A low-frequency power suppression method and system based on power module topology

Abstract

The present invention proposes a low-frequency power suppression method and system based on power module topology; obtain the first transfer function of electromagnetic coupling between the power decoupling branch and the main power circuit, obtain the target path and target coupling strength of electromagnetic coupling between the power decoupling branch and the main power circuit through simulation analysis, and the transfer characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power circuit, and construct a filter circuit for suppressing electromagnetic interference to adjust and optimize the circuit layout and wiring of the power decoupling branch and the main power circuit, and adjust the device parameters until the optimized circuit topology meets the preset power suppression requirements. The present invention solves the problem that the parameters such as parasitic inductance and capacitance in the circuit design of the prior art are difficult to control, and the interference of the coupling effect on the circuit cannot be effectively controlled, resulting in poor circuit stability. The present invention effectively reduces the electromagnetic coupling interference between the power decoupling branch and the main power circuit, and improves the electromagnetic compatibility performance of the power module.

Description

Technical Field

[0001] The present invention relates to the field of electronic information technology, and in particular to a low-frequency power suppression method and system based on power module topology. Background Art

[0002] There is a complex coupling mechanism between the power decoupling branch and the main power circuit in the power module. Electromagnetic coupling effect will occur between the circuit elements, interfering with the switching characteristics of the main circuit and causing fluctuations and instability in the branch voltage. Therefore, the wiring of the decoupling branch and the main circuit must be strictly isolated to avoid common impedance coupling.

[0003] At present, the actual circuit board layout of the circuit design of the existing technology is limited, and ideal isolation cannot be achieved. Parameters such as parasitic inductance and capacitance are difficult to control. The complex coupling mechanism inside the power module is difficult to be effectively coordinated in the conventional design of the current existing technology. The inability to effectively control the interference of the coupling effect on the circuit will lead to poor circuit stability. Summary of the invention

[0004] In order to solve the above problems, the present invention proposes a low-frequency power suppression method and system based on power module topology, which can effectively reduce the electromagnetic coupling interference between the power decoupling branch and the main power circuit and improve the electromagnetic compatibility performance of the power module.

[0005] To achieve the above object, an embodiment of the present invention provides a low-frequency power suppression method based on a power module topology, comprising:

[0006] Constructing an equivalent circuit model of electromagnetic coupling between a power decoupling branch and a main power loop of a power module, performing circuit parameter mapping on the equivalent circuit model by presetting a first neural network model, and obtaining a first transfer function of electromagnetic coupling between the power decoupling branch and the main power loop;

[0007] Based on the first transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit, a first simulation analysis is performed on the electromagnetic distribution of the power module by using a preset second neural network to obtain a target path and a target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit;

[0008] Based on the target approach and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, the switching characteristics of the main power circuit switch device are tested by using several different inductance parameters to obtain electromagnetic field simulation results;

[0009] Based on the first transfer function, target path, target coupling strength and electromagnetic field simulation results, a second simulation analysis is performed on the power module circuit to obtain the transfer characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power loop;

[0010] Constructing a filter circuit for suppressing electromagnetic interference, and adjusting and optimizing the circuit layout and wiring of the power decoupling branch and the main power circuit through the filter circuit for suppressing electromagnetic interference based on the transmission characteristics and influence degree of the electromagnetic interference between the power decoupling branch and the main power circuit, so as to obtain an optimized circuit topology;

[0011] Based on the optimized circuit topology, a coupling suppression scheme for the power module is constructed, and device parameters are adjusted until the optimized circuit topology meets the preset power suppression requirements.

[0012] Furthermore, the equivalent circuit model of electromagnetic coupling between the power decoupling branch and the main power loop of the power module is constructed, and the circuit parameters of the equivalent circuit model are mapped by a preset first neural network model to obtain a first transfer function of electromagnetic coupling between the power decoupling branch and the main power loop, including:

[0013] Based on the circuit topology of the power decoupling branch and the main power loop in the power module, the voltage and current of the main power loop and the power decoupling branch are synchronously sampled by setting a number of circuit test sites to obtain an equivalent circuit model of electromagnetic coupling between the power decoupling branch and the main power loop;

[0014] Based on the equivalent circuit model of electromagnetic coupling between the power decoupling branch and the main power loop, the magnetic flux density and electromagnetic field strength values ​​at the test points corresponding to the circuit are obtained;

[0015] Based on the magnetic flux density and electromagnetic field strength values, construct a second transfer function of electromagnetic coupling between the power decoupling branch and the main power loop;

[0016] Based on the second transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop, the circuit parameters of the equivalent circuit model are mapped by a preset first neural network model to obtain a surface diagram of the coupling transfer function;

[0017] Based on the coupling transfer function surface diagram, a coupling correction is performed on the second transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop to obtain a first transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop.

[0018] Further, based on the coupling transfer function surface diagram, a coupling correction is performed on the second transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit to obtain a first transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit, including:

[0019] Perform parameter combination screening on the coupling transfer function surface graph by using a preset screening algorithm to obtain the optimal parameter combination point;

[0020] Based on the second transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop, calculating the transmission loss ratio of the circuit parameters corresponding to the optimal parameter combination point and the corresponding test point;

[0021] Based on the transmission loss ratio, a coupling correction is performed on the second transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop to obtain a first transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop.

[0022] Further, based on the first transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit, a first simulation analysis is performed on the electromagnetic distribution of the power module through a preset second neural network to obtain a target path and a target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, including:

[0023] According to the spatial position of the power decoupling branch and the main power loop, a three-dimensional orthogonal grid is constructed;

[0024] Based on the first transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop, obtaining key circuit parameters of the electromagnetic coupling between the power decoupling branch and the main power loop;

[0025] Calculating the circuit magnetic vector potential value based on the three-dimensional orthogonal grid and key circuit parameters of electromagnetic coupling between the power decoupling branch and the main power loop;

[0026] Based on a preset second neural network, a simulation fitting is performed on the magnetic vector potential value of the circuit to obtain spatial magnetic field distribution data of the power module;

[0027] Based on the spatial magnetic field distribution data of the power module, the magnetic flux density of each distribution site is determined, and the distribution site that meets the preset density threshold is marked as a hot spot area;

[0028] Based on the magnetic field distribution data of the hot spot area, the attenuation curve of the coupling transfer function is calculated;

[0029] Based on the attenuation curve of the coupling transfer function, a target path and a target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power loop are obtained.

[0030] Furthermore, based on the target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, the switching characteristics of the main power circuit switch device are tested by using several different inductance parameters to obtain electromagnetic field simulation results, including:

[0031] Based on the target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power loop, a parameter scan is performed on the power decoupling branch to obtain parameter values ​​of several different inductance parameters;

[0032] Based on the parameter values ​​of the several different inductance parameters, an impact test is performed on the switching characteristics of the main power circuit switch device to obtain a voltage and current stress impact curve;

[0033] Calculating the switching loss value of the main power circuit switch device based on the voltage and current stress influence curve;

[0034] A corresponding relationship between the several different inductance parameters and the switching loss value is constructed to obtain an electromagnetic field simulation result.

[0035] Further, based on the first transfer function, target path, target coupling strength and electromagnetic field simulation results, a second simulation analysis is performed on the power module circuit to obtain the transfer characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power circuit, including:

[0036] Based on the first transfer function, the target pathway, the target coupling strength and the electromagnetic field simulation result, constructing a simulation netlist for the circuit wiring of the power module;

[0037] Based on the current change rate and the induced voltage ratio of the simulation netlist, the mutual inductance coefficient is extracted and a coupling model between the power decoupling branch and the main power loop is constructed;

[0038] Based on the coupling model between the power decoupling branch and the main power loop, a power module circuit is simulated and trained by a preset deep network algorithm to obtain a simulated transfer characteristic spectrum;

[0039] Based on the simulated transfer characteristic spectrum, the transfer characteristic and influence degree of the electromagnetic interference between the power decoupling branch and the main power loop are obtained.

