Design method and system of GaN gate driver

Abstract

The present invention relates to the technical field of gallium nitride gate drivers, and discloses a design method and system for a gallium nitride gate driver, wherein the method comprises: collecting working signals of a gallium nitride gate driver to obtain original waveform sampling data, and establishing a signal integrity model under the action of electromagnetic interference; separating the original waveform sampling data into effective driving signals, low-frequency interference signals and high-frequency noise signals, and extracting characteristic parameters to obtain a signal characteristic parameter set; setting switching time and circuit power consumption as optimization variables, and setting constraint conditions according to the signal integrity model to obtain an optimization objective function; performing online identification and optimization calculation on driving parameters to obtain a driving control parameter sequence; performing trapezoidal point discretization processing and optimization solution to obtain target driver working parameters, and the method improves the driver's adaptability to changes in working conditions, while ensuring the stability and reliability of the driving waveform.

Description

Technical Field

[0001] The present invention relates to the technical field of gallium nitride gate drivers, and in particular to a design method and system for a gallium nitride gate driver. Background Art

[0002] Gallium nitride (GaN) power devices are increasingly used in high-frequency and high-efficiency power electronic systems due to their high breakdown voltage, low on-resistance and fast switching characteristics. However, due to the extremely fast switching speed of GaN devices, their gate drivers face serious electromagnetic interference and parasitic effects when working at high frequencies. Traditional silicon-based gate driver design methods are difficult to meet the requirements of GaN devices for drive signal integrity and switching performance.

[0003] In practical applications, GaN gate drivers often need to strike a balance between switching speed and power consumption. Too fast a switching speed will cause severe electromagnetic interference and switching oscillation, while too slow a switching speed will increase switching losses. In addition, the driver's parasitic parameters and load characteristics will change with operating conditions, which poses a challenge to the optimal design of drive parameters. Summary of the invention

[0004] The present invention provides a design method and system for a gallium nitride gate driver, which are used to improve the adaptability of the driver to changes in working conditions while ensuring the stability and reliability of the driving waveform.

[0005] In a first aspect, the present invention provides a method for designing a gallium nitride gate driver, the method for designing a gallium nitride gate driver comprising:

[0006] Collect working signals of GaN gate drivers, obtain original waveform sampling data, and establish a signal integrity model under electromagnetic interference;

[0007] Based on the signal integrity model, the original waveform sampling data is separated into an effective driving signal, a low-frequency interference signal and a high-frequency noise signal, and characteristic parameters are extracted to obtain a signal characteristic parameter set;

[0008] Based on the signal characteristic parameter set, the switching time and the circuit power consumption are set as optimization variables, and the constraint conditions are set according to the signal integrity model to obtain the optimization objective function;

[0009] An adaptive controller is constructed according to the optimization objective function, and the drive parameters are identified and optimized online to obtain a drive control parameter sequence;

[0010] The driving control parameter sequence is subjected to trapezoidal point discretization processing and optimization solution to obtain a Pareto optimal solution;

[0011] The target driver operating parameters are set based on the Pareto optimal solution, and the target driver operating parameters include a driving voltage value, a driving current value, a waveform modulation parameter, and a protection threshold parameter.

[0012] In a second aspect, the present invention provides a design system for a gallium nitride gate driver, the design system for a gallium nitride gate driver comprising:

[0013] The acquisition module is used to collect the working signal of the GaN gate driver, obtain the original waveform sampling data, and establish a signal integrity model under the action of electromagnetic interference;

[0014] An extraction module, used to separate the original waveform sampling data into an effective driving signal, a low-frequency interference signal and a high-frequency noise signal based on the signal integrity model, and perform feature parameter extraction to obtain a signal feature parameter set;

[0015] A setting module, used to set the switching time and circuit power consumption as optimization variables based on the signal characteristic parameter set, and set constraint conditions according to the signal integrity model to obtain an optimization objective function;

[0016] A calculation module is used to construct an adaptive controller according to the optimization objective function, and perform online identification and optimization calculation on the drive parameters to obtain a drive control parameter sequence;

[0017] A solution module, used for performing trapezoidal point discretization processing and optimization solution on the driving control parameter sequence to obtain a Pareto optimal solution;

[0018] A setting module is used to set target driver operating parameters based on the Pareto optimal solution, wherein the target driver operating parameters include a driving voltage value, a driving current value, a waveform modulation parameter and a protection threshold parameter.

[0019] A third aspect of the present invention provides a gallium nitride gate driver design device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the gallium nitride gate driver design device performs the above-mentioned gallium nitride gate driver design method.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned gallium nitride gate driver design method.

[0021] In the technical solution provided by the present invention, by establishing a signal integrity model under the action of electromagnetic interference and combining the wavelet transform multi-scale decomposition technology, the accurate separation of the driving signal, the interference signal and the noise signal is achieved, and accurate feature information is provided for the optimization of the driving parameters. A dual-objective optimization model is adopted to take the switching time and the circuit power consumption as optimization variables. By setting reasonable constraints, the switching loss is minimized while ensuring the driving performance. An adaptive controller is designed, and the online optimization and dynamic adjustment of the driving parameters are realized through the synergy of the parameter identification layer, the control calculation layer and the waveform generation layer. A multi-objective optimization solution strategy based on the trapezoidal point matching method and the ε-constraint method is proposed. By converting the continuous optimization problem into a discrete nonlinear programming problem, the optimal trade-off solution of the switching time and the power consumption is obtained. The hierarchical control structure and the adaptive optimization algorithm are adopted to improve the adaptability of the driver to the changes in working conditions, while ensuring the stability and reliability of the driving waveform.

[0022] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0023] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of an embodiment of a design method of a gallium nitride gate driver in an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of an embodiment of a design system for a gallium nitride gate driver according to an embodiment of the present invention;

[0026] Figure 3 FIG. 1 is a schematic diagram of an embodiment of a design device for a gallium nitride gate driver according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are 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.

[0028] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or device ends.

[0029] To facilitate understanding of this embodiment, a design method of a gallium nitride gate driver disclosed in an embodiment of the present invention is first described in detail. Figure 1 As shown, the method comprises the following steps:

[0030] 101. Collect working signals of the GaN gate driver to obtain original waveform sampling data, and establish a signal integrity model under electromagnetic interference;

[0031] It is understandable that the execution subject of the present invention may be a design system of a gallium nitride gate driver, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as an execution subject as an example.

[0032] Specifically, waveform data sampling is performed on the target nodes of the gallium nitride gate driver to obtain the original waveform data of the driver in actual operation. The target nodes include the input end of the gate drive signal, the output end of the level conversion, and the output end of the gate. These nodes are the key paths for signal flow in the gate driver circuit. By collecting the waveform data of these nodes, the dynamic changes in the circuit can be mastered. The time domain features of the original waveform data are extracted, and the time domain feature parameter set containing this information is obtained by analyzing the characteristic parameters such as the amplitude, peak value, rise time, and fall time of the waveform. At the same time, the original waveform data is fast Fourier transformed to extract the frequency domain characteristics of the signal and obtain the frequency domain feature parameter set. The frequency domain feature parameter set can help analyze the energy distribution of the signal on different frequency components and reveal the potential sources of interference and noise during the operation of the driver. The interference source is identified according to the frequency domain feature parameter set, and the interference source is classified into common mode interference source and differential mode interference source through spectrum energy distribution analysis. Classification can accurately determine the nature and location of the interference source and provide the necessary basic data for the establishment of subsequent models. After completing the classification of interference sources, an equivalent model of the transmission channel is established based on the results. By analyzing the distributed parameters in the circuit, such as resistance, inductance, capacitance and the coupling coefficient between them, the characteristic parameters of the transmission channel are calculated to reflect the propagation characteristics and interference suppression capabilities of the signal in the channel. The transmission channel characteristic parameters and the time domain characteristic parameter set are input into the signal propagation analysis model, and the propagation characteristic data of the signal under different electromagnetic interference conditions are calculated by solving the electromagnetic field equation. The electromagnetic field equation calculation simulates the behavior of the signal in a complex circuit environment, including reflection, attenuation and coupling, so as to comprehensively analyze the impact of electromagnetic interference on signal propagation. The characteristics of the driver load port are calculated according to the signal propagation characteristic data to obtain the load port model. The load port model can describe the response behavior of the driver under different load conditions. By calculating the load port model and the signal propagation characteristic data together, the signal behavior of the driver during operation is accurately simulated, and a signal integrity model containing complete signal transmission characteristics is obtained.

