Temperature compensation method and device for silicon carbide MOS (Metal Oxide Semiconductor) driving control chip

Abstract

The invention provides a temperature compensation method and device for a silicon carbide MOS drive control chip, and the method comprises the steps: obtaining a junction temperature-electrical parameter set of a silicon carbide MOS, and carrying out the thermoelectric coupling simulation modeling processing of a junction temperature-electrical parameter mapping data set, and obtaining three-dimensional temperature-efficiency characteristic data; performing real-time parameter analysis and dynamic execution on the three-dimensional temperature-efficiency characteristic data to obtain an optimized switching characteristic waveform; extracting the optimized switching characteristic waveform frequency domain characteristics, and generating a dynamic frequency modulation parameter; and performing dead time iterative optimization on the dynamic frequency modulation parameters to obtain a control parameter combination of the silicon carbide MOS.

Description

Technical Field

[0001] The present application relates to the field of electronic control technology, and in particular to a temperature compensation method and device for a silicon carbide MOS drive control chip. Background Technology

[0002] Silicon carbide (SiC) MOSFET has shown significant advantages in the fields of renewable energy power generation and electric vehicles due to its high withstand voltage, low conduction loss and high frequency characteristics. However, as the junction temperature of the device increases, its electrical characteristics show significant nonlinear changes: the on-resistance (Rds(on)) increases as a quadratic function with rising temperature, the switching loss (Esw) is positively correlated with temperature, and the temperature sensitivity of the gate charge (Qg) causes the switching speed to decrease. Existing temperature compensation methods mostly use fixed thresholds or linear compensation strategies, which are difficult to accurately match the thermoelectric coupling characteristics of SiC MOSFET, resulting in a sharp drop in efficiency (>15%) and reliability degradation under high temperature conditions. Especially in ultra-high voltage (≥1200V) applications, the static temperature compensation model of traditional driver chips cannot adapt to the thermal accumulation effect of the switching transient process, which can easily cause local thermal runaway. SUMMARY OF THE INVENTION

[0003] This application provides a temperature compensation method and device for a silicon carbide MOS drive control chip, which is used to perform dynamic temperature compensation on the silicon carbide MOS according to the thermoelectric coupling characteristics of the silicon carbide MOS.

[0004] In a first aspect, an embodiment of the present application provides a temperature compensation method for a silicon carbide MOS drive control chip, the method comprising: Obtain a junction temperature-electrical parameter set of silicon carbide MOS, perform thermoelectric coupling simulation modeling based on the junction temperature-electrical parameter mapping data set, and obtain three-dimensional temperature-efficiency characteristic data; Perform real-time parameter analysis and dynamic execution on three-dimensional temperature-efficiency characteristic data to obtain optimized switching characteristic waveform; Extract the frequency domain characteristics of the optimized switching characteristic waveform to generate dynamic frequency modulation parameters; The dynamic frequency modulation parameters are iteratively optimized for dead time to obtain the control parameter combination of the silicon carbide MOS.

[0005] In a second aspect, an embodiment of the present application provides a temperature compensation device for a silicon carbide MOS drive control chip, the device comprising: A data acquisition module is used to obtain the junction temperature-electrical parameter set of the silicon carbide MOS, and perform thermoelectric coupling simulation modeling based on the junction temperature-electrical parameter mapping data set to obtain three-dimensional temperature-efficiency characteristic data; Data optimization module, used for real-time parameter analysis and dynamic execution of three-dimensional temperature-efficiency characteristic data to obtain optimized switching characteristic waveform; A parameter generation module is used to extract the frequency domain characteristics of the optimized switching characteristic waveform and generate dynamic frequency modulation parameters; The result output module is used to perform iterative optimization of the dead time of the dynamic frequency modulation parameters to obtain the control parameter combination of the silicon carbide MOS.

[0006] In a third aspect, an embodiment of the present application provides an electronic device, wherein the electronic device includes a memory and a processor; The memory is used to store computer programs; The processor is used to execute the computer program and implement the temperature compensation method of the silicon carbide MOS drive control chip as described in any one of the embodiments of the present application when executing the computer program.

[0007] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements a temperature compensation method for a silicon carbide MOS drive control chip as described in any one of the embodiments of the present application.