[0040] Furthermore, based on the coupling model between the power decoupling branch and the main power loop, the power module circuit is simulated and trained by a preset deep network algorithm to obtain a simulated transfer characteristic spectrum, including:

[0041] Dynamically predicting the switching noise of the main power circuit in the coupling model between the power decoupling branch and the main power circuit by using a preset first deep network algorithm to obtain a plurality of predicted switching noises of the main power circuit;

[0042] Performing timing extraction on a plurality of predicted switching noises of the main power loop by using a preset second deep neural network algorithm to obtain switching timing parameters;

[0043] Based on the switch timing parameters, a coupling model between the power decoupling branch and the main power loop is solved by a preset third deep neural network algorithm to obtain interference voltage and current waveforms;

[0044] The interference voltage and current waveforms are subjected to effective value extraction and normalization processing by using a preset fourth deep neural network algorithm to obtain a simulated transfer characteristic spectrum.

[0045] Furthermore, a filter circuit for suppressing electromagnetic interference is constructed. Based on the transmission characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power circuit, the circuit layout and wiring of the power decoupling branch and the main power circuit are adjusted and optimized by the filter circuit for suppressing electromagnetic interference to obtain an optimized circuit topology, including:

[0046] Based on the electromagnetic field simulation results, obtaining magnetic flux density distribution data of the filter device in the power decoupling branch and obtaining device design coordinates based on the magnetic flux density distribution data;

[0047] Based on the device design coordinates, a filter circuit layout diagram is obtained to construct a filter circuit for suppressing electromagnetic interference;

[0048] Based on the transmission characteristics and influence degree of the electromagnetic interference between the power decoupling branch and the main power circuit, the circuit device parameters of the power decoupling branch and the main power circuit are adjusted by the filter circuit for suppressing electromagnetic interference until the electromagnetic interference of the circuit layout and wiring of the power decoupling branch and the main power circuit meets the preset design requirements, thereby obtaining an optimized circuit topology.

[0049] Furthermore, based on the optimized circuit topology, a coupling suppression scheme for the power module is constructed, and device parameters are adjusted until the optimized circuit topology meets the preset power suppression requirements, including:

[0050] Based on the optimized circuit topology, a coupling suppression scheme of the power module is constructed by the filter circuit for suppressing electromagnetic interference;

[0051] Based on the coupling suppression scheme of the power module, the layout and wiring of the optimized circuit topology are tested to obtain a corresponding table of test device layout positions;

[0052] Based on the test device layout position correspondence table, device parameters are adjusted until the optimized circuit topology meets the preset power suppression requirements.

[0053] The embodiment of the present invention also provides a low-frequency power suppression system based on power module topology, including: a transfer function construction module, a first simulation analysis module, an electromagnetic field simulation module, a second simulation analysis module, a circuit topology optimization module and a circuit device adjustment module;

[0054] The transfer function construction module is used to construct an equivalent circuit model of electromagnetic coupling between a power decoupling branch and a main power loop of a power module, and to map circuit parameters of the equivalent circuit model by a preset first neural network model to obtain a first transfer function of electromagnetic coupling between a power decoupling branch and a main power loop;

[0055] The first simulation analysis module is used to perform a first simulation analysis on the electromagnetic distribution of the power module based on the first transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit by means of a preset second neural network, so as to obtain a target path and a target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit;

[0056] The electromagnetic field simulation module is used to test the influence of a number of different inductance parameters on the switching characteristics of the main power circuit switch device based on the target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, and obtain an electromagnetic field simulation result;

[0057] The second simulation analysis module is used to perform a second simulation analysis on the power module circuit based on the first transfer function, target path, target coupling strength and electromagnetic field simulation result to obtain the transfer characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power circuit;

[0058] The circuit topology optimization module is used to construct a filter circuit for suppressing electromagnetic interference. Based on the transmission characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power circuit, the circuit layout and wiring of the power decoupling branch and the main power circuit are adjusted and optimized through the filter circuit for suppressing electromagnetic interference to obtain an optimized circuit topology.

[0059] The circuit device adjustment module is used to construct a coupling suppression scheme for the power module based on the optimized circuit topology, and adjust device parameters until the optimized circuit topology meets the preset power suppression requirements.

[0060] Beneficial effects:

[0061] (1) By constructing an equivalent circuit model for the electromagnetic coupling between the power decoupling branch and the main power circuit of the power module, the transfer function of the power decoupling branch and the main power circuit is further obtained. Then, the difference in switching characteristics under different inductance parameters is analyzed by combining electromagnetic field simulation to determine the influence of inductance parameters on the switching characteristics of the main power circuit. In addition, through secondary analysis of the power module, the target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, as well as the transfer characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power circuit are obtained, which provides an important basis for the subsequent optimization of power module layout and improvement of electromagnetic compatibility.

[0062] (2) By designing a filter circuit to suppress electromagnetic interference, the circuit layout and wiring of the power decoupling branch and the main power circuit are optimized and adjusted. From the layout of the device, the selection of parameters and the optimization of the topology, the optimal adjustment results are obtained through strict quantitative analysis to achieve the ideal interference suppression effect of low-frequency power. The coupling suppression scheme of the power module is also designed to continuously adjust the optimized circuit topology. By verifying the effectiveness of the coupling suppression scheme, the suppression scheme is continuously improved and perfected, and the circuit layout and wiring are adjusted. Finally, through the coordination of multi-dimensional data, a reasonable optimization scheme is designed to effectively reduce the electromagnetic coupling interference between the power decoupling branch and the main power circuit and improve the electromagnetic compatibility performance of the power module. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 A schematic flow chart of the steps of a low-frequency power suppression method based on a power module topology provided in an embodiment of the present invention;

[0064] Figure 2 A schematic diagram of the module structure of a low-frequency power suppression system based on power module topology provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] Example 1

[0067] See also Figure 1 , Figure 1 A schematic flow chart of a low-frequency power suppression method based on a power module topology is provided in accordance with an embodiment of the present invention. Figure 1 As shown, the embodiment of the present invention proposes a low-frequency power suppression method based on a power module topology, including steps 101 to 106, each of which is specifically as follows:

[0068] Step 101, constructing an equivalent circuit model of electromagnetic coupling between a power decoupling branch and a main power loop of a power module, performing circuit parameter mapping on the equivalent circuit model by using a preset first neural network model, and obtaining a first transfer function of electromagnetic coupling between the power decoupling branch and the main power loop;

[0069] As an example of this embodiment, based on the circuit topology of the power decoupling branch and the main power circuit in the power module, the voltage and current of the main power circuit and the power decoupling branch are synchronously sampled by setting a number of circuit test sites to obtain an equivalent circuit model of the electromagnetic coupling between the power decoupling branch and the main power circuit; based on the equivalent circuit model of the electromagnetic coupling between the power decoupling branch and the main power circuit, the magnetic flux density and electromagnetic field strength values ​​of the corresponding test sites of the circuit are obtained; based on the magnetic flux density and electromagnetic field strength values, a second transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit is constructed; based on the second transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit, the circuit parameters of the equivalent circuit model are mapped by a preset first neural network model to obtain a coupling transfer function surface graph; based on the coupling transfer function surface graph, the second transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit is coupled and corrected to obtain a first transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit. Specifically, parameter combinations are screened on the coupling transfer function surface diagram through a preset screening algorithm to obtain an optimal parameter combination point; based on a second transfer function of electromagnetic coupling between a power decoupling branch and a main power circuit, a transmission loss ratio of circuit parameters corresponding to the optimal parameter combination point and a corresponding test site is calculated; based on the transmission loss ratio, a coupling correction is performed on the second transfer function of electromagnetic coupling between the power decoupling branch and the main power circuit to obtain a first transfer function of electromagnetic coupling between the power decoupling branch and the main power circuit.

[0070] As an implementable embodiment of the present invention, a Hall sensor is used to synchronously sample the voltage and current of the main power circuit and the power decoupling branch, and the magnetic flux density and the electromagnetic field strength values ​​are obtained according to the sampled data. Specifically, the Hall sensors are arranged at four measuring points of the main power circuit and the power decoupling branch, and the voltage and current are synchronously sampled with a sampling frequency of 20kHz and a time window of 100μs to obtain the magnetic flux density and the electromagnetic field strength values ​​at the corresponding positions and construct an equivalent circuit model of the electromagnetic coupling between the power decoupling branch and the main power circuit; in an application of this embodiment, a magnetic flux density distribution diagram is obtained by arranging a Hall sensor array between the master and slave windings, and four Hall sensors are arranged on the power module PCB board, which are respectively located at four positions of the upper, lower, left and right of the master and slave windings, and the sampling frequency of each sensor is 20kHz. When a 10A switching current passes through the main power circuit, the magnetic flux density detected by the Hall sensor is 0.5T, and the corresponding electromagnetic field strength is 4000A / m. In this way, the voltage and current are synchronously sampled by four Hall sensors to obtain the magnetic flux density distribution diagram and construct an equivalent circuit model of the power module.