[0033] 102. Based on the signal integrity model, the original waveform sampling data is separated into an effective driving signal, a low-frequency interference signal and a high-frequency noise signal, and characteristic parameters are extracted to obtain a signal characteristic parameter set;

[0034] Specifically, the wavelet transform technology is used to perform multi-scale decomposition on the original waveform sampling data, decompose the signal into different time and frequency scales, and obtain the wavelet decomposition coefficients at each scale. These wavelet decomposition coefficients can accurately capture the local characteristics of the signal and provide detailed distribution information of the signal in the frequency domain. Based on the wavelet decomposition coefficients, the original waveform sampling data is divided according to the frequency band. Among them, the frequency band from DC to 100kHz is defined as the effective drive signal, which mainly reflects the effective components of the core work of the gate driver; the frequency band from 100kHz to 1MHz is defined as the low-frequency interference signal, which is caused by power supply fluctuations or external interference; the frequency band above 1MHz is defined as the high-frequency noise signal, which is mostly related to switching operations or high-frequency parasitic effects. Through frequency band division, the original waveform sampling data is accurately separated into three main components, so that the source and characteristics of each signal can be analyzed independently. For the separated effective drive signal, the gate drive feature analysis is performed. By extracting key feature data such as the rise time, fall time, switch overshoot value and steady-state on-voltage of the gate voltage, the performance and quality of the drive signal can be accurately characterized. At the same time, these characteristic data are of great significance for evaluating the switching speed, power efficiency and steady-state performance of the driver. For low-frequency interference signals, the envelope extraction method is used to obtain the envelope characteristics of the interference signal. The envelope characteristics clearly reflect the amplitude variation law of the interference signal, thereby helping to identify and suppress low-frequency interference. For high-frequency noise signals, the energy distribution is calculated. By analyzing the distribution of noise energy in different frequency bands, the main components and influence range of the noise can be accurately located, providing a basis for the formulation of noise suppression strategies. The driving signal characteristic data, interference signal envelope characteristics and noise characteristic parameters are fused to form a comprehensive signal characteristic description. The fused data is normalized so that different types of characteristic parameters can be compared and optimized on the same scale. Normalization eliminates the impact of dimension on data analysis and can improve the applicability and versatility of the signal characteristic parameter set.

[0035] 103. Based on the signal characteristic parameter set, the switching time and the circuit power consumption are set as optimization variables, and the constraint conditions are set according to the signal integrity model to obtain the optimization objective function;

[0036] Specifically, the switching time feature extraction is performed on the signal characteristic parameter set. By analyzing the time required for the gate voltage to rise from 10% to 90% and the time required for the gate voltage to drop from 90% to 10%, the initial data of the gate switching time is obtained to characterize the dynamic response capability of the gate drive. Based on the initial data of the gate switching time, a mapping model between the switching time and the driving parameters is established. In this model, the switching time is used as the target variable and is associated with parameters such as the driving voltage, current and waveform shape to obtain the target function of the switching time. At the same time, the power consumption of the driving waveform is analyzed based on the signal integrity model to calculate the static power consumption data of the driving circuit. The static power consumption comes from the energy consumption of the circuit when there is no dynamic switching. On the basis of obtaining the static power consumption data, the dynamic loss is calculated to reflect the energy consumption generated by the driver during the rapid change of the gate voltage and current. By combining the static power consumption and the dynamic loss, the target function of the circuit power consumption is established. The switching time target function and the circuit power consumption target function are weighted to form a dual-objective optimization model. According to the signal integrity model, the limit range of the driving voltage and driving current is set. The limit range of the driving voltage needs to ensure that the GaN device can be effectively turned on and off, while avoiding device damage or signal distortion caused by excessive voltage; the limit range of the driving current needs to ensure that the gate charging and discharging speed is within a reasonable range, neither causing current spikes nor ensuring switching efficiency. While determining these limit ranges, the waveform quality constraint analysis is performed on the signal characteristic parameter set, focusing on the analysis of waveform overshoot, ringing, rise time and fall time, etc., to ensure that the quality of the driving signal meets the design requirements and form signal quality constraints. The driving voltage limit range, driving current limit range and signal quality constraints are substituted into the dual-objective optimization model to obtain the final optimization objective function. This function can coordinately optimize the switching time and power consumption while meeting the signal integrity and power consumption requirements, thereby achieving efficient and stable operation of the GaN gate driver.

[0037] 104. Construct an adaptive controller according to the optimization objective function, and perform online identification and optimization calculation on the drive parameters to obtain a drive control parameter sequence;

[0038] Specifically, the optimization objective function is input into the parameter identification layer of the adaptive controller, which uses the recursive least squares algorithm to identify the driving voltage and current values ​​in real time. The recursive least squares algorithm estimates the dynamic characteristics of the driving parameters by continuously analyzing the input signal and output response, and calculates the root mean square value of the parameter estimation error to judge the convergence of the identification process. After the parameter identification converges, the initial estimated value of the driving parameter is obtained. The adaptive weight calculation is performed on the initial estimated value of the driving parameter. By introducing the exponentially weighted recursive least squares algorithm with a forgetting factor of 0.95, the parameter identification can dynamically track the changes in the working state of the driver. The setting of the forgetting factor ensures that the algorithm can respond more sensitively to the changes in the current data on the basis of long-term tracking of historical data, thereby improving the accuracy and real-time performance of the identification. Through this process, a more accurate driving parameter identification result is obtained to reflect the transient response characteristics of the driver. The driving parameter identification result is input into the control calculation layer of the adaptive controller. This layer uses a nonlinear model predictive control algorithm to make full use of the nonlinear characteristics of the driver. By setting the prediction time domain to 10 sampling cycles and the control time domain to 5 sampling cycles, the algorithm can predict the dynamic behavior of the driver in a rolling manner, optimize the driving parameters in real time, and generate an optimized control sequence. The predictive control algorithm ensures the control effect of the driver in a dynamic environment by re-optimizing the objective function in each sampling cycle. In order to make the optimized control sequence meet the requirements of actual hardware implementation, constraints are processed. By setting the hard constraints that the driving voltage change rate does not exceed 1V / μs and the driving current change rate does not exceed 0.5A / μs, it is ensured that the generated control quantity is within the hardware achievable range, while avoiding overshoot or ringing caused by rapid changes. After completing the constraint processing, the control quantity that meets the constraints is input into the waveform generation layer of the adaptive controller. The waveform generation layer uses a piecewise linear interpolation algorithm to smoothly reconstruct the driving waveform by calculating the slope coefficient between adjacent sampling points. Piecewise linear interpolation can simplify the calculation and ensure the continuity and smoothness of the waveform, generate a target driving waveform that meets the design requirements, and reflects the ideal working state of the driver under current conditions. The generated target driving waveform is input into the closed-loop controller, and the control parameters are corrected online by comparing the error between the actual driving waveform and the target waveform. The error feedback control mechanism can improve the robustness and accuracy of the system. After the error correction is completed, the parameters of the driver are reconstructed based on the corrected control amount to obtain the final drive control parameter sequence.

