A litz wire winding intelligent design optimization method

Abstract

The application discloses a kind of litz wire winding intelligent design optimization methods, and the litz wire winding ac resistance and porosity are obtained by neural network and hfss, the model of litz wire winding is constructed using hfss simulation software to simulate the ac resistance of litz wire winding under different materials, different number of turns, different litz wire winding radius and different frequency, to complete the training data and test data used to litz wire winding neural network model. Manual grid division is used to simplify the solving steps and time, and the accuracy is ensured. The litz wire winding neural network model is used to automatically train and test using particle swarm optimization algorithm to improve the network model, and finally the litz wire winding neural network model can quickly and accurately obtain the required litz wire winding ac resistance. The application can be used to quickly and accurately obtain the litz wire winding ac resistance under different requirements, reduce the amount of calculation and working time, greatly improve the efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wire design, and more particularly to an intelligent design optimization method for litz wire winding. BACKGROUND

[0002] In recent years, with the rapid development of human science and technology and the continuous improvement of living standards, a variety of electronic products, such as 5G mobile phones, tablet computers, wireless earphones, etc., have gradually entered people's daily life. However, it is inevitable that the charging problem of these electronic devices cannot be avoided. The traditional power supply generally adopts a wired mode, that is, using a wire to supply power to the device. This method is relatively simple and has low loss, and is the mainstream power supply method at present. However, with the increasing number of electronic products in people's life, chargers and charging wires will bring great cost and inconvenience, and in some special environments or special industries, the wired power supply method will have considerable limitations, for example: in underground or underwater environments, damage to the wire may cause a major accident; in flammable and explosive environments, it may cause a fire or explosion; portable wearable human devices will be extremely inconvenient and extremely high in cost if they do not use rechargeable batteries. However, if wireless energy transmission technology is used, it can solve these problems well.

[0003] As one of the most popular emerging technologies in the 21st century, magnetic coupling resonance wireless power transmission (Wireless Power Transfer, WPT) technology has the advantages of high transmission efficiency, strong safety and less affected by transmission distance and orientation, and has a very broad application prospect: today, from the use of enterprise employee cards to the power transmission of large devices in space, whether in the living environment closely related to everyone's life or in the industrial environment that countries are developing, magnetic coupling resonance WPT technology is playing its great value at all times.

[0004] However, this technology still has a high potential practical value, so in recent years it has become a research hotspot for scholars around the world. With the continuous development of litz wire winding alternating current resistance technology, the optimization scheme for litz wire winding design will also be iteratively updated, which is beneficial to reduce the energy loss in the WPT system and greatly reduce the transportation cost of energy. However, the current mainstream method is to use artificial calculation to obtain the alternating current resistance of the litz wire winding by simplifying the calculation formula, which has the problems of long time-consuming, complicated calculation, and inaccuracy of the analytical calculation method. SUMMARY

[0005] The present application aims at the defects of requiring a large amount of manual calculation in the prior art, and provides a time-saving automatic calculation method of litz wire winding AC resistance, so as to reduce the time required for calculation; since the automatic operation can reduce the error caused by manual operation, the AC resistance calculation accuracy is improved by continuously improving and training the litz wire winding neural network model, and the litz wire winding is automatically calculated through the neural network model, so that the optimal resistance value can be quickly and accurately calculated.

[0006] The technical scheme adopted by the present application is a litz wire winding intelligent design optimization method, which is divided into two parts: the first part is to establish a litz wire winding simulation model based on hfss software, and the second part is to optimize the litz wire winding simulation model based on a neural network model.

[0007] S1, establishing a litz wire winding simulation model based on hfss software;

[0008] Step 1.1: establish a plane litz wire winding simulation model through hfss simulation software to obtain AC resistance values under different parameters, obtain various data under different frequencies, different materials, different sizes and different litz wire winding radii, and perform neural network training and testing on the obtained various data. The winding types of the plane litz wire winding simulation model are circular wire, rectangular wire and square wire.

[0009] Step 1.2: in the hfss simulation software, analyze the actual result gap between the simulation results of the established different types of plane litz wire winding and the actual litz wire winding, and compare and confirm that the error gap between different types of wire is small, in order to save simulation time, select square wire as the winding type. Then obtain AC resistance values under different parameters through hfss simulation, and obtain a large amount of data under different frequencies, different materials (different materials result in different electrical conductivity), different sizes and different litz wire winding radii for neural network training and testing.