[0071] Then, a group of nonlinear differential equations is constructed according to the numerical values ​​of the magnetic flux density and the electromagnetic field strength, and the relationship between the change of the magnetic flux density is calculated by Maxwell's equations. Specifically, the coupling coefficient k is calculated by multiplying the magnetic flux density by the integral value of the electromagnetic field strength, and the master-slave coupling circuit transfer function (equivalent to the second transfer function) H(s)=(k×M) / (L1×L2) is constructed, where M is the mutual inductance coefficient, L1 and L2 are the master and slave winding inductance values. In one application case, when determining the coupling coefficient and the transfer function parameters, the principle of magnetic flux continuity needs to be considered. If a sinusoidal alternating current flows through the main power circuit, the induced current in the decoupling branch has the same frequency as the main circuit current, and the amplitude and phase are determined by the coupling coefficient. For example, for a coupling circuit with a turns ratio of 1:1, when the main circuit current is 10A and the frequency is 10kHz, the induced current amplitude in the decoupling branch is 8A, and the phase lags by 45 degrees.

[0072] A BP neural network is used to establish a mapping relationship between the flux linkage and the current change rate data, and a main power circuit current amplitude and a power decoupling branch resistance parameter combination are extracted from the BP neural network output data. Specifically, a second-order nonlinear differential equation group is constructed according to the main power circuit voltage and current sampling values ​​and the power decoupling branch sampling value data. The magnetic flux density change relationship is calculated using the Maxwell equations under the condition of unit magnetic permeability, and the flux linkage and current change rate data corresponding to the coupling coefficient are extracted from the transfer function. The BP neural network is used to establish a corresponding relationship between the flux linkage and the current change rate data, and the main power circuit current amplitude is extracted from the BP neural network output data at 0 .1A to 10A range and the power decoupling branch resistance value range of 10Ω to 1000Ω parameter combination, according to the ten-fold cross validation rule to calculate the coupling transfer function value of all parameter combinations, to construct the coupling transfer function surface diagram; in one application case, the BP neural network adopts a three-layer structure to fit the magnetic flux and current change rate data. The input layer receives the magnetic flux density and electromagnetic field strength data, the hidden layer sets 16 neurons, and the output layer generates the coupling coefficient prediction value. By scanning the main power loop current in the range of 0.1A to 10A, the coupling coefficient curves at different working points are obtained to construct the coupling transfer function surface diagram.

[0073] The transfer function value is calculated according to the parameter combination, and the optimal parameter combination point is obtained from the transfer function surface diagram by the gradient descent method to obtain the modified electromagnetic coupling transfer function parameters. Specifically, the optimal parameter combination point is obtained from the surface diagram according to the gradient descent method, and the magnetic flux density amplitude and phase angle are calculated according to the transfer function formula according to the current value and resistance value corresponding to the optimal parameter combination point. The transmission loss ratio of the master and slave coils at the same measuring point is calculated in combination with the Hall sensor measurement data, and the coupling coefficient correction value and position relationship are reversed from the transmission loss ratio data. The master-slave winding turns ratio and the coupling coefficient k are corrected to obtain the modified electromagnetic coupling transfer function (equivalent to the first transfer function) and its corresponding key parameters; in one application case, the coupling transmission The mutual inductance coefficient in the transfer function decays exponentially with the increase of the distance between the master and slave windings. When the distance increases from 5mm to 50mm, the mutual inductance coefficient decreases from 0.8 to 0.08. The transmission loss ratio reflects the energy transfer efficiency between the master and slave windings. Under the ideal coupling state, the transmission loss ratio is close to 1. In the actual measurement, when the distance between the master and slave windings is 10mm, the transmission loss ratio is 0.85, indicating that 85% of the magnetic energy is effectively transferred. The coupling coefficient correction value calculated by the transmission loss ratio enables the transfer function to more accurately describe the electromagnetic coupling characteristics in the actual circuit. When the main power circuit operates at a switching frequency of 50kHz, the error between the predicted value of the corrected transfer function and the measured induced voltage is less than 5%.

[0074] It is worth mentioning that the establishment of the electromagnetic coupling transfer function originates from the crosstalk problem existing in the power electronic switching circuit. In the power module, electromagnetic coupling is formed between the main power circuit and the power decoupling branch through mutual inductance, where the mutual inductance coefficient is related to the number of winding turns, the core material and the winding distance. When the current change rate of the main power circuit is large, an induced electromotive force proportional to the current change rate of the main power circuit is induced in the power decoupling branch. For example, in a switching power supply with a power of 1000W, when the main power circuit current rises from 0 to 10A with a rise time of 1μs, a peak voltage of 100V is induced in the decoupling branch 10mm away from the main power circuit.

[0075] Step 102, based on the first transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop, a first simulation analysis is performed on the electromagnetic distribution of the power module by using a preset second neural network to obtain a target path and a target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power loop;

[0076] As an example of this embodiment, a three-dimensional orthogonal grid is constructed according to the spatial position of the power decoupling branch and the main power loop; based on the first transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop, the key circuit parameters of the electromagnetic coupling between the power decoupling branch and the main power loop are obtained; based on the three-dimensional orthogonal grid and the key circuit parameters of the electromagnetic coupling between the power decoupling branch and the main power loop, the circuit magnetic vector potential value is calculated; based on a preset second neural network, the circuit magnetic vector potential value is simulated and fitted to obtain the spatial magnetic field distribution data of the power module; based on the spatial magnetic field distribution data of the power module, the magnetic flux density of each distribution site is judged, and the distribution site that meets the preset density threshold is marked as a hot spot area; based on the magnetic field distribution data of the hot spot area, the attenuation curve of the coupling transfer function is calculated; based on the attenuation curve of the coupling transfer function, the target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power loop are obtained.

[0077] As an implementable method of this embodiment, a three-dimensional orthogonal grid is constructed according to the spatial position of the main power circuit and the power decoupling branch, and the carrier drift velocity and current density data are calculated through the three-dimensional orthogonal grid. Specifically, a three-dimensional orthogonal grid is constructed in the plane where the main power circuit and the power decoupling branch are located, and the side length of the grid unit is set to 0.5 mm. The drift velocity distribution of the carrier in the cross section of the wire is calculated according to the current flowing through the loop wire, and the current density data is generated according to the product of the drift velocity and the carrier concentration. The magnetic field energy is calculated by integrating each grid unit according to the product formula of the current density and the magnetic flux density, and a finite element node matrix containing position coordinates and field strength data is generated.

[0078] The magnetic vector potential value is calculated using the current density data and the magnetic permeability parameter of the air medium, and the magnetic vector potential value is fitted through a feedforward neural network to obtain the spatial magnetic field distribution data of the power module. Specifically, on the basis of the finite element node matrix, the magnetic vector potential and the magnetic scalar potential values ​​are calculated according to the magnetic permeability and relative magnetic permeability parameters of the air medium, and the flux continuity condition and the current density boundary condition are applied to the conductor boundary. The magnetic vector potential field distribution is solved by constructing a lumped parameter expression on the conductor surface, and a feedforward neural network with a depth of 5 layers and 32 neurons is used to generate a fitting curve for the magnetic field distribution data to obtain the spatial magnetic field distribution data of the power module.

[0079] The coordinates of the sampling points are extracted from the spatial magnetic field distribution data, and the angle between the direction of the magnetic field line and the axis of the power decoupling branch is obtained by a spatial vector cross multiplication operation. Specifically, 1000 sampling points are extracted from the magnetic field distribution data generated by a feedforward neural network (equivalent to the preset second neural network) according to the coordinates of the center points of the grid units, and the magnetic field direction vectors between 8 adjacent points around each sampling point are calculated. The angle between the direction of the magnetic field line and the axis of the power decoupling branch is obtained by a spatial vector cross multiplication operation, and the angle data is spatially mapped and transformed using a three-dimensional Cramer matrix, so that the spatial magnetic field direction data is constructed into a three-dimensional distribution line diagram of the magnetic field lines.