[0039] 105. Perform trapezoidal point discretization and optimization on the drive control parameter sequence to obtain the Pareto optimal solution;

[0040] Specifically, the time interval distribution point processing is performed on the drive control parameter sequence. According to the working frequency of the switching scenario, the time period is reasonably divided: in the high-frequency switching scenario, the switching cycle is divided into 50 equally spaced sub-intervals to ensure the capture of the fast-changing parameter dynamic characteristics; in the low-frequency switching scenario, the switching cycle is divided into 20 equally spaced sub-intervals to optimize the calculation efficiency and retain sufficient time domain accuracy. Through the division, the time discrete distribution points are obtained. On the basis of the time discrete distribution points, the Lagrangian interpolation calculation is performed on the drive control parameter sequence. By constructing the third-order Lagrangian basis function, the driving voltage value and the driving current value are piecewise interpolated and fitted, so that the change trend of these driving parameters between discrete distribution points can be smoothly described. The Lagrangian interpolation method can improve the accuracy of parameter fitting and effectively avoid the distortion caused by data discretization, and obtain an accurate parameter discrete mapping function to reflect the change law of the driving parameters of the driver at different time points. The parameter discrete mapping function is numerically integrated in a trapezoidal shape. By calculating the product of the mean value of the driving parameters of adjacent distribution points and the time step, the state equation is discretized to form a discrete state equation group. The trapezoidal numerical integration method can simplify the solution process of the continuous state equation while maintaining the calculation accuracy, so that it can adapt to the requirements of discrete optimization solution. The obtained discrete state equation group contains the dynamic information of the driving voltage and driving current, and can describe the key characteristics of signal propagation and energy conversion. The discrete state equation group, the switching time objective function and the circuit power consumption objective function are substituted into the nonlinear programming model to construct a nonlinear programming equation group containing driving voltage constraints, driving current constraints and waveform quality constraints. In this equation group, the range of driving voltage and current is limited to the range allowed by the hardware, and the waveform quality constraint is used to ensure the integrity and stability of the signal during the switching process. By introducing these constraints, the possibility of the optimization results deviating from the actual engineering application can be effectively avoided. Based on the nonlinear programming equation group, ε-constraint solution pre-processing is performed. The circuit power consumption objective function is converted into a parameterized constraint condition, and a series of constraint parameter sequences are set to make it an auxiliary condition of the optimization model; at the same time, the switching time objective function is used as the main optimization objective to ensure that the optimization result can maximize the dynamic response capability of the driver. Through this processing, a solution framework for multi-constraint optimization problems is established, making the trade-off relationship between different objective functions clearer. The constraint parameter sequence is input into the interior point optimization solver, and the single objective optimization problem under each constraint condition is solved using the barrier function method. The interior point method can effectively handle complex constraints by introducing barrier terms in the objective function, and ensure the convergence and stability of the optimization results while maintaining the solution speed. Through this process, a series of optimal solution sequences are obtained, each of which corresponds to the best driving parameters under a specific constraint parameter combination. The Pareto optimality judgment is performed on the optimal solution sequence.By calculating the relative gain coefficient of switching time and circuit power consumption, the optimization trade-off between the two can be quantitatively analyzed; at the same time, the non-dominated sorting method is used to screen the solution set, and those solutions that achieve the optimal balance between circuit power consumption and switching time are extracted to obtain the Pareto optimal solution.

[0041] The circuit power consumption objective function in the nonlinear programming equations is Benchmarking is performed. It is a parameter vector containing the driving voltage value and the driving current value. By calculating the circuit power consumption and the minimum driving power consumption The ratio of , and obtain the benchmark power consumption objective function Based on the benchmarked power consumption objective function , construct inequality constraints. By setting the upper limit constraint of circuit power consumption and lower bound constraints , we get the power consumption constraint boundary equations .in, is the standardized power consumption constraint parameter, which is used to control the range of power consumption limit. The setting of upper and lower limits ensures that the optimization problem strikes a balance between the efficiency of power consumption and the actual engineering requirements. Perform component decoupling. Decomposed into opening time components and the off-time component , and by introducing the weight coefficient and Constructing the weighted sum objective function The main optimization objective function is It is possible to comprehensively consider different stages of switching time and dynamically balance their importance during the optimization process. With the main optimization objective function Combine them to build a complete parametric planning model. By introducing the solution vector ,in is the Lagrange multiplier vector, and contains the equality constraints and inequality constraints The Karush-Kuhn-Tucker (KKT) condition of the optimization solution is formed to form a complete optimization solution framework. The application of the KKT condition ensures that the main optimization objective function is minimized while satisfying all constraints. Perform discrete sampling to generate a Parameter sampling sequence. By traversing each sampling value of the sampling sequence, a set of constrained optimization sub-problems is constructed. Each sub-problem represents a specific The optimization problem under the parameter value is substituted into the parameterized planning model to generate a complete set of optimization problems. Constraint gradient calculation is performed on the optimization problem group. By solving the constraint matrix and the objective function gradient, the set of constraint sensitivity matrices for each optimization problem is obtained. The sensitivity matrix reflects the sensitivity of the constraints to the changes in the optimization variables. Constraint feasibility is judged based on the set of constraint sensitivity matrices. By calculating the rank of each Jacobian matrix, it is checked whether it meets the linear independence constraint qualification (LICQ) condition. Only the constraint parameter sequence that meets the LICQ condition can pass the screening and is considered to be a valid and feasible result in the optimization process. Through the above steps, the constraint parameter sequence finally generated not only meets the optimization goal, but also meets the constraint requirements in practical applications.

[0042] 106. Set target driver operating parameters based on the Pareto optimal solution, where the target driver operating parameters include a driving voltage value, a driving current value, a waveform modulation parameter, and a protection threshold parameter.

[0043] Specifically, the target driver operating parameters are set based on the Pareto optimal solution. The target driver operating parameters include driving voltage value, driving current value, waveform modulation parameters and protection threshold parameters. By analyzing the solution on the Pareto frontier, the trade-off information between multiple objectives such as circuit power consumption, switching time and signal integrity is obtained. The setting of the driving voltage value is based on the performance requirements of the driver under different loads and switching frequencies. By analyzing the distribution of driving voltage values ​​in the Pareto optimal solution, the optimal voltage range between signal integrity and energy efficiency in high-frequency and low-frequency scenarios is identified. In order to balance dynamic performance and stability, the driving voltage is set in the high-efficiency range within the device safe operating area, while avoiding overshoot or ringing caused by excessive voltage. The setting of the driving current value needs to comprehensively consider the limitations of the gate charging and discharging speed and the current change rate. Through the Pareto optimal solution, the range of the driving current is determined to achieve a sufficiently fast gate charge transfer speed while suppressing the noise and heat dissipation caused by rapid current changes. Based on this, the driving current value is set within the minimum current range that meets the switching speed requirements, which can reduce circuit power consumption and ensure signal waveform quality. The setting of waveform modulation parameters requires optimizing signal integrity. By modulating the amplitude, frequency and shape of the waveform, the ringing and overshoot caused by electromagnetic interference during the switching process can be reduced. Combined with the optimization results of waveform quality in the Pareto optimal solution, the key parameters of waveform modulation, such as modulation frequency and slope variation range, are determined to improve signal stability in dynamic environments. The setting of protection threshold parameters is directly related to the safety and reliability of the driver. By analyzing the power consumption distribution and boundary conditions of the voltage and current waveforms of the Pareto optimal solution, the optimal thresholds of overcurrent protection, overvoltage protection and overtemperature protection are determined. The selection of protection thresholds must ensure that the driver can still maintain stable operation without damaging the device under extreme working conditions. Especially in complex electromagnetic environments, these protection parameters can effectively shield unexpected failures caused by external interference. The operating parameters of the target driver are generated by comprehensively analyzing the driving voltage, driving current, waveform modulation parameters and protection threshold parameters in the Pareto optimal solution.

[0044] In the embodiment of the present invention, by establishing a signal integrity model under the action of electromagnetic interference and combining the wavelet transform multi-scale decomposition technology, the accurate separation of the driving signal, the interference signal and the noise signal is achieved, and accurate feature information is provided for the optimization of the driving parameters. The dual-objective optimization model is adopted to take the switching time and the circuit power consumption as the optimization variables. By setting reasonable constraints, the switching loss is minimized while ensuring the driving performance. An adaptive controller is designed, and the online optimization and dynamic adjustment of the driving parameters are realized through the synergy of the parameter identification layer, the control calculation layer and the waveform generation layer. A multi-objective optimization solution strategy based on the trapezoidal point matching method and the ε-constraint method is proposed. By converting the continuous optimization problem into a discrete nonlinear programming problem, the optimal trade-off solution of the switching time and power consumption is obtained. The hierarchical control structure and the adaptive optimization algorithm are adopted to improve the adaptability of the driver to the changes in working conditions, while ensuring the stability and reliability of the driving waveform.