[0008] The embodiment of the present application provides a temperature compensation method for a silicon carbide MOS drive control chip, the method comprising: obtaining a junction temperature-electrical parameter set of the silicon carbide MOS, performing thermoelectric coupling simulation modeling based on the junction temperature-electrical parameter mapping data set, and obtaining three-dimensional temperature-efficiency characteristic data; performing real-time parameter analysis and dynamic execution on the three-dimensional temperature-efficiency characteristic data to obtain an optimized switching characteristic waveform; extracting the frequency domain characteristics of the optimized switching characteristic waveform to generate dynamic frequency modulation parameters; performing dead time iterative optimization on the dynamic frequency modulation parameters to obtain a control parameter combination of the silicon carbide MOS. In the above method, a junction temperature-electrical parameter set is obtained and thermoelectric coupling simulation modeling is performed, and real-time parameter analysis based on the three-dimensional temperature-efficiency characteristic data optimizes the switching performance and reduces the switching loss. The dynamic frequency modulation parameters are generated by extracting the frequency domain characteristics of the optimized waveform, and the dead time iterative optimization is performed to achieve adaptive adjustment of the operating frequency of the silicon carbide MOS according to the operating environment, effectively reducing the shoot-through current and electromagnetic interference, which not only improves the overall energy conversion efficiency, but also prolongs the device life and ensures stable and reliable operation under various working conditions. Brief Description of the Figures

[0009] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for describing the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1A schematic flow chart of a temperature compensation method for a silicon carbide MOS drive control chip provided in an embodiment of the present application; Figure 2 This is a schematic block diagram of a temperature compensation device for a silicon carbide MOS drive control chip provided in an embodiment of the present application. Specific implementation method

[0011] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of them. 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.

[0012] The flowcharts shown in the attached figures are only examples and do not necessarily include all contents and operations / steps, nor do they have to be executed in the order described. For example, some operations / steps can be decomposed, combined or partially merged, so the actual execution order may change according to the actual situation.

[0013] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, the singular forms "a", "an" and "the" are intended to include plural forms unless the context clearly indicates otherwise.

[0014] It should also be further understood that the term "and / or" used in the present specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0015] Please refer to Figure 1 , Figure 1 is a schematic flow chart of a temperature compensation method for a silicon carbide MOS drive control chip provided in an embodiment of the present application. Figure 1 The temperature compensation method of the silicon carbide MOS drive control chip shown in the figure includes the following specific steps: S101-S104.

[0016] S101, obtain the junction temperature-electrical parameter set of silicon carbide MOS, perform thermoelectric coupling simulation modeling based on the junction temperature-electrical parameter mapping data set, and obtain three-dimensional temperature-efficiency characteristic data.

[0017] ​For example, the junction temperature test interval is set to [25°C, 250°C], covering the wide temperature operating range of silicon carbide devices, and precise control of temperature points is achieved through precision temperature control testing. At each temperature point, a set of key electrical parameters are measured, including gate charge Qg, on-resistance Rds (on), switching time tr / tf, and switching energy loss E on / E off , forming a complete temperature-electrical parameter mapping data set. To ensure data reliability, the four-terminal Kelvin measurement method is used to eliminate the influence of parasitic resistance, the double pulse test technology captures transient switching characteristics, and the IGBT power module provides stable test conditions as an inductive load. The acquired raw data matrix is ​​digitally filtered to remove noise interference and improve data quality. Based on these temperature-electrical parameter mapping data, the finite element analysis method is used to establish a chip-level thermoelectric coupling simulation model. The model uses detailed SiC device structural parameters, including the thermal conductivity characteristics of the semiconductor layer, metal layer and packaging material. Furthermore, the on-resistance parameter is fitted with a quadratic polynomial, and the least squares method is used to generate Rds(on)(T)=R 0 [1+α(TT 0 )+β(TT 0 ) 2 ] mathematical model, where R 0 indicates reference temperature T 0 , α and β are temperature coefficients. Perform linear regression analysis on the switching loss parameters to establish E sw (T)=E 0 [1+γ(TT 0 )] loss model, where E sw (T) is the switching loss parameter, E 0 is the switching loss at the reference temperature, and γ is the loss temperature coefficient. Through the above processing, a three-dimensional mapping matrix containing temperature coordinates, electrical parameter coordinates, and efficiency coordinates is constructed to form a complete three-dimensional temperature-efficiency characteristic model, laying the foundation for subsequent analysis and parameter optimization.