[0080] If the magnetic flux density value is greater than a preset threshold, it is marked as a hot spot area. The attenuation curve of the coupling transfer function is calculated based on the spatial position data of the hot spot area and the magnetic permeability of the medium. Specifically, the area with a magnetic flux density greater than 0.5T in the three-dimensional distribution line diagram of magnetic lines of force is marked as a hot spot area, and the relative spatial position data of the main power loop and the decoupling branch in the hot spot area are extracted. The amplitude attenuation curve of the coupling transfer function in this area is calculated based on the magnetic permeability of the master and slave coil media and the eddy current loss density. The characteristic vector of the change trend of the coupling coefficient k in the coupling transfer function on the spatial coordinate axis is extracted to obtain the target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power loop.

[0081] In an practicable application, a switching power supply module with a power of 1000W is used as a specific example, and the cross-sectional area of ​​the main power loop wire is 4mm 2 , the cross-sectional area of ​​the power decoupling branch conductor is 2mm 2 The center distance between the two loops is 10mm. When the main loop current increases from 0 to 10A, the carriers are Gaussian distributed on the wire cross section, and the current density in the center area reaches 250A / mm 2 , the current density in the edge area is reduced to 50A / mm 2In the above power module simulation, a grid size of 0.5mm is selected to strike a balance between the amount of calculation and the accuracy. For the conductor cross section, it is divided into 8×8 grid units in the xy plane and 200 grid units in the z direction to form a three-dimensional grid structure; each grid node records the magnetic field components in three directions, and the magnetic field energy distribution obtained by current density calculation shows obvious radial attenuation characteristics. In the process of fitting the magnetic field distribution by the feedforward neural network (equivalent to the preset second neural network), the input layer receives the spatial coordinates and current density data of the grid nodes, and extracts features through 5 hidden layers; it is worth mentioning that when the main circuit current is 10A, the magnetic field strength at 1mm away from the center of the conductor is 7960A / m, and it drops to 1592A at 5mm. / m, showing a typical inverse attenuation law, and the error between the prediction result of the neural network and the theoretical calculation value is within 5%; in the process of spatial mapping transformation of the angle data using the three-dimensional Cramer matrix, the sampling point selection follows the principle of uniform distribution, and 1000 sampling points are arranged in a space range of 10mm×10mm×20mm. The spacing between adjacent sampling points is 1mm. Each sampling point determines the direction of the magnetic field line by calculating the magnetic field vector of the surrounding 8 points. The angle distribution between the magnetic field line and the axis of the decoupling branch reflects the spatial orientation characteristics of the magnetic field; in the identification process of the hot spot area, it can be found that the areas with a magnetic flux density exceeding 0.5T are mainly distributed at the intersection of the master and slave loops. The magnetic permeability of these areas is 5 times that of the ferrite core material, and the eddy current loss density is about 0.2W / cm 3 By analyzing the spatial distribution of the hot spot area, it is found that the coupling coefficient k decays exponentially with the increase of the axial distance. The coupling coefficient is 0.8 at 5mm from the center of the main loop, and decays to 0.2 when it increases to 15mm. After processing and analyzing the above steps, the target approach and target coupling strength for obtaining the electromagnetic coupling between the power decoupling branch and the main power loop can be obtained.

[0082] Step 103, based on the target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, a switching characteristic of the main power circuit switch device is tested by using several different inductance parameters to obtain an electromagnetic field simulation result;

[0083] As an example of this embodiment, based on the target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, a parameter scan is performed on the power decoupling branch to obtain parameter values ​​of several different inductance parameters; based on the parameter values ​​of the several different inductance parameters, an impact test is performed on the switching characteristics of the main power circuit switching device to obtain a voltage and current stress influence curve; based on the voltage and current stress influence curve, the switching loss value of the main power circuit switching device is calculated; and a corresponding relationship between the several different inductance parameters and the switching loss value is constructed to obtain an electromagnetic field simulation result.

[0084] As an implementable method of this embodiment, a circuit breaker is used to obtain scanning data of the inductance parameters in the power decoupling branch, and the scanning data is transformed through Fourier transform to obtain time domain parameters and harmonic amplitudes. Specifically, an analog circuit breaker is used to scan the inductance parameters in the power decoupling branch in the range of 100μH to 1000μH at equal intervals, and the sampling frequency is set to 1GHz and the sampling depth is set to 10,000 points. The voltage and current waveforms of the main power circuit switch tube during the conduction and shutdown process are recorded, and a 1024-point Hanning window Fourier transform is applied to the sampled data. The four time domain parameters of rise time, fall time, overshoot value and resonance frequency are calculated according to the transformation results, and the amplitude data of the first ten harmonics are obtained from the spectrum curve.

[0085] The time domain parameters and harmonic amplitudes are input into a long short-term memory neural network for training. The long short-term memory neural network outputs a waveform parameter group to establish an inductance value comparison database. Specifically, a two-layer 32-node long short-term memory neural network is used to classify waveform features, and four time domain parameters of voltage rise time, fall time, overshoot value, resonance frequency and the first ten harmonic amplitudes are input into training samples. After the inductance parameters and corresponding waveform parameters are standardized, they are imported into the training sequence according to the time sequence, and the switching waveform parameter group is obtained from the neural network output. A switching waveform parameter and inductance value comparison database is established in the feedback data.

[0086] A reference inductance value is extracted from the inductance value reference database for parameter traversal to obtain traversal data, and the traversal data is processed by a Kalman filter to obtain a voltage and current stress response curve. Specifically, starting from extracting the reference inductance value from the reference database, the inductance parameter is traversed according to a step value of 10 μH, and the voltage stress and current stress of the switch tube in the on and off states are measured at each inductance value. The measurement noise variance of the Kalman filter is set to 0.01V and the process noise variance is set to 0.1V to filter the voltage and current waveforms, and the response curve (equivalent to the voltage and current stress influence curve) is drawn in a coordinate system where the horizontal axis is the inductance value and the vertical axis is the switch tube stress.

[0087] The switching loss value is calculated according to the voltage-current stress response curve. The switching loss value is normalized to establish a corresponding relationship with the inductance parameter and the switching characteristic. The switching loss value is calculated specifically according to the response curve coordinate data, the voltage-current product integral when the switch is on is taken as the conduction loss, and the voltage-current product integral when the switch is off is taken as the off loss. The switching losses of all working points are normalized based on the maximum switching loss value, and the corresponding relationship between the inductance parameter and the overshoot value of the switching frequency rise time and fall time is marked in the normalized numerical sequence to obtain the electromagnetic field simulation result.

[0088] In an implementable application, in a typical power converter circuit, the main power circuit uses a power field effect tube with a rated current of 20A as a switching device, and the inductor in the power decoupling branch uses a manganese-zinc ferrite core with a magnetic permeability of 2000. The inductance value is adjusted by changing the number of turns of the inductor winding. When the switching device operates at a frequency of 100kHz, a 1GHz sampling frequency is used to record the transient waveform during the switching process. Specifically, when the power field effect tube switches from off to on, the drain-source voltage drops from 400V to 2V, the drain current increases from 0 to 10A. The spectrum characteristics obtained by Fourier transform show that when the inductance value is 200μH, the 100kHz fundamental component accounts for 80%, the 300kHz harmonic component accounts for 15%, and the 500kHz harmonic component accounts for 5% in the switching waveform; in the process of using the long short-term memory neural network training data, the inductance value increases from 100μH to 1000μH, the switch tube voltage rise time is extended from 50ns to 200ns, and the voltage overshoot peak value is reduced from 600V to 450 V, the neural network establishes the mapping relationship between the inductance parameters and the switch waveform parameters by learning these timing characteristics; when the inductance value is 500μH, the drain-source voltage of the switch tube in the on state has a high-frequency oscillation with an amplitude of 100mV, and the drain current in the off state contains a spike interference with an amplitude of 200mA. After filtering with the Kalman filter, the obtained switch stress response curve clearly shows that for every 100μH increase in inductance value, the switch overshoot voltage decreases by 50V and the oscillation frequency decreases by 20kHz; at the reference operating point When the inductance value is 300μH, the switch tube conduction loss is 2W, and the turn-off loss is 3W. When the inductance value increases to 800μH, the conduction loss is reduced to 1.5W, and the turn-off loss is reduced to 2W, but the switching speed is significantly reduced. The normalized loss curve shows that in the range of 400μH to 600μH, the balance between switching characteristics and losses is optimal, the switching frequency can be maintained at 100kHz, and the voltage overshoot is limited to below 450V. This electromagnetic field simulation result can guide the selection of inductance parameters in the power decoupling branch.