[0045] In a specific embodiment, the process of executing step 101 may specifically include the following steps:

[0046] Sampling waveform data of a target node of a gallium nitride gate driver to obtain original waveform sampling data, wherein the target node includes a gate drive signal input terminal, a level conversion output terminal, and a gate output terminal;

[0047] Performing time domain feature extraction on the original waveform sampling data to obtain a time domain feature parameter set, and performing fast Fourier transform on the original waveform sampling data to obtain a frequency domain feature parameter set;

[0048] Interference sources are identified based on the frequency domain characteristic parameter set, and interference sources are divided into common mode interference sources and differential mode interference sources by analyzing the spectrum energy distribution to obtain the interference source classification result;

[0049] Based on the interference source classification results, an equivalent model of the transmission channel is established, and the distribution parameters and coupling coefficients are calculated to obtain the characteristic parameters of the transmission channel;

[0050] Input the transmission channel characteristic parameters and the time domain characteristic parameter set into the signal propagation analysis model to solve the electromagnetic field equations and obtain the signal propagation characteristic data;

[0051] The load port characteristics of the driver are calculated according to the signal propagation characteristic data to obtain a load port model, and the load port model and the signal propagation characteristic data are jointly calculated to obtain a signal integrity model.

[0052] Specifically, waveform data sampling is performed on key target nodes of the GaN gate driver, including the gate drive signal input, level conversion output, and gate output. The signal changes of these nodes are recorded by a high-precision oscilloscope or high-speed sampling device, and an original sampling data set containing transient voltage and current waveforms is formed. These data are stored in the form of a time series. Time domain feature extraction is performed on the collected original waveform data. The extraction of time domain features includes calculating key parameters, such as the rise time of the signal. , Fall time , Peak voltage , average current Etc. These parameters are calculated by the following formulas, for example, the rise time is defined as:

[0053] ;

[0054] in, and The peak voltage represents the maximum voltage in the waveform, and the average current is calculated by integrating the current waveform over time:

[0055] ;

[0056] After the time domain feature extraction is completed, the original waveform data is subjected to fast Fourier transform to convert the time domain signal into the frequency domain signal. This conversion is achieved by the following formula:

[0057] ;

[0058] in, is the amplitude distribution of the frequency domain signal, is the time domain signal, is the frequency, is an imaginary unit. Through fast Fourier transform, the characteristic parameter set in the spectrum is extracted, such as the main frequency, bandwidth and noise power distribution. After obtaining the frequency domain characteristic parameter set, the spectrum energy distribution is analyzed to identify the interference source. By comparing the energy distribution within a specific frequency range, the interference source is divided into common mode interference and differential mode interference. Common mode interference refers to interference with the same polarity and amplitude on two conductors. Its energy is mainly concentrated in the low frequency band. It is identified by calculating the common mode voltage:

[0059] ;

[0060] Differential mode interference refers to the difference signal between two conductors. Its energy is distributed in the high frequency band and is calculated by differential mode voltage:

[0061] ;

[0062] According to the classification results of interference sources, an equivalent model of the transmission channel is established. Transmission channel modeling requires the calculation of distributed parameters (such as resistance ,inductance ,capacitance and admittance ) and the coupling coefficient These parameters are calculated using the following formulas, for example, for the mutual inductance coupling coefficient between two conductors:

[0063] ;

[0064] in, For mutual induction, and is the self-inductance of the two conductors. Using the transmission channel characteristic parameters and the time domain characteristic parameter set, the signal propagation analysis model is input, and the signal propagation characteristic data is calculated by solving the electromagnetic field equation. The electromagnetic field equation is described by Maxwell's equations. For example, the electric field and magnetic field in propagation satisfy the following relationship:

[0065] ;

[0066] in, is the electric field strength, is the magnetic induction intensity, is the magnetic field strength, is the current density, is the electric field displacement. By calculating these equations, we can obtain the propagation characteristic data of the signal in the transmission channel, such as reflection coefficient, transmission delay, etc. Based on these data, we can calculate the load port characteristics of the driver. The load port characteristics describe the dynamic behavior of the driver when it is connected to the load, such as input impedance and reflection coefficient. The input impedance is calculated by the following formula:

[0067] ;

[0068] Combining the load port model with the signal propagation characteristic data, a complete signal integrity model is established through joint calculation. The signal integrity model can fully reflect the working characteristics of the driver in a complex electromagnetic environment.

[0069] In a specific embodiment, the process of executing step 102 may specifically include the following steps:

[0070] Based on the signal integrity model, the original waveform sampling data is decomposed by wavelet transform at multiple scales to obtain the wavelet decomposition coefficients at each scale.

[0071] The original waveform sampling data is divided into frequency bands according to the wavelet decomposition coefficients, where the DC to 100kHz frequency band corresponds to the effective drive signal, the 100kHz to 1MHz frequency band corresponds to the low-frequency interference signal, and the frequency band above 1MHz corresponds to the high-frequency noise signal;

[0072] Performing gate drive characteristic analysis on the effective drive signal to obtain drive signal characteristic data, the drive signal characteristic data including gate voltage rise time, fall time, switch overshoot value and steady-state on-voltage;

[0073] Perform envelope extraction on the low-frequency interference signal to obtain the interference signal envelope characteristics, and perform energy distribution calculation on the high-frequency noise signal to obtain the noise characteristic parameters;

[0074] The driving signal characteristic data, interference signal envelope characteristics and noise characteristic parameters are subjected to data fusion and normalization processing to obtain a signal characteristic parameter set.

[0075] Specifically, the original waveform sampling data is subjected to multi-scale decomposition by wavelet transform. The time-frequency localization characteristics of wavelet transform are used to decompose the signal into different scales, and the wavelet decomposition coefficients at each scale are extracted. Wavelet transform is expressed by the following formula:

[0076] ;

[0077] in, is the wavelet transform coefficient, is the original waveform sampling data, is the mother wavelet function, is the scale parameter (controls the frequency resolution), is the translation parameter (controlling time resolution), and * represents complex conjugate. By selecting a suitable mother wavelet function (such as Daubechies wavelet), the signal is decomposed into multiple scales corresponding to signal components in different frequency bands. Based on the wavelet decomposition coefficients, the original waveform sampling data is divided into frequency bands. By selecting the coefficient range of a specific scale, the frequency band from DC to 100kHz is extracted as the effective drive signal, the frequency band from 100kHz to 1MHz is extracted as the low-frequency interference signal, and the frequency band above 1MHz is extracted as the high-frequency noise signal. By using the frequency band separation ability of wavelet decomposition, the main components of various signals are clarified by analyzing the energy distribution at each scale. For example, assuming that the main energy of an original waveform is concentrated at 50kHz, the low-frequency interference energy is concentrated at 500kHz, and the high-frequency noise energy is concentrated above 2MHz, these frequency components are clearly identified and separated by wavelet decomposition. The gate drive characteristics of the separated effective drive signal are analyzed to extract the key performance indicators of the gate driver. The gate voltage rise time and fall time represent the time required for the signal to change from 10% to 90% amplitude and from 90% to 10% amplitude respectively. The formula is as follows:

[0078] ;

[0079] in, and The switching overshoot is defined as the amount by which the voltage or current exceeds the steady-state value during switching, and is calculated using the following formula:

[0080] ;

[0081] in, is the maximum value of the instantaneous voltage, is the steady-state voltage. The steady-state on-state voltage represents the average voltage value in the stable stage of the switch. For low-frequency interference signals, the envelope characteristics of the interference signal are obtained through envelope extraction technology. Envelope extraction uses Hilbert transform, which is defined as:

[0082] ;

[0083] in, is the Hilbert transform of the signal, is the envelope signal. The envelope characteristic reflects the amplitude variation law of the low-frequency interference signal and can reveal the periodicity and intensity of the interference. For high-frequency noise signals, the energy distribution parameters are calculated to analyze their impact on system performance. The energy distribution of high-frequency noise is calculated using the following formula:

[0084] ;

[0085] in, is the signal energy, is the frequency domain signal, and is the frequency range of the noise signal. This energy distribution can help evaluate the degree of interference of high-frequency noise on signal integrity. After completing the above feature extraction, the driving signal feature data, interference signal envelope features and noise feature parameters are fused and normalized to form a unified signal feature parameter set. Normalization is achieved through the following formula:

[0086] ;

[0087] in, is the normalized eigenvalue, is the original eigenvalue, and are the minimum and maximum values ​​of the characteristic values ​​respectively. The normalized signal characteristic parameter set is used for the optimal design of the driver.

[0088] In a specific embodiment, the process of executing step 103 may specifically include the following steps:

[0089] The switching time features are extracted from the signal characteristic parameter set. The initial gate switching time data is obtained by calculating the time required for the gate voltage to rise from 10% to 90% and the time required for the gate voltage to drop from 90% to 10%.