[0010] S2, optimization of the litz wire winding simulation model based on the neural network model;

[0011] The neural network based on the particle swarm optimization algorithm quickly obtains the porosity and AC resistance of the litz wire winding, and the porosity is the ratio of the litz wire diameter to the litz wire spacing. The specific steps are as follows:

[0012] Step 2.1: The simulation results from the Litz wire winding simulation model created in the previous step in HFSS are used to create the training and test sets in MATLAB for training and testing. The Litz wire AC resistance and porosity are used as the output values in the neural network, and the Litz wire geometry and frequency are used as the input values. In each step of the particle swarm optimization algorithm during the training process, each Litz wire dimension is searched around the minimum point it has found before and the minimum point found by the whole Litz wire dimension swarm. After iterations, the minimum point that the Litz wire dimension swarm has explored before is used as the minimum point of the Litz wire AC resistance function. Each Litz wire dimension in the algorithm represents a possible solution of the Litz wire AC resistance function.

[0013] Step 2.2: The size of each Litz wire dimension (e.g. Litz wire radius) and the frequency of each Litz wire are updated in each iteration under the influence of other Litz wire dimensions (except the Litz wire radius), and the frequency and dimension difference updates follow Equations (1-1) and (1-2).

[0014]

[0015]

[0016] where is the single Litz wire dimension extremum, is the multiple Litz wire dimension extremum, is the previous single Litz wire dimension, is the current single Litz wire dimension. Equation (1-1) is composed of three terms. In the first term, w is the weight that adjusts the current frequency of the Litz wire determines to what extent the Litz wire should keep its previous frequency The second term is the cognitive component of the dimension difference, with the factor adjusting the size of the Litz wire dimension, where rand refers to a random number between 0 and 1. The last term represents the Litz wire dimension adjustment, which is defined by the factor adjusting the dimension difference relative to the best solution of the group. The parameters and control how much weight the search result of perfecting the Litz wire itself and the search result of identifying the Litz wire dimension swarm should get.

[0017] Step 2.3: The single Litz wire dimension extremum and the multiple Litz wire dimension extremum The fitness value of the litz wire size will be updated in each new iteration, along with the position of the optimal solution explored by a single litz wire size and the size of the optimal solution explored by all litz wire sizes in the group of multiple litz wire sizes.

[0018] The PSO algorithm only needs a few parameters to run: the value of the limiting factor (w, and ), the number of litz wire sizes, the maximum number of iterations, and a fitness litz wire AC resistance function. The function of the fitness litz wire AC resistance function is to evaluate the performance of the litz wire AC resistance in each iteration and select the best porosity scheme that minimizes the litz wire AC resistance under the current litz wire conditions.

[0019] Compared with the prior art, the litz wire winding AC resistance and porosity are obtained by the neural network and the hfss. Since the existing technical solutions obtain the litz wire AC resistance and porosity by manual calculation, they are all focused on simplifying the formula, so the present application can greatly reduce the calculation amount and material loss. For the calculation of the AC resistance of the litz wire winding, the current mainstream method is to use the manual calculation method to obtain the AC resistance of the litz wire winding by simplifying the calculation formula. This method has the problems of long time-consuming, complicated calculation, and inaccuracy of analytical calculation. The present application uses the hfss (high-frequency electromagnetic simulation software) simulation software to construct a model of the litz wire winding to simulate the AC resistance value of the litz wire winding under different materials, different turns, different radii of the litz wire winding, and different frequencies, to complete the training data and test data used by the litz wire winding neural network model. For the simulation of the litz wire winding model, the square line grid division method is used to replace the grid division method of the circular line of the litz wire winding, the grid is manually divided, the solving steps and time are simplified, and the accuracy is ensured. Then the litz wire winding neural network model is used to automatically train and test to improve the network model. The trained network model receives the data simulated by the litz wire winding model of the hfss to improve the accuracy of the litz wire winding neural network model. Finally, the litz wire winding neural network model can quickly and accurately obtain the required AC resistance of the litz wire winding. The present application can be used to quickly and accurately obtain the AC resistance of the litz wire winding under different requirements, reduce the calculation amount and working time, and greatly improve the efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 The litz wire winding model to be established in the hfss.

[0021] Figure 2 The function extremum optimization algorithm of the PSO algorithm is a flowchart.

[0022] Figure 3 The joint structure diagram of the hfss and the litz wire winding neural network model. DETAILED DESCRIPTION

[0023] Figure 1 The planar litz wire winding AC resistance to be established in the hfss. First, the planar litz wire winding AC resistance is established in the hfss. When establishing the model, the square wire is compared, the grid division of the square wire is used to divide the grid to reduce the simulation time, the litz wire winding AC resistance is wound by different materials and different turns, the training data and test data are obtained by the frequency sweep mode and the variable scanning mode of the hfss to train the neural network. The obtained data is made into training set and test set by file for use.

[0024] Figure 2 The function extremum optimization algorithm of the PSO algorithm is a flowchart. Then the neural network is built in matlab, the particle swarm optimization algorithm is realized in matlab, and the function extremum optimization algorithm based on the PSO algorithm.

[0025] The design method of the two-layer perceptron neural network in matlab is as follows.