[0089] Step 104, based on the first transfer function, target path, target coupling strength and electromagnetic field simulation result, a second simulation analysis is performed on the power module circuit to obtain the transfer characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power loop;

[0090] As an example of this embodiment, based on the first transfer function, target path, target coupling strength and electromagnetic field simulation results, a simulation netlist is constructed for the circuit wiring of the power module; based on the current change rate and the induced voltage ratio of the simulation netlist, the mutual inductance coefficient is extracted and a coupling model between the power decoupling branch and the main power loop is constructed; based on the coupling model between the power decoupling branch and the main power loop, the power module circuit is simulated and trained through a preset deep network algorithm to obtain a simulated transfer characteristic spectrum; based on the simulated transfer characteristic spectrum, the transfer characteristics and influence degree of the electromagnetic interference between the power decoupling branch and the main power loop are obtained. Specifically, a first deep network algorithm is preset to dynamically predict the switching noise of the main power circuit in the coupling model between the power decoupling branch and the main power circuit, so as to obtain several predicted switching noises of the main power circuit; a second deep neural network algorithm is preset to perform timing extraction on the several predicted switching noises of the main power circuit, so as to obtain switching timing parameters; based on the switching timing parameters, a third deep neural network algorithm is preset to solve the coupling model between the power decoupling branch and the main power circuit, so as to obtain interference voltage and current waveforms; a fourth deep neural network algorithm is preset to perform effective value extraction and normalization on the interference voltage and current waveforms, so as to obtain a simulated transfer characteristic spectrum.

[0091] As an implementable method of this embodiment, a simulation netlist is constructed according to the circuit wiring of the power module, the mutual inductance coefficient is extracted by the ratio of the current change rate to the induced voltage in the simulation netlist, and a coupling path two-port network (equivalent to the coupling model between the power decoupling branch and the main power loop) is established by using the mutual inductance coefficient. Specifically, the simulation netlist is constructed according to the actual circuit wiring of the power module, the mutual inductance coefficient is extracted by using the ratio of the current change rate to the induced voltage in the transfer function, the coupling coefficient is calculated from the product of the magnetic flux density and the electromagnetic field intensity, the parasitic capacitance is obtained by measuring the ratio of the potential difference and the charge between the conductors, an impedance matrix is ​​established between the conductor nodes by the dual port theorem, and a coupling path two-port network between the main power loop and the power decoupling branch is constructed by using the port input impedance and the transfer impedance.

[0092] A third-order Wiener predictor (equivalent to the preset first deep network algorithm) is used to dynamically predict the switching noise of the main power circuit in the coupled path two-port network, and a five-layer deep residual network (equivalent to the preset second deep neural network algorithm) is used to extract the switching timing parameters from the main power circuit switching noise. Specifically, based on the coupled path two-port network, a third-order Wiener predictor is used to dynamically predict the switching noise source of the main power circuit, and the fundamental component and the fifth harmonic component in the switching voltage and current waveforms are decomposed by Fourier spectrum. The five-layer deep residual network is used to learn the time domain feature vector of the switching noise source, and the turn-on rise time, turn-off fall time and switch ringing frequency parameters of the main power circuit switching process are extracted from the noise source vector.

[0093] According to the coupling pathway two-port network and the switching timing parameters, the fourth-order Runge Kutta method (equivalent to the preset third deep neural network algorithm) is used to solve the power loop coupling state equation group, and the interference voltage and current waveforms are obtained from the power loop coupling state equation group. Specifically, according to the coupling pathway two-port network and the switching noise source parameters, the transient simulation time step is set to 1% of the switching period, and the fourth-order Runge Kutta method is used to solve the electromagnetic coupling state equation group composed of the power loop and the decoupling branch, and the voltage and current waveforms of each key node in the two loops are recorded, and the interference voltage amplitude and interference current amplitude during the conduction and shutdown process are extracted from the waveform data.

[0094] The interference voltage and current waveforms are divided into frequency sub-bands, and a fourth-order Butterworth filter (equivalent to the preset fourth deep neural network algorithm) is used to filter the sub-band signals to extract effective values. The transfer characteristic spectrum is obtained through normalization processing. Specifically, the extracted interference voltage and current waveforms are divided into frequency sub-bands, and sub-band signals in the range from the switching frequency to five times the switching frequency are obtained respectively. The fourth-order Butterworth filter is used to filter each sub-band signal to extract effective values. The proportion of sub-band signals in each frequency band to the total interference signal is calculated through normalization processing, and the transfer characteristic spectrum of the main power loop switching noise transmitted to the power decoupling branch through the coupling path is obtained.

[0095] In an implementable application, a power converter with a semiconductor switch tube with a rated current of 20A in the main power circuit, a 300μH inductor in the power decoupling branch, and a distance of 10mm between the two circuits is taken as an example. When the current change rate of the main circuit reaches 1A / ns, a 50mV interference voltage is induced in the decoupling branch, the mutual inductance coefficient between the main and slave circuits is 0.8μH, the corresponding coupling coefficient is 0.6, the parasitic capacitance between the two wires is 10pF, and the common mode impedance is 50Ω; the switching noise source parameters are at a switching frequency of 100kHz, the amplitude of the fundamental component of the main circuit switching waveform is 400V, the 300kHz harmonic component is 80V, and the 500kHz harmonic component is 40V. The third-order Wiener predictor dynamically tracks these noise components, and the prediction error is controlled within 5%; the deep residual network accurately identifies the rise time of 20ns and the fall time of 30ns in the switching process by learning these features; At a switching frequency of 100kHz, a time step of 100ps is selected to ensure that there are 10,000 sampling points in each switching cycle. When the fourth-order Runge-Kutta method is used to solve the state equation, when the voltage of the main circuit switch node rises from 0 to 400V, the decoupling branch induces a common-mode interference voltage with an amplitude of 2V and a loop interference current with an amplitude of 100mA; in the range of switching frequency to five times the switching frequency, the fourth-order Butterworth filter divides the interference signal into five frequency bands. The interference signal in the 100kHz band accounts for 50% of the total interference, the 300kHz band accounts for 30%, and the 500kHz band accounts for 15%. The proportion of higher frequency bands gradually decreases. This spectrum distribution feature shows that the interference at the switching fundamental frequency is the most significant, and the high-order harmonic components gradually attenuate after passing through the coupling path. By comparing the simulation results under different working conditions, it is found that factors such as switching frequency, current change rate, and wire layout will affect the transmission characteristics of interference. When the switching frequency increases from 100kHz to 200kHz, the interference peak increases by 50% and the proportion of high-frequency components increases. When the master-slave loop distance increases from 10mm to 20mm, the coupling coefficient decreases to 0.3 and the interference amplitude decreases by 40%. Finally, based on the above steps, the simulated transfer characteristic spectrum is obtained.

[0096] Step 105, constructing a filter circuit for suppressing electromagnetic interference, and adjusting and optimizing the circuit layout and wiring of the power decoupling branch and the main power circuit through the filter circuit for suppressing electromagnetic interference based on the transmission characteristics and influence degree of the electromagnetic interference between the power decoupling branch and the main power circuit, so as to obtain an optimized circuit topology;

[0097] As an example of this embodiment, based on the electromagnetic field simulation results, the magnetic flux density distribution data of the filter device in the power decoupling branch is obtained and the device design coordinates are obtained based on the magnetic flux density distribution data; based on the device design coordinates, the filter circuit layout diagram is obtained, and a filter circuit for suppressing electromagnetic interference is constructed; based on the transfer characteristics and influence degree of the electromagnetic interference between the power decoupling branch and the main power circuit, the circuit device parameters of the power decoupling branch and the main power circuit are adjusted through the filter circuit for suppressing electromagnetic interference until the electromagnetic interference of the circuit layout and wiring of the power decoupling branch and the main power circuit meets the preset design requirements, thereby obtaining an optimized circuit topology.

[0098] As an implementable method of this embodiment, the magnetic flux density distribution data of the filter device in the power decoupling branch is obtained according to the magnetic field simulation results, and the device placement coordinates are determined by the magnetic flux density data. Specifically, the filter bandwidth is divided according to the main power circuit switching frequency to five times the switching frequency, and the layout of the filter device in the power decoupling branch is planned in a 30mm×30mm area. According to the magnetic field simulation results, the device placement coordinates are determined at a position where the magnetic flux density is less than 0.1T.