[0090] A mapping model between switching time and driving parameters is established based on the initial data of gate switching time to obtain the switching time target function, and the power consumption of the driving waveform is analyzed based on the signal integrity model to obtain the static power consumption data of the driving circuit;

[0091] The dynamic loss is calculated according to the static power consumption data of the driving circuit to obtain the circuit power consumption objective function, and the switching time objective function and the circuit power consumption objective function are weighted to obtain a dual-objective optimization model;

[0092] The driving voltage limit range and the driving current limit range are set according to the signal integrity model, and the waveform quality constraint analysis is performed on the signal characteristic parameter set to obtain the signal quality constraint condition;

[0093] The driving voltage limit range, driving current limit range and signal quality constraints are substituted into the dual-objective optimization model to obtain the optimization objective function.

[0094] Specifically, the switching time feature extraction is performed on the signal feature parameter set. Based on the dynamic change of the gate voltage, the initial data of the gate switching time is obtained by calculating the time required for the gate voltage to rise from 10% to 90% and the time required for the gate voltage to fall from 90% to 10%. These time features are calculated using the following formula:

[0095] ;

[0096] in, and Represent the time when the gate voltage rises and falls, and These are the time points when the gate voltage reaches 90% and 10% of its amplitude, respectively. By measuring these key time parameters, a basic switching time feature data set is formed. Based on the initial gate switching time data, a mapping model between switching time and drive parameters is established. Gate switching time and drive voltage , drive current and load capacitance The following relationship exists:

[0097] ;

[0098] in, represents the gate switching time, is the load capacitance connected to the gate driver, is the driving voltage, is the driving current. Through experimental measurement and fitting, this mapping model is constructed and described in mathematical form to generate the switching time objective function ,in is the driving parameter vector. Based on the signal integrity model, the driving waveform is analyzed for power consumption to obtain the static power consumption data of the driving circuit. The static power consumption mainly comes from the leakage current loss of the circuit under steady-state conditions and the maintenance power consumption of the gate drive. The calculation formula is:

[0099] ;

[0100] in, is the static power consumption, is the leakage current. Combined with the actual parameters of the gate driver, the corresponding static power consumption value is obtained through testing and model analysis. Based on the static power consumption data, the dynamic loss is calculated to fully describe the total power consumption characteristics of the circuit. The dynamic loss is mainly generated by the energy supply of the drive circuit to the gate capacitor during the charging and discharging process. The formula is:

[0101] ;

[0102] in, is the dynamic power consumption, is the switching frequency, and The load capacitance and driving voltage are respectively. Through the comprehensive analysis of static power consumption and dynamic loss, the circuit power consumption objective function is constructed. . The switching time objective function And circuit power consumption objective function Combine them and assign weights to them to build a dual-objective optimization model. The objective function of the optimization model is expressed as:

[0103] ;

[0104] in, and are weight coefficients, which reflect the optimization preferences for switching speed and energy consumption respectively. By adjusting the weights, different optimization strategies between performance and power consumption can be achieved. While building the optimization model, the limit range of driving voltage and driving current is set according to the signal integrity model. For example, the range of driving voltage is limited to , the driving current range is , to ensure safe operation. The signal characteristic parameter set is subjected to constraint analysis of waveform quality, including limiting key indicators such as overshoot, ringing, and settling time. These constraints are expressed by the following inequality conditions:

[0105] ;

[0106] in, is the waveform overshoot value, is the maximum overshoot threshold allowed, is the signal stabilization time, is the maximum allowable stabilization time. Substitute the driving voltage limit range, driving current limit range and waveform quality constraints into the dual-objective optimization model to obtain the final optimization objective function. By solving this optimization objective function, the optimal driving parameters that meet signal integrity, power consumption requirements and dynamic performance are determined.

[0107] In a specific embodiment, the process of executing step 104 may specifically include the following steps:

[0108] The optimization objective function is input into the parameter identification layer of the adaptive controller. The parameter identification layer uses a recursive least squares algorithm to identify the driving voltage and driving current in real time, and calculates the root mean square value of the parameter estimation error to make a convergence judgment, and obtains the initial estimated value of the driving parameter.

[0109] The initial estimated values ​​of the driving parameters are adaptively weighted, and an exponentially weighted recursive least squares algorithm with a forgetting factor of 0.95 is constructed to dynamically track the driving parameters to obtain the driving parameter identification results.

[0110] The identification results of the driving parameters are input into the control calculation layer of the adaptive controller. The control calculation layer adopts the nonlinear model predictive control algorithm. By setting the prediction time domain to 10 sampling periods and the control time domain to 5 sampling periods, the driving parameters are optimized in a rolling manner to obtain the optimized control sequence.

[0111] The optimized control sequence is constrained, and the control quantity that meets the constraints is obtained by setting the hard constraints that the driving voltage change rate does not exceed 1V / μs and the driving current change rate does not exceed 0.5A / μs;

[0112] The control quantity that meets the constraints is input into the waveform generation layer of the adaptive controller. The waveform generation layer uses a piecewise linear interpolation algorithm to reconstruct the driving waveform by calculating the slope coefficient between adjacent sampling points to obtain the target driving waveform.

[0113] The target driving waveform is input into the closed-loop controller. By comparing the error between the actual driving waveform and the target waveform, the control parameters are corrected online to obtain the corrected control quantity. The corrected control quantity is then parameter-reconstructed to obtain a driving control parameter sequence.

[0114] Specifically, the optimization objective function is input into the parameter identification layer of the adaptive controller. The parameter identification layer uses a recursive least squares algorithm to identify the drive voltage and drive current values ​​in real time to capture the changing characteristics of the drive parameters in dynamic operation. The recursive least squares algorithm minimizes the sum of squared errors by continuously updating the parameter estimates. Its mathematical expression is:

[0115] ;

[0116] in, is the parameter estimation vector at the current moment, including the estimated values ​​of the driving voltage and driving current; is the input signal vector, which contains the state variables of the system; is the output signal; is the gain matrix, which is calculated by the following formula:

[0117] ;

[0118] ;

[0119] in, is the covariance matrix, is the forgetting factor. The root mean square value of the parameter estimation error is calculated as the convergence criterion:

[0120] ;

[0121] When the RMS value is lower than the preset threshold, the identification process is considered to have converged and the initial estimated value of the driving parameter is obtained. On this basis, the initial estimated value is adaptively weighted. By introducing an exponentially weighted recursive least squares algorithm with a forgetting factor of 0.95, the tracking ability of dynamic changes is enhanced. The update formula is:

[0122] ;

[0123] This improvement emphasizes the weight of recent data while reducing the impact of historical data, making the drive parameter estimation more accurate. Through multiple iterations, the identification results of the drive parameters are obtained, including the instantaneous drive voltage and drive current values. The drive parameter identification results are input into the control calculation layer of the adaptive controller. The control calculation layer uses the nonlinear model predictive control (NMPC) algorithm to predict the system state at future moments and optimize the drive parameters in a rolling manner. NMPC focuses on the following optimization problems:

[0124] ;

[0125] in, is the system state vector, is the target reference state, is the control input (drive voltage and drive current), is the prediction time domain length (set to 10 sampling periods), and is the weight matrix. By setting the control time domain to 5 sampling periods, the optimized control sequence is recalculated in each time domain. The optimized control sequence needs to be constrained to ensure the feasibility of physical implementation. The hard constraints are set as follows:

[0126] ;

[0127] The optimized control sequence is modified by the constrained optimization method (such as the projected gradient method) to obtain the control quantity that meets the constraint conditions. The control quantity that meets the constraint is input into the waveform generation layer of the adaptive controller, and the waveform generation layer reconstructs the driving waveform through the piecewise linear interpolation algorithm. Assume that the values ​​of two adjacent sampling points are and , the interpolation formula is:

[0128] ;

[0129] The slope coefficients of adjacent points are calculated by interpolation to achieve continuity and smoothness of the drive waveform, thereby generating the target drive waveform. The target drive waveform is input into the closed-loop controller. The closed-loop controller corrects the error between the actual drive waveform and the target waveform online, and its correction strategy is based on the following update formula:

[0130] ;

[0131] in, is the corrected control quantity, is the error gain matrix, is the current error (the difference between the target waveform and the actual waveform). By adjusting the control parameters in real time, the error is gradually reduced to generate the corrected control quantity. The corrected control quantity is reconstructed to obtain the final drive control parameter sequence.

