well in terms of convergence speed and accuracy, and is suitable for application scenarios of high-precision converter performance estimation.

[0109] Phase 4: PSO selects the best design parameter combination .

[0110] Based on the converter performance estimation model trained using PINN-AM and the capacitance and inductance lookup tables, the PSO algorithm is applied to search for the optimal 、 、 、 Parameter combination. Internal Search , from the Select from the inductor lookup table within , and from 、 Select the capacitor in the lookup table and Combined with the design constraint of minimum cost, the optimal design parameters are =941uF, =28.5uF, =182uH and = 37.37kHz. They provide the best converter performance =99.04%, =0.15%. Figure 11 The design results of the user-side two-level energy storage converter using the proposed PINN-AM-PSO-D method are summarized.

[0111] Figure 12 This is a structural block diagram of an automatic circuit parameter design device provided by the present invention. Figure 12 ,include: Determination module 11, used to determine circuit design parameter specifications and performance parameter specifications; A lookup table creation module 12 is used to create a circuit design parameter lookup table for storing available circuit design parameters; A data set construction module 13 is used to construct a data set based on the circuit design parameters of the power converter and the performance parameters obtained by simulation using circuit simulation software; A model construction module 14 is used to construct a PINN-AM model; wherein the PINN-AM model is obtained by introducing an attention mechanism into the PINN model; A training module 15 is used to train the PINN-AM model to obtain a power converter performance prediction model; A first parameter determination module 16 is configured to optimize the circuit parameters of the power converter using an optimization algorithm based on the converter performance prediction model, and output a combination of all circuit parameters that meet the circuit design parameter specification, the performance parameter specification, and the circuit design parameter lookup table as a target circuit parameter set; The second parameter determination module 17 takes cost as a target and selects the optimal circuit parameters from the target circuit parameter set as target design parameters.

[0112] Figure 13 This is a structural block diagram of an electronic device provided by the present invention. Figure 13 , Figure 13 This is a structural block diagram of an electronic device provided by the present invention. Figure 13 , electronic devices may include Figure 12 The automatic circuit parameter design device generally includes a processor 21 and a memory 22 .

[0113] Processor 21 may include one or more processing cores, such as a quad-core processor or an octa-core processor. Processor 21 may be implemented in hardware using at least one of the following: a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), or a PLA (Programmable Logic Array). Processor 21 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. Memory 22 may include one or more computer-readable storage media, which may be non-transitory. Memory 22 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in memory 22 is used to store at least one instruction, which is executed by processor 21 to implement the automatic circuit parameter design method performed by an electronic device provided in the method embodiments of this application.

[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for automatic design of circuit parameters, characterized in that: include: Determine circuit design parameter specifications and performance parameter specifications; Creating a circuit design parameter lookup table for storing available circuit design parameters; Using circuit simulation software, a data set is constructed based on the circuit design parameters of the power converter and the performance parameters obtained by simulation; Construct a PINN-AM model; the PINN-AM model is obtained by introducing the attention mechanism into the PINN model; The PINN-AM model is trained to obtain a power converter performance prediction model; Based on the converter performance prediction model, an optimization algorithm is used to optimize the circuit parameters of the power converter. All circuit parameters that meet the circuit design parameter specifications, performance parameter specifications, and the circuit design parameter lookup table are combined and output as the target circuit parameter set. Taking cost as a target, the optimal circuit parameters are selected from the target circuit parameter set as target design parameters.

2. The automatic circuit parameter design method according to claim 1, characterized in that: The circuit design parameter lookup table includes: Inductance lookup table and capacitance lookup table; the inductance lookup table includes all available inductors arranged by inductance size ; The capacitance lookup table includes all available capacitors arranged by capacitance size .

3. The automatic circuit parameter design method according to claim 1, characterized in that: The steps for constructing a data set based on the circuit design parameters and simulated performance parameters of the power converter using circuit simulation software include: Based on the circuit design parameter specification, generate multiple groups of initial circuit design parameter combinations; According to the initial circuit design parameter combination, the power converter is simulated using simulation software to obtain the initial performance parameter combination of the power converter; wherein the initial performance parameters include efficiency and total harmonic distortion Total harmonic distortion The efficiency is obtained from the simulation software. Calculated according to the following formula: in, represents the input power of the power converter, represents the power loss of the power converter, represents the power loss caused by the capacitance of the power converter, Indicates the power loss caused by inductance, Indicates the power loss caused by the switching tube.

4. The automatic circuit parameter design method according to any one of claims 1 to 3, characterized in that: The loss function of the PINN-AM model is as follows: in, represents the total loss, represents the loss of the data-driven part, represents the loss of the driving part of the physical model, , Indicates the The time point of the group input parameters, represents the parameters of the estimated PINN-AM model output, represents the parameters of the real PINN-AM model output, Indicates the input Group parameters, Indicates the number of input parameter combinations, represents the attention weight, represents the input power of the circuit, Indicates the output power of the circuit.

5. The automatic circuit parameter design method according to claim 4, characterized in that: During the training of the PINN-AM model, the prediction accuracy of the model is used as the optimization target, and the attention weight is dynamically adjusted. .

6. The automatic circuit parameter design method according to any one of claims 1 to 3, characterized in that: Based on the converter performance prediction model, an optimization algorithm is used to optimize the circuit design parameters of the power converter. All circuit parameter combinations that meet the circuit design parameter specifications, the power converter performance parameter specifications, and the circuit design parameter lookup table are output as the target circuit parameter set. The steps include: Taking the performance parameters of the circuit as the optimization target, optimizing the circuit parameters using an optimization algorithm, and taking all the circuit parameter combinations that meet the performance parameter specifications as the first circuit parameter set; Replacing the circuit parameters in each circuit parameter combination in the first circuit parameter set according to the circuit design parameter lookup table to obtain a second circuit parameter set; The second circuit parameter set is screened using the circuit design parameter specification to obtain a target circuit parameter set.

7. The automatic circuit parameter design method according to claim 6, characterized in that: The step of replacing the circuit parameters in each circuit parameter combination in the first circuit parameter set according to the circuit parameter lookup table to obtain the second circuit parameter set includes: Based on the available inductance in the inductance lookup table and capacitor lookup table for available capacitors Inductance in circuit parameter combination and capacitors Replace; if , then select As ,like , then select As ,in, is a coefficient with a value range of 0.85~0.95; if , then select As .

8. An automatic circuit parameter design device, characterized in that: include: A determination module, used to determine circuit design parameter specifications and performance parameter specifications; A lookup table creation module, used for creating a circuit design parameter lookup table for storing available circuit design parameters; A data set construction module is used to construct a data set based on circuit design parameters of the power converter and performance parameters obtained by simulation using circuit simulation software; The model construction module is used to build the PINN-AM model; the PINN-AM model is obtained by introducing the attention mechanism into the PINN model; A training module is used to train the PINN-AM model to obtain a power converter performance prediction model; A first parameter determination module is configured to optimize the circuit parameters of the power converter using an optimization algorithm based on the converter performance prediction model, and output a combination of all circuit parameters that meet the circuit design parameter specification, the performance parameter specification, and the circuit design parameter lookup table as a target circuit parameter set; The second parameter determination module takes cost as a target and selects the optimal circuit parameters from the target circuit parameter set as target design parameters.

9. An electronic device, characterized in that: The electronic device comprises the automatic circuit parameter design device according to claim 8.

10. A computer-readable storage medium, characterized in that At least one program code is stored in the computer-readable storage medium, and the program code is executed by a processor to implement the automatic circuit parameter design method according to any one of claims 1 to 7.

Images