[0099] The device placement coordinates are used to train a deep neural network model, and a filter circuit layout diagram is obtained from the output of the deep neural network model. Specifically, a deep neural network composed of a five-layer convolutional structure and a fully connected layer is used to optimize the device layout. The magnetic field interference intensity and the wiring area are set as optimization objective functions, and the filter circuit layout diagram is obtained from the neural network output.

[0100] A filter circuit topology is generated according to the filter circuit layout diagram, wherein the filter circuit topology includes a fundamental frequency low-pass filter channel and a harmonic frequency band-stop suppression channel. Specifically, the filter circuit topology is generated through the layout diagram, a second-order low-pass filter channel is set at the fundamental frequency, and a band-stop suppression channel is set at the second harmonic to the fifth harmonic frequency. A linear regressor is used to model the frequency response of the filter channel, and the impedance of each node of the filter circuit is calculated from the frequency response data. The filter device layout diagram is converted into a circuit node network diagram, and a resonant node and impedance matching point for suppressing interference are constructed.

[0101] The frequency response is calculated for the filter circuit topology. If the frequency response data deviates from the set range (equivalent to the preset design requirements), the Butterworth function is used to correct the filter device parameters to obtain the corrected filter channel transfer characteristics. Specifically, the capacitance value is scanned in the range of 100pF to 1000pF according to the impedance matching point, and the inductance value is scanned in the range of 1μH to 100μH. The attenuation response and phase delay of the filter circuit at each frequency point are calculated using the sixth-order Butterworth function. The filter channel transfer characteristics are solved by complex frequency domain operations, and the corresponding attenuation multiple and phase offset are calculated for the interference components of each frequency band. The filter device parameters are corrected using the attenuation multiple and phase offset data, the attenuation depth at the fundamental frequency is set as the reference value, and the capacitor and inductor parameters are corrected according to the multiple relationship at the harmonic frequency. The corrected filter branch frequency response is extracted from the Butterworth function attenuation curve, a magnetic ring suppression winding is added to the area where the magnetic flux density exceeds 0.1T, and a differential mode suppression capacitor is arranged at the overlap of the master and slave loops to complete the optimization of the filter circuit topology and device layout.

[0102] In an implementable application, taking a typical power converter as an example, when the main power circuit operates at a switching frequency of 100kHz, fundamental and harmonic interference will be generated. The magnetic field simulation shows that the magnetic flux density is 0.2T at 10mm away from the main circuit, and it drops to 0.05T at 20mm, which provides an important reference for the layout of the filter device. In the layout area of ​​30mm×30mm, the low-pass filter capacitor is placed at a position where the magnetic field strength is lower than 0.1T, and the resonant inductor is arranged at the edge area away from the main circuit. The neural network obtains the optimal layout coordinates of the device by optimizing the two objective functions of magnetic field interference intensity and wiring area. In specific measurements, when the capacitor device is located 20mm outside the main circuit, the induced interference voltage is reduced by 80%. The filter channel is designed based on the frequency segmentation principle. At the fundamental frequency of 100kHz, a second-order low-pass filter with a cutoff frequency of 150kHz is used, and the attenuation slope is 40dB / dec. For the harmonic frequency band of 300kHz to 500kHz, a band-stop filter with a quality factor of 10 is set. The filter provides 30dB attenuation at the center frequency; the linear regressor predicts the impedance matching point of each frequency band by learning the frequency response characteristics; the device parameter selection and optimization process includes parameter scanning when the capacitance value is in the range of 100pF to 1000pF and the inductance value is in the range of 1μH to 100μH. The calculation results of the sixth-order Butterworth function show that the selection of 470pF capacitors and 47μH inductors at 100kHz can achieve 25dB attenuation, and the use of 220pF capacitors at 300kHz can achieve 25dB attenuation. The parallel resonance with the 22μH inductor provides 35dB attenuation; the optimization of the filter circuit is to add a ferrite ring with a magnetic permeability of 2000 in the area where the magnetic flux density exceeds 0.1T, and provide an additional 15dB of magnetic field attenuation through a 4-turn winding. A 100pF differential mode suppression capacitor is arranged at the overlap of the master and slave loops to reduce 80% of the differential mode interference current. This multi-level suppression measure reduces the interference voltage at the switching frequency from 2V to 0.1V, and the interference current at the harmonic frequency from 100mA to 5mA.

[0103] Step 106: construct a coupling suppression scheme for the power module based on the optimized circuit topology, and adjust device parameters until the optimized circuit topology meets the preset power suppression requirements.

[0104] As an example of this embodiment, based on the optimized circuit topology, a coupling suppression scheme of the power module is constructed by the filter circuit for suppressing electromagnetic interference; based on the coupling suppression scheme of the power module, the layout and wiring of the optimized circuit topology are tested to obtain a test device layout position correspondence table; based on the test device layout position correspondence table, the device parameters are adjusted until the optimized circuit topology meets the preset power suppression requirements.

[0105] As an implementable mode of this embodiment, a voltage and current probe is used to synchronously sample the power module test node, and the power module test node includes a main power loop switch tube drive end node and a power decoupling branch node. Specifically, eight test nodes are set in the power module, including two points of the main power loop switch tube drive end, two points of the power end, two points of the ground end, and two points of the power decoupling branch. The voltage and current probe with a sampling frequency of 1 GHz and a sampling depth of 10,000 points is used for synchronous sampling;

[0106] A neural network model is trained according to the synchronous sampling data, and the neural network model outputs impedance amplitude and phase data between test nodes. Specifically, the test point impedance data is trained using a five-layer feedforward neural network, each containing 64 neurons. The input layer receives the voltage and current sampling values ​​of eight test nodes, and the impedance amplitude and phase data between each test node are obtained from the output layer.

[0107] The impedance data is subjected to Fourier transform to obtain a spectrum feature vector, and the spectrum feature vector is classified by a random forest regressor to obtain a device layout position correspondence table. Specifically, a 1024-point Fourier transform is performed in the fundamental frequency to fifth harmonic frequency range according to the impedance data of the measuring point, and the transformed spectrum is segmented to extract amplitude and phase information at a frequency interval of 20kHz. A random forest regressor with a tree depth of 8 is used to classify the spectrum feature vector, and the feature vector of each frequency band is obtained at the output end of the regressor. The filter device parameters and position coordinates are associated and calibrated through impedance matching calculation to form a device value and layout position correspondence table. The layout is adjusted according to the data in the correspondence table. A ferrite ring with a magnetic permeability of 2000 and a four-turn suppression winding are added at the position where the magnetic flux density exceeds 0.1T. The position of the inductor in the filter branch is offset to the weakest magnetic field. The device placement spacing is optimized using the resonant cavity parameters with a quality factor of 10. A loss matching circuit with a quality factor of 20 is arranged at the node of the master-slave loop, and the coordinates of the key points of the device layout are calibrated from the loss parameter curve.

[0108] A PCB stacking structure is generated according to the device layout position correspondence table. The interlayer impedance and the crosstalk coefficient of the PCB stacking structure are determined by distributed parameter calculation. If the interlayer impedance exceeds a preset threshold, a magnetic flux suppression slot is added in the overlapping area of ​​the main loop and the decoupling branch. Specifically, a four-layer PCB stacking structure is generated from the layout coordinates, a three-times line width spacing wiring rule is adopted for the routing layer, a dielectric layer with a relative dielectric constant of 4.4 is filled between the power layer and the ground layer, and a two-layer 20mil thick copper foil shielding is adopted for key signal nodes. The interlayer impedance and the crosstalk coefficient are solved by distributed parameter calculation, and the stacking order and wiring rules of each layer are corrected by using the solution results, and a magnetic flux suppression slot and a ground shielding hole are added in the overlapping area of ​​the main loop and the decoupling branch.