[0132] In a specific embodiment, the process of executing step 105 may specifically include the following steps:

[0133] The drive control parameter sequence is processed by time interval point distribution. The switching cycle is divided into 50 equally spaced sub-intervals in the high-frequency switching scenario and the switching cycle is divided into 20 equally spaced sub-intervals in the low-frequency switching scenario to obtain time discrete point distribution.

[0134] Based on the time discrete collocation point, the driving control parameter sequence is calculated by Lagrangian interpolation, and the driving voltage value and the driving current value are piecewise interpolated and fitted by constructing the third-order Lagrangian basis function to obtain the parameter discrete mapping function;

[0135] The parameter discrete mapping function is numerically integrated in a trapezoidal manner, and the state equation is discretized by calculating the product of the mean value of the driving parameters of adjacent points and the time step to obtain a discrete state equation group.

[0136] Substitute the discrete state equations, the switching time objective function and the circuit power consumption objective function into the nonlinear programming model to construct a nonlinear programming equation group including driving voltage constraints, driving current constraints and waveform quality constraints;

[0137] Based on the nonlinear programming equations, the ε-constraint solution pre-processing is performed to obtain the constraint parameter sequence by converting the circuit power consumption objective function into parameterized constraint conditions and taking the switching time objective function as the main optimization target.

[0138] Input the constraint parameter sequence into the interior point optimization solver, solve the single objective optimization problem under each constraint parameter by using the barrier function method, and obtain the optimal solution sequence;

[0139] The Pareto optimality of the optimal solution sequence is determined. The relative gain coefficient of switching time and circuit power consumption is calculated, and the optimal trade-off solution between circuit power consumption and switching time is screened out using the non-dominated sorting method to obtain the Pareto optimal solution.

[0140] Specifically, the drive control parameter sequence is processed with time interval points to meet the time resolution requirements of different switching scenarios. In high-frequency switching scenarios, the signal changes quickly, and the switching cycle is divided into 50 equally spaced sub-intervals to capture fast dynamic characteristics; in low-frequency switching scenarios, since the signal changes relatively slowly, it is divided into 20 equally spaced sub-intervals. Assuming the switching cycle is , the time discrete collocation is expressed as:

[0141] ;

[0142] in, is the time step, is the number of subintervals (50 for high frequency and 20 for low frequency), For the The time coordinates of the points are obtained through this step. Based on the time discrete points, the Lagrangian interpolation calculation is performed on the drive control parameter sequence to achieve the segmented interpolation fitting of the drive voltage value and the drive current value. The Lagrangian interpolation formula is:

[0143] ;

[0144] in, is the interpolation function, and are the time coordinates of the known data points, is the corresponding driving parameter value. By constructing the third-order Lagrangian basis function, local interpolation is performed on every three adjacent matching points to obtain the continuous fitting curve of the driving parameters and generate the parameter discrete mapping function. The parameter discrete mapping function is numerically integrated in a trapezoidal manner to discretize the state equation. The trapezoidal numerical integration approximates the integral value of the state equation by calculating the product of the mean driving parameter of the adjacent matching points and the time step. The calculation formula is:

[0145] ;

[0146] The state equation of the actuator is discretized by integrating and accumulating over all time intervals. For example, for the state variable , its discretized form is:

[0147] ;

[0148] in, and are the state variables at adjacent time points, and is the driving parameter value, is the functional form of the state equation. Through this process, a discrete state equation group is obtained. Substitute the discrete state equation group, the switching time objective function and the circuit power consumption objective function into the nonlinear programming model to construct a nonlinear programming equation group containing driving voltage constraints, driving current constraints and waveform quality constraints. The objective function of nonlinear programming is expressed as:

[0149] ;

[0150] in, is the switching time objective function, is the circuit power consumption objective function, and is the weight coefficient. The constraints include:

[0151] ;

[0152] And limitations on waveform quality, such as:

[0153] ;

[0154] Based on the nonlinear programming equations, the ε-constraint solution pre-processing is performed by converting the circuit power consumption objective function into parameterized constraints, such as:

[0155] ;

[0156] The switching time objective function is used as the main optimization objective to generate different constraint parameter sequences. The constraint parameter sequence is input into the interior point method optimization solver. The interior point method uses the barrier function form to transform the inequality constraint into a part of the objective function. The form of the optimization problem is:

[0157] ;

[0158] in, is the obstacle parameter, is the constraint function, by gradually reducing The optimizer can approach the boundary of the feasible region by calculating the relative gain coefficient of switching time and circuit power consumption, for example:

[0159] ;

[0160] Evaluate the performance tradeoffs of different solutions. Use the non-dominated sorting method to screen the solution set, exclude those points that are dominated by other solutions on all objectives, and obtain the optimal trade-off solution between circuit power consumption and switching time, forming a Pareto optimal solution.

[0161] In a specific embodiment, the execution step performs ε-constraint solving pre-processing based on the nonlinear programming equation group, converts the circuit power consumption objective function into a parameterized constraint condition, and takes the switching time objective function as the main optimization objective, and the process of obtaining the constraint parameter sequence can specifically include the following steps:

[0162] The circuit power consumption objective function F(x) in the nonlinear programming equation group is benchmarked, where x is a parameter vector including the driving voltage value and the driving current value. By calculating the circuit power consumption and the minimum driving power consumption P min The ratio F(x) / P min , and obtain the benchmark power consumption objective function Fn (x);

[0163] Based on the benchmark power consumption objective function F n (x) Construct inequality constraints by setting the upper limit of circuit power consumption F n (x) ≤ ε and the circuit power consumption lower limit constraint F n (x) ≥ 0.8ε, where ε is the normalized power constraint parameter, and the power constraint boundary equation group G(x,ε) is obtained;

[0164] The switching time objective function T(x) is decomposed into the turn-on time component T on (x) and the off time component T off (x), and introduce weight coefficients w1=0.6 and w2=0.4 to construct the weighted sum T'(x)=w1T on (x)+w2T off (x), and obtain the main optimization objective function T'(x);

[0165] The power consumption constraint boundary equation group G(x,ε) and the main optimization objective function T'(x) are combined to obtain a parameterized planning model by constructing a solution vector [x,λ], where λ is a Lagrange multiplier vector, and a complete KKT condition including an equality constraint h(x)=0 and an inequality constraint G(x,ε)≤0;

[0166] Performing discrete sampling of ε parameters on the parameterized planning model to obtain an ε parameter sampling sequence;

[0167] Constructing a set of constrained optimization sub-problems based on the ε parameter sampling sequence, and obtaining an optimization problem group by substituting each sampling value in the ε parameter sampling sequence into the parameterized planning model;

[0168] Performing constraint gradient calculation on the optimization problem group, and obtaining a set of constraint sensitivity matrices by solving the constraint matrix and the objective function gradient;

[0169] Based on the constraint condition sensitivity matrix set, constraint feasibility judgment is performed, and the constraint parameter sequence that meets the constraint specification is screened out by calculating the rank of each Jacobian matrix and checking whether the LICQ condition is satisfied.