[0026] 1. PSO algorithm parameter setting

[0027] The individual code length is calculated according to formula (1-3), the population size is 30, and the evolution times are 100. The particle swarm algorithm parameters and are both 1.49445.

[0028] length = inpnt n *hidden n +hidden n *output n +hidden n +output n (1-3)

[0029] Wherein, length is the individual code length, input is the input particle, output is the output value, and hidden is the hidden layer number.

[0030] 2. Litz wire size initialization

[0031] The litz wire size initialization refers to assigning a random value to the size and frequency of each litz wire in all litz wire sizes, so as to calculate the litz wire size fitness value according to the fitness litz wire ac resistance function. The method updates the weights of the network through the number of nodes of the input, output and hidden layers, and trains the network.

[0032] 3. Find the first extreme value

[0033] According to the initial random assignment of litz wire size fitness, find single and multiple extreme litz wire sizes in the population.

[0034] 4. Iterative optimization

[0035] The litz wire frequency and size are updated according to formula (1-1) and formula (1-2). The fitness value of the litz wire is calculated according to formula (1-1), and the new single extreme value and multiple extreme values are determined.

[0036] Figure 3 The hfss and litz wire winding neural network model joint block diagram is combined through matlab code, so that manual data updating is not required every time, and the litz wire winding neural network model will automatically call from the hfss.

[0037] Embodiment

[0038] A litz wire winding intelligent design optimization method is divided into two parts: the first part is to establish a litz wire winding simulation model based on hfss software, and the second part is to optimize the litz wire winding simulation model based on a neural network model.

[0039] S1, based on hfss software, establish a litz wire winding simulation model;

[0040] Step 1.1: Since the ac resistance value of litz wire winding is different at different frequencies, the ac resistance value of litz wire winding is also different when the number of turns, radius, etc. of litz wire winding are different, and it is necessary to calculate repeatedly through the formula, which is very complex and time-consuming. At the same time, the planar litz wire winding is made of metal wire and inductive litz wire winding, and the litz wire winding is wound in turn from inside to outside on the same plane. In different environments, different materials will also affect the ac resistance value of litz wire winding, thereby increasing the required time; the hfss simulation software is used to establish a planar litz wire winding to obtain the ac resistance value under different parameters, and a large amount of data under different frequencies, different materials (different materials resulting in different conductivity), different sizes and different litz wire winding radii are obtained for neural network training and testing.

[0041] Step 1.2: In hfss, by studying the simulation results and actual results gap between different kinds of plane litz wire winding (circular wire, rectangular wire and square wire) and the actual litz wire winding, the results are as follows:

[0042]

[0043] It is found that the error gap between different types of lines is small, but the grid division is different between different types, and the simulation time is also different. For rectangular wire, uniform grid division will greatly reduce the simulation time, so in order to reduce the simulation time and improve the calculation efficiency, the circular wire is replaced by the uniform grid of square wire in simulation, so that the simulation data can be quickly called when the neural network is called.

[0044] Second part of neural network model establishment and improvement

[0045] The neural network used in this patent uses particle swarm optimization algorithm (Particle Swarm Optimization, PSO), which optimizes the data to quickly get the appropriate porosity, that is, the ratio of wire diameter to wire spacing. PSO is a simple algorithm for searching for optimal solutions in solution space. It is different from other optimization algorithms, which only need objective function, and do not depend on gradient or any differential form of objective. It needs very few hyperparameters.

[0046] PSO is best suited for finding the maximum or minimum value of a function defined on a multidimensional vector space. Its optimization method is similar to bird flocking for food, starting from some random points in the plane, that is, particles, and searching for the minimum value in a random direction. The specific steps are as follows:

[0047] Step 2.1: In each step, each litz wire size should search around the minimum point it has found and the minimum point found by the whole litz wire size group. After a certain number of iterations, we take the minimum point that the litz wire size group has explored as the minimum point of the litz wire ac resistance function. In this way, each litz wire size in the algorithm represents a possible solution of the litz wire ac resistance function.

[0048] Step 2.2: The size of each litz wire size (such as litz wire radius) and the frequency of each litz wire are updated in each iteration under the influence of other litz wire sizes (except litz wire radius), and the frequency and size difference are updated according to formulas (1-1) and (1-2).

[0049]

[0050]

[0051] Equation (1-1) consists of three terms. In the first term, w is the weight that adjusts the current frequency of the litz wire and determines how much the litz wire should maintain its previous frequency. The second term is the cognitive component of the size difference, with factor cl adjusting the size of the litz wire, where rand is a random number between 0 and 1. The last term represents the litz wire size adjustment, which is adjusted by factor c2 and defines the size difference relative to the best solution of the group. Parameters cl and c2 control how much weight the search results of perfecting the litz wire itself and the search results of identifying the litz wire size group should receive.