[0109] In an implementable application, a power converter with a rated current of 20A in the main power circuit and a 300μH inductor in the power decoupling branch is used for interpretation. By arranging test probes at key nodes to obtain measured data, the test point at the driving end of the main circuit switch tube records a switch drive signal with an amplitude of 20V, the power end records a switch stress of 400V, the ground end detects a common mode noise of 200mV, and the decoupling branch measures an induced voltage of 2V; the feedforward neural network establishes an impedance relationship model between the test points by training 1000 sets of test data. At a switching frequency of 100kHz, the coupling impedance between the main circuit and the decoupling branch is 50Ω, with a phase lag of 45 degrees. The spectrum analysis shows that the coupling impedance of the 300kHz harmonic component is reduced to 25Ω, and the 500kHz component is further reduced to 10Ω. The processing of the impedance measurement data reflects the interference transmission characteristics; the magnetic flux density 0 is measured at a distance of 5mm from the main circuit. .5T, and it is reduced to 0.1T by adding a 4-turn magnetic winding group. After the filter inductor is offset from the original position by 15mm, the induced voltage is reduced from 2V to 0.5V. The optimization of the resonance cavity parameters increases the spacing between adjacent devices from 5mm to 10mm, and the crosstalk attenuation is improved by 15dB. The loss matching circuit reduces the standing wave ratio at the node from 3 to 1.2. For the design of PCB, the four-layer board adopts the signal layer-ground layer-power layer-signal layer structure, and the signal line adopts the spacing rule of three times the line width. The measured crosstalk coefficient is less than 0.1. The 20mil copper foil shielding layer reduces the radiation intensity of the main circuit switch node by 80%. A 5mm wide flux suppression groove and a 0.4mm deep ground shielding hole array are added in the overlapping area, and the common mode impedance is reduced from 50Ω to 10Ω. After actual measurement of the PCB design, it is concluded that the interference voltage at the main circuit switching frequency is reduced from 2V to 0.1V, and the interference current at the harmonic frequency is reduced from 100mA to 5mA. After the wiring optimization, the crosstalk voltage dropped from 200mV to 20mV, and the radiation magnetic field strength dropped from 80dBμV / m to 50dBμV / m. The interlayer shielding reduced the formation noise from 100mV to 10mV; the above test data constituted a complete verification system, from impedance characteristics to layout effects, from interlayer crosstalk to radiation suppression, and comprehensively evaluated the implementation effect of the coupling suppression scheme. The data from multiple test dimensions corroborated each other and jointly confirmed the rationality of the design scheme. Through actual measurement verification and continuous optimization, the expected suppression target was finally achieved.

[0110] An embodiment of the present invention proposes a low-frequency power suppression method based on power module topology. An equivalent circuit model is constructed for the electromagnetic coupling between the power decoupling branch and the main power circuit of the power module, and the transfer function between the power decoupling branch and the main power circuit is further obtained. The difference in switching characteristics under different inductance parameters is analyzed in combination with electromagnetic field simulation to determine the influence of the inductance parameters on the switching characteristics of the main power circuit. The target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, as well as the transfer characteristics and influence degree of the electromagnetic interference between the power decoupling branch and the main power circuit are obtained through secondary analysis of the power module, which provides an important basis for the subsequent optimization of the power module layout and improvement of electromagnetic compatibility. According to; by designing a filter circuit to suppress electromagnetic interference, the circuit layout and wiring of the power decoupling branch and the main power circuit are optimized and adjusted. From the layout of the device, the selection of parameters and the optimization of the topology, the optimal adjustment results are obtained through strict quantitative analysis to achieve the ideal interference suppression effect of low-frequency power. The coupling suppression scheme of the power module is also designed to continuously adjust the optimized circuit topology. By verifying the effectiveness of the coupling suppression scheme, the suppression scheme is continuously improved and perfected, and the circuit layout and wiring are adjusted. Finally, through the coordination of multi-dimensional data, a reasonable optimization scheme is designed to effectively reduce the electromagnetic coupling interference between the power decoupling branch and the main power circuit and improve the electromagnetic compatibility performance of the power module.

[0111] Example 2

[0112] See also Figure 2 , Figure 2 A schematic diagram of the module structure of a low-frequency power suppression system based on a power module topology is provided in one embodiment of the present invention. Figure 2 As shown, an embodiment of the present invention proposes a method including: a transfer function construction module 201, a first simulation analysis module 202, an electromagnetic field simulation module 203, a second simulation analysis module 204, a circuit topology optimization module 205 and a circuit device adjustment module 206;

[0113] The transfer function construction module 201 is used to construct an equivalent circuit model of electromagnetic coupling between a power decoupling branch and a main power loop of a power module, and to map circuit parameters of the equivalent circuit model by a preset first neural network model to obtain a first transfer function of electromagnetic coupling between a power decoupling branch and a main power loop;

[0114] The first simulation analysis module 202 is used to perform a first simulation analysis on the electromagnetic distribution of the power module based on the first transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit by using a preset second neural network to obtain a target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit;

[0115] The electromagnetic field simulation module 203 is used to perform an influence test on the switching characteristics of the main power circuit switch device through a number of different inductance parameters based on the target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, and obtain an electromagnetic field simulation result;

[0116] The second simulation analysis module 204 is used to perform a second simulation analysis on the power module circuit based on the first transfer function, target path, target coupling strength and electromagnetic field simulation result to obtain the transfer characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power loop;

[0117] The circuit topology optimization module 205 is used to construct a filter circuit for suppressing electromagnetic interference. Based on the transmission characteristics and influence degree of the electromagnetic interference between the power decoupling branch and the main power circuit, the circuit layout and wiring of the power decoupling branch and the main power circuit are adjusted and optimized by the filter circuit for suppressing electromagnetic interference to obtain an optimized circuit topology.

[0118] The circuit device adjustment module 206 is used to construct a coupling suppression scheme for the power module based on the optimized circuit topology, and adjust device parameters until the optimized circuit topology meets the preset power suppression requirements.

[0119] A low-frequency power suppression system based on a power module topology is proposed in an embodiment of the present invention. For the electromagnetic coupling interference problem between a power decoupling branch and a main power circuit in a power module, an electromagnetic coupling equivalent circuit model is established to obtain a transfer function, coupling paths and strengths are analyzed in combination with electromagnetic field simulation, the influence of inductance parameters on the switching characteristics of the main power circuit is determined, the transfer characteristics and influence degree of electromagnetic interference are verified based on circuit simulation, the circuit layout and wiring scheme of the power module are optimized, and a suitable filter circuit is designed to suppress electromagnetic interference. Through actual test verification and continuous improvement, a complete coupling suppression scheme is finally formed, which can effectively reduce the electromagnetic coupling interference between the power decoupling branch and the main power circuit, improve the electromagnetic compatibility of the power module, and ensure the reliable operation of the switching device, and has important engineering application value.

[0120] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

[0121] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the described specific features, structures, materials or characteristics may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0122] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

Claims

1. A low-frequency power suppression method based on power module topology, characterized in that: include: Constructing an equivalent circuit model of electromagnetic coupling between a power decoupling branch and a main power loop of a power module, performing circuit parameter mapping on the equivalent circuit model by presetting a first neural network model, and obtaining a first transfer function of electromagnetic coupling between the power decoupling branch and the main power loop; Based on the first transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit, a first simulation analysis is performed on the electromagnetic distribution of the power module by using a preset second neural network to obtain a target path and a target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit; Based on the target approach and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, the switching characteristics of the main power circuit switch device are tested by using several different inductance parameters to obtain electromagnetic field simulation results; Based on the first transfer function, target path, target coupling strength and electromagnetic field simulation results, a second simulation analysis is performed on the power module circuit to obtain the transfer characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power loop; Constructing a filter circuit for suppressing electromagnetic interference, and adjusting and optimizing the circuit layout and wiring of the power decoupling branch and the main power circuit through the filter circuit for suppressing electromagnetic interference based on the transmission characteristics and influence degree of the electromagnetic interference between the power decoupling branch and the main power circuit, so as to obtain an optimized circuit topology; Based on the optimized circuit topology, a coupling suppression scheme for the power module is constructed, and device parameters are adjusted until the optimized circuit topology meets the preset power suppression requirements.