[0170] Specifically, for the circuit power consumption objective function in the nonlinear programming equations Benchmarking is performed, where the parameter vector Contains driving voltage value and drive current value By calculating the circuit power consumption With minimum drive power consumption The ratio of , obtains the benchmark power consumption objective function:

[0171] ;

[0172] in, Represents the total power consumption, including static power consumption and dynamic power consumption, and is defined as follows:

[0173] ;

[0174] in, is the leakage current, is the switching frequency, is the load capacitance, is the driving voltage. After benchmarking, Normalized to a dimensionless form relative to the minimum power consumption, providing a consistent measure for the optimization process. , construct inequality constraints. By setting the upper limit constraint and lower bound constraints , define the power consumption constraint boundary equations:

[0175] ;

[0176] in, To standardize the power consumption constraint parameters, reflect the target range of system power consumption. , flexibly control the tightness of power consumption constraints. For the switching time objective function Decouple the components and decompose them into the opening time components and the off-time component The turn-on and turn-off times are defined as:

[0177] ;

[0178] in, is the threshold voltage in the turn-off phase. By introducing the weight coefficient and , construct the weighted and main optimization objective function:

[0179] ;

[0180] The objective function comprehensively considers the importance ratio of the turn-on and turn-off phases, providing a unified indicator for optimization. And the main optimization objective function Combine and construct a parametric planning model. By defining the solution vector ,in is the Lagrange multiplier vector, and combined with the equality constraints and inequality constraints , forming the complete Karush–Kuhn–Tucker (KKT) condition:

[0181] ;

[0182] The Lagrangian function is defined as:

[0183] ;

[0184] The parameterized planning model Parameter discrete sampling, The value range of is discrete For example, take 10 equally spaced points:

[0185] ;

[0186] Each The value corresponds to a sub-optimization problem, forming a set of constrained optimization sub-problems. For each sub-optimization problem, the constraint sensitivity matrix is ​​solved by constraint gradient calculation. The constraint matrix is ​​defined as:

[0187] ;

[0188] The objective function gradient is:

[0189] ;

[0190] Combine the constraint matrix and the objective function gradient to calculate the sensitivity of the constraint conditions and form a set of sensitivity matrices. Based on the set of sensitivity matrices, the feasibility of the constraints is judged. By calculating the rank of each Jacobian matrix:

[0191] ;

[0192] Determine whether the linear independence constraint qualification (LICQ) condition is met. The parameter sequence that meets the LICQ condition is a feasible solution, and the parameter sequence that finally meets the constraint specification is obtained through screening.

[0193] The above describes the design method of the GaN gate driver in the embodiment of the present invention. The following describes the design system of the GaN gate driver in the embodiment of the present invention. Figure 2 , an embodiment of the design system of the gallium nitride gate driver in the embodiment of the present invention includes:

[0194] The acquisition module 201 is used to acquire the working signal of the gallium nitride gate driver, obtain the original waveform sampling data, and establish a signal integrity model under the action of electromagnetic interference;

[0195] The extraction module 202 is used to separate the original waveform sampling data into effective driving signals, low-frequency interference signals and high-frequency noise signals based on the signal integrity model, and extract characteristic parameters to obtain a signal characteristic parameter set;

[0196] A setting module 203 is used to set the switching time and circuit power consumption as optimization variables based on the signal characteristic parameter set, and set constraint conditions according to the signal integrity model to obtain an optimization objective function;

[0197] The calculation module 204 is used to construct an adaptive controller according to the optimization objective function, and perform online identification and optimization calculation on the drive parameters to obtain a drive control parameter sequence;

[0198] A solution module 205 is used to perform trapezoidal point distribution discretization processing and optimization solution on the drive control parameter sequence to obtain a Pareto optimal solution;

[0199] The setting module 206 is used to set the target driver operating parameters based on the Pareto optimal solution, and the target driver operating parameters include a driving voltage value, a driving current value, a waveform modulation parameter and a protection threshold parameter.

[0200] Through the synergy of the above components, by establishing a signal integrity model under the action of electromagnetic interference and combining the wavelet transform multi-scale decomposition technology, the precise separation of the driving signal, interference signal and noise signal is achieved, providing accurate characteristic information for the optimization of driving parameters. The dual-objective optimization model is adopted to take the switching time and circuit power consumption as optimization variables. By setting reasonable constraints, the switching loss is minimized while ensuring the driving performance. An adaptive controller is designed. Through the synergy of the parameter identification layer, the control calculation layer and the waveform generation layer, the online optimization and dynamic adjustment of the driving parameters are realized. A multi-objective optimization solution strategy based on the trapezoidal collocation method and the ε-constraint method is proposed. By converting the continuous optimization problem into a discrete nonlinear programming problem, the optimal trade-off solution of the switching time and power consumption is obtained. The hierarchical control structure and adaptive optimization algorithm are adopted to improve the adaptability of the driver to changes in working conditions, while ensuring the stability and reliability of the driving waveform.

[0201] above Figure 2 The design system of the gallium nitride gate driver in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The design device of the gallium nitride gate driver in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0202] Figure 33 is a schematic diagram of the structure of a gallium nitride gate driver design device provided by an embodiment of the present invention. The gallium nitride gate driver design device 300 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the gallium nitride gate driver design device 300. Furthermore, the processor 310 may be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the gallium nitride gate driver design device 300 to implement the steps of the gallium nitride gate driver design method described above.

[0203] The design device 300 of the gallium nitride gate driver may further include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 3 The illustrated design device structure of the gallium nitride gate driver does not constitute a limitation on the design device of the gallium nitride gate driver provided by the present invention, and may include more or fewer components than illustrated, or a combination of certain components, or a different arrangement of components.

[0204] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions, and when the instructions are executed on a computer, the computer executes the steps of the gallium nitride gate driver design method.

[0205] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0206] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0207] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A design method for a gallium nitride gate driver, characterized in that: The method comprises: Collect working signals of GaN gate drivers, obtain original waveform sampling data, and establish a signal integrity model under electromagnetic interference; Based on the signal integrity model, the original waveform sampling data is separated into an effective driving signal, a low-frequency interference signal and a high-frequency noise signal, and characteristic parameters are extracted to obtain a signal characteristic parameter set; Based on the signal characteristic parameter set, the switching time and the circuit power consumption are set as optimization variables, and the constraint conditions are set according to the signal integrity model to obtain the optimization objective function; An adaptive controller is constructed according to the optimization objective function, and the drive parameters are identified and optimized online to obtain a drive control parameter sequence; The driving control parameter sequence is subjected to trapezoidal point discretization processing and optimization solution to obtain a Pareto optimal solution; The target driver operating parameters are set based on the Pareto optimal solution, and the target driver operating parameters include a driving voltage value, a driving current value, a waveform modulation parameter, and a protection threshold parameter.

2. The gallium nitride gate driver design method according to claim 1, characterized in that: The method collects working signals of the gallium nitride gate driver to obtain original waveform sampling data and establishes a signal integrity model under electromagnetic interference, including: Sampling waveform data of a target node of a gallium nitride gate driver to obtain original waveform sampling data, wherein the target node includes a gate drive signal input terminal, a level conversion output terminal, and a gate output terminal; Performing time domain feature extraction on the original waveform sampling data to obtain a time domain feature parameter set, and performing fast Fourier transform on the original waveform sampling data to obtain a frequency domain feature parameter set; Identify interference sources according to the frequency domain characteristic parameter set, and classify interference sources into common mode interference sources and differential mode interference sources by analyzing spectrum energy distribution to obtain interference source classification results; Based on the interference source classification result, a transmission channel equivalent model is established, and distribution parameters and coupling coefficients are calculated to obtain transmission channel characteristic parameters; Inputting the transmission channel characteristic parameters and the time domain characteristic parameter set into a signal propagation analysis model to solve electromagnetic field equations and obtain signal propagation characteristic data; The driver load port characteristics are calculated according to the signal propagation characteristic data to obtain a load port model, and the load port model is jointly calculated with the signal propagation characteristic data to obtain a signal integrity model.

3. The gallium nitride gate driver design method according to claim 2, characterized in that: Based on the signal integrity model, the original waveform sampling data is separated into an effective driving signal, a low-frequency interference signal and a high-frequency noise signal, and characteristic parameters are extracted to obtain a signal characteristic parameter set, including: Based on the signal integrity model, the original waveform sampling data is subjected to wavelet transform multi-scale decomposition to obtain wavelet decomposition coefficients at each scale; The original waveform sampling data is divided into frequency bands according to the wavelet decomposition coefficients, wherein the DC to 100kHz frequency band corresponds to the effective driving signal, the 100kHz to 1MHz frequency band corresponds to the low-frequency interference signal, and the frequency band above 1MHz corresponds to the high-frequency noise signal; Performing gate drive characteristic analysis on the effective drive signal to obtain drive signal characteristic data, wherein the drive signal characteristic data includes gate voltage rise time, fall time, switch overshoot value and steady-state on-voltage; Performing envelope extraction on the low-frequency interference signal to obtain interference signal envelope characteristics, and performing energy distribution calculation on the high-frequency noise signal to obtain noise characteristic parameters; The driving signal characteristic data, the interference signal envelope characteristics and the noise characteristic parameters are subjected to data fusion and normalization processing to obtain a signal characteristic parameter set.