[0018] S102, perform real-time parameter analysis and dynamic execution on the three-dimensional temperature-efficiency characteristic data to obtain an optimized switching characteristic waveform.

[0019] For example, FPGA is used as the real-time analysis and processing core to implement the temperature data stream analysis algorithm and establish an accurate mapping relationship table between temperature values ​​and key gate drive parameters (Vgs, Rg). Temperature acquisition monitors the junction temperature changes of SiC devices in real time through thermocouples or infrared thermal imaging devices, and the data is sent to the FPGA processing unit after AD conversion and digital filtering. Gate voltage compensation is performed according to the formula, specifically: ΔVgs=K v ·(TT 0 )2 +ε v , calculate the gate voltage adjustment ΔVgs, where K v is the coefficient of the quadratic term, ε v is the correction value to achieve accurate compensation for threshold voltage drift at high temperature. At the same time, a gate resistance compensation algorithm is constructed according to the formula: ΔRg=K r ·(E sw -E 0 ) / E 0 +ε r , calculate the gate resistance adjustment ΔRg, K r is the proportional coefficient to achieve active control of switching loss. Through these two compensation algorithms, a discrete parameter combination solution set including Vgs∈[12V, 20V] and Rg∈[1Ω, 20Ω] is generated to cover the optimal drive parameter configuration under different temperature conditions. After priority sorting and conflict elimination, the parameter solution set is converted into a gate drive parameter adjustment instruction set. A digital potentiometer array is used to achieve 0.5Ω step fine adjustment of the Rg value, and a high-precision DAC module is used to generate a 0.1V resolution output of the Vgs value to ensure the accuracy of parameter adjustment. The drive circuit is designed with a switching cycle synchronization mechanism to ensure that at t d Complete parameter updates within the (deadtime) period to avoid transient anomalies caused by parameter changes. Through the above mechanism, an optimized switching characteristic waveform with a variable rising edge slope tr(Vgs, Rg) and falling edge slope tf(Vgs, Rg) is generated, achieving precise control of the switching characteristics of silicon carbide MOS, reducing switching losses and suppressing electromagnetic interference.

[0020] S103, extract the frequency domain characteristics of the optimized switching characteristic waveform and generate dynamic frequency modulation parameters.

[0021] For example, a high-speed data acquisition module is used to capture the optimized switching transient waveform, with a sampling rate of 1GS / s to ensure the capture of high-frequency harmonic components. The captured time domain waveform data is fast Fourier transformed through the FFT analysis module to extract the fundamental wave and high-order harmonic components H of the switching waveform n (n=1, 2, 3...), the analysis spectrum range covers 10kHz to 100MHz. The extracted harmonic components are normalized, the amplitude ratio of each harmonic to the fundamental wave is calculated, and the degree of waveform distortion is evaluated. Combined with temperature data, the temperature-related frequency attenuation coefficient η(f)=Σ(H n (f)·T n ) / H 1(f), this coefficient reflects the changing law of harmonic attenuation characteristics at different temperatures. Based on the frequency domain feature analysis, a temperature-adaptive segmented frequency control function is constructed to realize dynamic modulation of the switching frequency. The control function adopts a segmented definition method: when the temperature T∈[25℃, 120℃), f(T)=120kHz-0.4kHz / ℃×(T-25℃); when the temperature T∈[120℃, 150℃), f(T)=80kHz-0.2kHz / ℃×(T-120℃); when the temperature T∈[150℃, 250℃], f(T)=60kHz-0.1kHz / ℃×(T-150℃). This segmented control strategy uses a higher switching frequency in the low temperature range to increase the power density, and gradually reduces the switching frequency as the temperature rises to reduce the loss, reflecting the nonlinear relationship between temperature and switching frequency.

[0022] S104, perform iterative optimization of the dead time of the dynamic frequency modulation parameters to obtain a control parameter combination of the silicon carbide MOS.