2. A low-frequency power suppression method based on power module topology as claimed in claim 1, characterized in that: The equivalent circuit model of electromagnetic coupling between the power decoupling branch and the main power loop of the power module is constructed, and the circuit parameter mapping of the equivalent circuit model is performed by presetting a first neural network model to obtain a first transfer function of electromagnetic coupling between the power decoupling branch and the main power loop, including: Based on the circuit topology of the power decoupling branch and the main power loop in the power module, the voltage and current of the main power loop and the power decoupling branch are synchronously sampled by setting a number of circuit test sites to obtain an equivalent circuit model of electromagnetic coupling between the power decoupling branch and the main power loop; Based on the equivalent circuit model of electromagnetic coupling between the power decoupling branch and the main power loop, the magnetic flux density and electromagnetic field strength values ​​at the test points corresponding to the circuit are obtained; Based on the magnetic flux density and electromagnetic field strength values, construct a second transfer function of electromagnetic coupling between the power decoupling branch and the main power loop; Based on the second transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop, the circuit parameters of the equivalent circuit model are mapped by a preset first neural network model to obtain a coupling transfer function surface diagram; Based on the coupling transfer function surface diagram, a coupling correction is performed on the second transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop to obtain a first transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop.

3. A low-frequency power suppression method based on power module topology as claimed in claim 2, characterized in that: Based on the coupling transfer function surface diagram, a coupling correction is performed on the second transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit to obtain a first transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit, including: Perform parameter combination screening on the coupling transfer function surface graph by using a preset screening algorithm to obtain the optimal parameter combination point; Based on the second transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop, calculating the transmission loss ratio of the circuit parameters corresponding to the optimal parameter combination point and the corresponding test point; Based on the transmission loss ratio, a coupling correction is performed on the second transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop to obtain a first transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop.

4. A low-frequency power suppression method based on power module topology as claimed in claim 1, characterized in that: Based on the first transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit, a first simulation analysis is performed on the electromagnetic distribution of the power module by presetting a second neural network to obtain a target path and a target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, including: According to the spatial position of the power decoupling branch and the main power loop, a three-dimensional orthogonal grid is constructed; Based on the first transfer function of the electromagnetic coupling between the power decoupling branch and the main power loop, obtaining key circuit parameters of the electromagnetic coupling between the power decoupling branch and the main power loop; Calculating the circuit magnetic vector potential value based on the three-dimensional orthogonal grid and key circuit parameters of electromagnetic coupling between the power decoupling branch and the main power loop; Based on a preset second neural network, a simulation fitting is performed on the magnetic vector potential value of the circuit to obtain spatial magnetic field distribution data of the power module; Based on the spatial magnetic field distribution data of the power module, the magnetic flux density of each distribution site is determined, and the distribution site that meets the preset density threshold is marked as a hot spot area; Based on the magnetic field distribution data of the hot spot area, the attenuation curve of the coupling transfer function is calculated; Based on the attenuation curve of the coupling transfer function, a target path and a target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power loop are obtained.

5. A low-frequency power suppression method based on power module topology as claimed in claim 1, characterized in that: Based on the target approach and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, the switching characteristics of the main power circuit switch device are tested by using several different inductance parameters to obtain electromagnetic field simulation results, including: Based on the target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power loop, a parameter scan is performed on the power decoupling branch to obtain parameter values ​​of several different inductance parameters; Based on the parameter values ​​of the several different inductance parameters, an impact test is performed on the switching characteristics of the main power circuit switch device to obtain a voltage and current stress impact curve; Calculating the switching loss value of the main power circuit switch device based on the voltage and current stress influence curve; A corresponding relationship between the several different inductance parameters and the switching loss value is constructed to obtain an electromagnetic field simulation result.

6. A low-frequency power suppression method based on power module topology as claimed in claim 1, characterized in that: Based on the first transfer function, target path, target coupling strength and electromagnetic field simulation results, a second simulation analysis is performed on the power module circuit to obtain the transfer characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power circuit, including: Based on the first transfer function, the target pathway, the target coupling strength and the electromagnetic field simulation result, constructing a simulation netlist for the circuit wiring of the power module; Based on the current change rate and the induced voltage ratio of the simulation netlist, the mutual inductance coefficient is extracted and a coupling model between the power decoupling branch and the main power loop is constructed; Based on the coupling model between the power decoupling branch and the main power loop, a power module circuit is simulated and trained by a preset deep network algorithm to obtain a simulated transfer characteristic spectrum; Based on the simulated transfer characteristic spectrum, the transfer characteristic and influence degree of the electromagnetic interference between the power decoupling branch and the main power loop are obtained.

7. A low-frequency power suppression method based on power module topology as claimed in claim 6, characterized in that: Based on the coupling model between the power decoupling branch and the main power loop, the power module circuit is simulated and trained by a preset deep network algorithm to obtain a simulated transfer characteristic spectrum, including: Dynamically predicting the switching noise of the main power circuit in the coupling model between the power decoupling branch and the main power circuit by using a preset first deep network algorithm to obtain a plurality of predicted switching noises of the main power circuit; Performing timing extraction on a plurality of predicted switching noises of the main power loop by using a preset second deep neural network algorithm to obtain switching timing parameters; Based on the switch timing parameters, a coupling model between the power decoupling branch and the main power loop is solved by a preset third deep neural network algorithm to obtain interference voltage and current waveforms; The interference voltage and current waveforms are subjected to effective value extraction and normalization processing by using a preset fourth deep neural network algorithm to obtain a simulated transfer characteristic spectrum.

8. A low-frequency power suppression method based on power module topology as claimed in claim 1, characterized in that: A filter circuit for suppressing electromagnetic interference is constructed. Based on the transmission characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power circuit, the circuit layout and wiring of the power decoupling branch and the main power circuit are adjusted and optimized by the filter circuit for suppressing electromagnetic interference to obtain an optimized circuit topology, including: Based on the electromagnetic field simulation results, obtaining magnetic flux density distribution data of the filter device in the power decoupling branch and obtaining device design coordinates based on the magnetic flux density distribution data; Based on the device design coordinates, a filter circuit layout diagram is obtained to construct a filter circuit for suppressing electromagnetic interference; Based on the transmission characteristics and influence degree of the electromagnetic interference between the power decoupling branch and the main power circuit, the circuit device parameters of the power decoupling branch and the main power circuit are adjusted by the filter circuit for suppressing electromagnetic interference until the electromagnetic interference of the circuit layout and wiring of the power decoupling branch and the main power circuit meets the preset design requirements, thereby obtaining an optimized circuit topology.

9. A low-frequency power suppression method based on power module topology as claimed in claim 1, characterized in that: Based on the optimized circuit topology, a coupling suppression scheme for the power module is constructed, and device parameters are adjusted until the optimized circuit topology meets the preset power suppression requirements, including: Based on the optimized circuit topology, a coupling suppression scheme of the power module is constructed by the filter circuit for suppressing electromagnetic interference; Based on the coupling suppression scheme of the power module, the layout and wiring of the optimized circuit topology are tested to obtain a corresponding table of test device layout positions; Based on the test device layout position correspondence table, device parameters are adjusted until the optimized circuit topology meets the preset power suppression requirements.

10. A low-frequency power suppression system based on power module topology, characterized in that: include: A transfer function building module, a first simulation analysis module, an electromagnetic field simulation module, a second simulation analysis module, a circuit topology optimization module, and a circuit device adjustment module; The transfer function construction module is used to construct an equivalent circuit model of electromagnetic coupling between a power decoupling branch and a main power loop of a power module, and to map circuit parameters of the equivalent circuit model by a preset first neural network model to obtain a first transfer function of electromagnetic coupling between a power decoupling branch and a main power loop; The first simulation analysis module is used to perform a first simulation analysis on the electromagnetic distribution of the power module based on the first transfer function of the electromagnetic coupling between the power decoupling branch and the main power circuit by means of a preset second neural network, so as to obtain a target path and a target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit; The electromagnetic field simulation module is used to test the influence of a number of different inductance parameters on the switching characteristics of the main power circuit switch device based on the target path and target coupling strength of the electromagnetic coupling between the power decoupling branch and the main power circuit, and obtain an electromagnetic field simulation result; The second simulation analysis module is used to perform a second simulation analysis on the power module circuit based on the first transfer function, target path, target coupling strength and electromagnetic field simulation result to obtain the transfer characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power circuit; The circuit topology optimization module is used to construct a filter circuit for suppressing electromagnetic interference. Based on the transmission characteristics and influence degree of electromagnetic interference between the power decoupling branch and the main power circuit, the circuit layout and wiring of the power decoupling branch and the main power circuit are adjusted and optimized through the filter circuit for suppressing electromagnetic interference to obtain an optimized circuit topology. The circuit device adjustment module is used to construct a coupling suppression scheme for the power module based on the optimized circuit topology, and adjust device parameters until the optimized circuit topology meets the preset power suppression requirements.

Images