4. The gallium nitride gate driver design method according to claim 3, characterized in that: The method of setting the switching time and the circuit power consumption as optimization variables based on the signal characteristic parameter set and setting the constraint conditions according to the signal integrity model to obtain the optimization objective function includes: Extracting switching time features from the signal characteristic parameter set, and obtaining initial gate switching time data by calculating the time required for the gate voltage to rise from 10% to 90% and the time required for the gate voltage to drop from 90% to 10%; A mapping model between switching time and driving parameters is established according to the initial gate switching time data to obtain a switching time target function, and a power consumption analysis is performed on the driving waveform based on the signal integrity model to obtain static power consumption data of the driving circuit; Calculating the dynamic loss according to the static power consumption data of the driving circuit to obtain the circuit power consumption target function, and performing weight allocation on the switching time target function and the circuit power consumption target function to obtain a dual-objective optimization model; Setting a driving voltage limit range and a driving current limit range according to the signal integrity model, and performing a waveform quality constraint analysis on the signal characteristic parameter set to obtain a signal quality constraint condition; The driving voltage limit range, the driving current limit range and the signal quality constraint condition are substituted into the dual-objective optimization model to obtain an optimization objective function.

5. The gallium nitride gate driver design method according to claim 4, characterized in that: The adaptive controller is constructed according to the optimization objective function, and the drive parameters are identified and optimized online to obtain a drive control parameter sequence, including: The optimization objective function is input into the parameter identification layer of the adaptive controller, the parameter identification layer uses a recursive least squares algorithm to perform real-time identification of the driving voltage value and the driving current value, and performs convergence judgment by calculating the root mean square value of the parameter estimation error to obtain the initial estimated value of the driving parameter; Adaptively weighting the initial estimated values ​​of the driving parameters, dynamically tracking the driving parameters by constructing an exponentially weighted recursive least squares algorithm with a forgetting factor of 0.95, and obtaining a driving parameter identification result; The driving parameter identification result is input into the control calculation layer of the adaptive controller, and the control calculation layer adopts a nonlinear model predictive control algorithm, and performs rolling optimization on the driving parameters by setting the prediction time domain to 10 sampling periods and the control time domain to 5 sampling periods to obtain an optimized control sequence; The optimized control sequence is subjected to constraint processing, and a control quantity satisfying the constraints is obtained by setting hard constraint conditions that the driving voltage change rate does not exceed 1V / μs and the driving current change rate does not exceed 0.5A / μs; The control quantity satisfying the constraint is input into the waveform generation layer of the adaptive controller, and the waveform generation layer uses a piecewise linear interpolation algorithm to reconstruct the driving waveform by calculating the slope coefficient between adjacent sampling points to obtain a target driving waveform; The target driving waveform is input into a closed-loop controller, and the control parameters are corrected online by comparing the error between the actual driving waveform and the target waveform to obtain a corrected control quantity, and the corrected control quantity is parameter reconstructed to obtain a driving control parameter sequence.

6. The gallium nitride gate driver design method according to claim 5, characterized in that: The step of performing trapezoidal point distribution discretization processing and optimization on the driving control parameter sequence to obtain a Pareto optimal solution includes: Performing time interval point distribution processing on the drive control parameter sequence, by dividing the switching cycle into 50 equally spaced sub-intervals in a high-frequency switching scenario and dividing the switching cycle into 20 equally spaced sub-intervals in a low-frequency switching scenario, to obtain a time discrete point distribution; Based on the time discrete matching points, Lagrangian interpolation calculation is performed on the driving control parameter sequence, and a third-order Lagrangian basis function is constructed to perform piecewise interpolation fitting on the driving voltage value and the driving current value to obtain a parameter discrete mapping function; Performing trapezoidal numerical integration on the parameter discrete mapping function, discretizing the state equation by calculating the product of the mean value of the driving parameters of adjacent points and the time step, and obtaining a discrete state equation group; Substituting the discrete state equation group, the switching time objective function and the circuit power consumption objective function into a nonlinear programming model, and constructing a nonlinear programming equation group including a driving voltage constraint, a driving current constraint and a waveform quality constraint; Based on the nonlinear programming equation group, ε-constraint solving pre-processing is performed to obtain a constraint parameter sequence by converting the circuit power consumption objective function into a parameterized constraint condition and taking the switching time objective function as the main optimization objective; Inputting the constraint parameter sequence into an interior point optimization solver, solving the single-objective optimization problem under each constraint parameter by using a barrier function method, and obtaining an optimal solution sequence; The optimal solution sequence is subjected to Pareto optimality determination, and the relative gain coefficient of the switching time and the circuit power consumption is calculated, and the optimal trade-off solution between the circuit power consumption and the switching time is screened out using a non-dominated sorting method to obtain a Pareto optimal solution.

7. The gallium nitride gate driver design method according to claim 6, characterized in that: The ε-constraint solving pre-processing based on the nonlinear programming equation group is performed by converting the circuit power consumption objective function into parameterized constraint conditions, and taking the switching time objective function as the main optimization objective to obtain a constraint parameter sequence, including: The circuit power consumption objective function F(x) in the nonlinear programming equation group is benchmarked, where x is a parameter vector including the driving voltage value and the driving current value. By calculating the circuit power consumption and the minimum driving power consumption P min The ratio F(x) / P min , and obtain the benchmark power consumption objective function F n (x); Based on the benchmark power consumption objective function F n (x) Construct inequality constraints by setting the upper limit of circuit power consumption F n (x) ≤ ε and the circuit power consumption lower limit constraint F n (x) ≥ 0.8ε, where ε is the normalized power constraint parameter, and the power constraint boundary equation group G(x,ε) is obtained; The switching time objective function T(x) is decomposed into the turn-on time component T on (x) and the off time component T off (x), and introduce weight coefficients w1=0.6 and w2=0.4 to construct the weighted sum T'(x)=w1T on (x)+w2T off (x), and obtain the main optimization objective function T'(x); The power consumption constraint boundary equation group G(x,ε) and the main optimization objective function T'(x) are combined to obtain a parameterized planning model by constructing a solution vector [x,λ], where λ is a Lagrange multiplier vector, and a complete KKT condition including an equality constraint h(x)=0 and an inequality constraint G(x,ε)≤0; Performing discrete sampling of ε parameters on the parameterized planning model to obtain an ε parameter sampling sequence; Constructing a set of constrained optimization sub-problems based on the ε parameter sampling sequence, and obtaining an optimization problem group by substituting each sampling value in the ε parameter sampling sequence into the parameterized planning model; Performing constraint gradient calculation on the optimization problem group, and obtaining a set of constraint sensitivity matrices by solving the constraint matrix and the objective function gradient; Based on the constraint condition sensitivity matrix set, constraint feasibility judgment is performed, and the constraint parameter sequence that meets the constraint specification is screened out by calculating the rank of each Jacobian matrix and checking whether the LICQ condition is satisfied.

8. A gallium nitride gate driver design system, characterized in that: A method for designing a gallium nitride gate driver according to any one of claims 1 to 7, the system comprising: The acquisition module is used to collect the working signal of the GaN gate driver, obtain the original waveform sampling data, and establish a signal integrity model under the action of electromagnetic interference; An extraction module, used to separate the original waveform sampling data into an effective driving signal, a low-frequency interference signal and a high-frequency noise signal based on the signal integrity model, and extract characteristic parameters to obtain a signal characteristic parameter set; A setting module, used to set the switching time and circuit power consumption as optimization variables based on the signal characteristic parameter set, and set constraint conditions according to the signal integrity model to obtain an optimization objective function; A calculation module is used to construct an adaptive controller according to the optimization objective function, and perform online identification and optimization calculation on the drive parameters to obtain a drive control parameter sequence; A solution module, used for performing trapezoidal point discretization processing and optimization solution on the driving control parameter sequence to obtain a Pareto optimal solution; A setting module is used to set target driver operating parameters based on the Pareto optimal solution, wherein the target driver operating parameters include a driving voltage value, a driving current value, a waveform modulation parameter and a protection threshold parameter.

9. A design device for a gallium nitride gate driver, characterized in that: The gallium nitride gate driver design device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instruction in the memory to enable the gallium nitride gate driver design device to perform the gallium nitride gate driver design method according to any one of claims 1 to 7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the method for designing a gallium nitride gate driver according to any one of claims 1 to 7 is implemented.

Images