[0052] Step 2.3: The single litz wire size extreme value xPbesth and the multiple litz wire size extreme value xGbest are updated with the updated fitness value of the litz wire size in each new iteration. They are the position of the optimal solution explored by the single litz wire size and the size of the optimal solution explored by all litz wire sizes in the multiple litz wire size group, respectively.

[0053] The PSO algorithm only needs a few parameters to run: the limiting factor values (w, cl and c2), the number of litz wire sizes, the maximum number of iterations, and a fitness litz wire ac resistance function. The fitness litz wire ac resistance function is used to evaluate the performance of the litz wire ac resistance in each iteration and select the best solution.

[0054] Step 2.4: By simplifying the simulation model of the ac resistance of the planar litz wire winding and sending the data to the neural network for integration, the optimal planar litz wire winding size and porosity under the corresponding external conditions can be obtained through the neural network, which can greatly reduce the amount of calculation and material loss.

Claims

1. A method for intelligent design and optimization of Litz wire windings, characterized in that, Includes the following steps, S1. Establish a simulation model of the Litz wire winding based on HFSS software; Step 1.1: Establish a planar Litz wire winding simulation model using HFSS simulation software to obtain AC resistance values ​​under different parameters, and obtain various data under different frequencies, materials, sizes, and Litz wire winding radii. Use the obtained data to train and test neural networks. The winding types of the planar Litz wire winding simulation model are circular wire, rectangular wire, and square wire. Step 1.2: In the HFSS simulation software, analyze the difference between the simulation results of different types of planar Litz wire windings and the actual results of the actual Litz wire windings. By comparison, it is confirmed that the error difference between different types of wires is small, and square wire is selected as the winding type. The AC resistance value under different parameters is obtained through HFSS simulation. Data under different frequencies, different materials, different sizes, and different Litz wire winding radii are obtained for neural network training and testing. S2 is an optimization of the simulation model of the litz wire winding based on a neural network model; A neural network based on particle swarm optimization algorithm quickly obtains the porosity and AC resistance of the Litz wire winding. The porosity is the ratio of the Litz wire diameter to the Litz wire spacing. The specific steps are as follows: Step 2.1: Using the Litz wire winding simulation model established in the HFSS simulation in the previous step, the simulation results are used to create training and testing sets in MATLAB. In the neural network, the AC resistance and porosity of the Litz wire are used as output values, and the geometric dimensions and frequency of the Litz wire are used as input values. During the training process, in each step of the particle swarm optimization algorithm, each Litz wire dimension searches around the previously found minimum point and the minimum point found by the entire group of Litz wire dimensions. After iteration, the minimum point explored by this group of Litz wire dimensions is taken as the minimum point of the Litz wire AC resistance function. Each Litz wire dimension represents a possible solution to the Litz wire AC resistance function. Step 2.2: The size of each litz line and the frequency of each litz line are updated in each iteration under the influence of the sizes of other litz lines. The update of the frequency and size difference follows formulas (1-1) and (1-2). in For the extreme values ​​of a single litz line size, For multiple litz line size extrema, For the previous single litz line size, The current size of a single litz line; Equation (1-1) consists of three terms, in the first term, w is the weight that adjusts the current frequency of the litz line. This determines the extent to which the litz line should maintain its previous frequency. The second term is the cognitive component of size difference, factor. Adjusts the size of the litz line; rand refers to a random number between 0 and 1; the last term represents the litz line size adjustment, determined by a factor. Adjustment, defining the size difference relative to the optimal solution in this set; parameters and How much weight should be given to the search results for perfecting the litz line itself and the search results for identifying litz line size groups? Step 2.3: Extreme values ​​of individual litz line dimensions and multiple litz line size extrema It will be updated in each new iteration along with the fitness value of the litz line size; the two are the position of the optimal solution explored by a single litz line size and the size of the optimal solution explored by all litz line sizes in the group of multiple litz line sizes, respectively.

2. The intelligent design and optimization method for Litz wire windings according to claim 1, characterized in that, In the simulation model of the Litz wire winding, a square wire mesh is used instead of the circular wire mesh for the Litz wire winding in order to facilitate accurate calculation of the electromagnetic field distribution inside the wire.

3. The intelligent design and optimization method for Litz wire windings according to claim 1, characterized in that, The Litz wire winding model built using HFSS is autonomously adjustable. By setting a certain frequency range and Litz wire winding size variables in HFSS, different environments can be simulated and different requirements can be achieved by changing the variable values ​​and frequency range.

4. The intelligent design and optimization method for Litz wire windings according to claim 1, characterized in that, The neural network uses the Particle Swarm Optimization (PSO) algorithm.

5. The intelligent design and optimization method for Litz wire windings according to claim 1, characterized in that, The neural network autonomously chooses whether to change the number of neurons to reach the set convergence value, and uses code to call HFSS simulation results at any time to improve the Litz wire winding neural network model.

Images