[0023] For example, the switch delay time differential equation dtd / dT=α·ln(Rg / Rg0)+β·(Vgs-Vgs0) is established to describe the relationship between the gate resistance Rg, the gate-source voltage Vgs and the temperature T, where α and β are experimental calibration coefficients, and Rg0 and Vgs0 are reference operating point parameters. This equation reflects the mechanism of the influence of temperature change on the switch delay time and provides a theoretical basis for the optimization of the dead time. The Newton iteration method is used to solve the optimal dead time, and the iteration formula is T dead (k+1) =T dead k -[f(T dead k )-T dead_min ] / f'(T dead k ), where f(T dead ) is the dead time objective function, T dead_min To ensure the minimum dead time for safety, f'(T dead ) is the derivative of the objective function. The iterative process sets the convergence condition |T dead (k+1) -T dead k |<1ns, to ensure the calculation accuracy. In order to prevent the theoretical calculation results from exceeding the practical feasible range, an anti-saturation control algorithm is designed with the constraint condition of 100ns≤T dead ≤t d (off)-t d ​​(on) + 50ns, ensuring that the dead time is neither too short to cause a shoot-through risk nor too long to cause additional loss. After the iterative optimization is completed, the frequency parameter f, gate voltage parameter Vgs, gate resistance parameter Rg and dead time parameter T dead Integrate into a four-dimensional control vector (f, Vgs, Rg, T dead ) and output it to the power drive unit. This control vector realizes full parameter optimization control of SiC MOS gate drive to adapt to different temperature conditions. In addition, a parameter smooth transition mechanism is used to prevent instability caused by sudden changes in control parameters. Through this complete parameter optimization process, efficient and stable operation of SiC MOS devices in a wide temperature range is achieved, significantly improving the performance and reliability of power conversion.

[0024] The embodiment of the present application provides a temperature compensation method for a silicon carbide MOS drive control chip, the method comprising: obtaining a junction temperature-electrical parameter set of the silicon carbide MOS, performing thermoelectric coupling simulation modeling based on the junction temperature-electrical parameter mapping data set, and obtaining three-dimensional temperature-efficiency characteristic data; performing real-time parameter analysis and dynamic execution on the three-dimensional temperature-efficiency characteristic data to obtain an optimized switching characteristic waveform; extracting the frequency domain characteristics of the optimized switching characteristic waveform to generate dynamic frequency modulation parameters; performing dead time iterative optimization on the dynamic frequency modulation parameters to obtain a control parameter combination of the silicon carbide MOS. In the above method, a junction temperature-electrical parameter set is obtained and thermoelectric coupling simulation modeling is performed, and real-time parameter analysis based on the three-dimensional temperature-efficiency characteristic data optimizes the switching performance and reduces the switching loss. The dynamic frequency modulation parameters are generated by extracting the frequency domain characteristics of the optimized waveform, and the dead time iterative optimization is performed to achieve adaptive adjustment of the operating frequency of the silicon carbide MOS according to the operating environment, effectively reducing the shoot-through current and electromagnetic interference, which not only improves the overall energy conversion efficiency, but also prolongs the device life and ensures stable and reliable operation under various working conditions.

[0025] In order to more clearly introduce the technical solution of the present application, the technical solution of the present application will be introduced through specific embodiments below. It should be noted that the specific embodiments are used to expand the technical solution of the present application, but are not intended to limit the present application.

[0026] In some embodiments, obtaining a junction temperature-electrical parameter set of a silicon carbide MOS includes: S1011-S1015.

[0027] S1011. Perform nonlinear division on the junction temperature test interval to be tested and determine the junction temperature sampling point.

[0028] Exemplarily, the junction temperature test interval [25°C, 250°C] is processed by nonlinear temperature sampling point division to generate a distribution set of temperature sampling points {T i}, where the sampling density is determined by a piecewise exponential function, the sampling interval in the high temperature zone (T≥150℃) is ΔT=5℃, and the sampling interval in the low temperature zone (T<150℃) is ΔT=10℃.

[0029] S1012, through the preset double pulse test platform and preset test conditions, the switching transient data of each junction temperature sampling point is collected to obtain the gate-source voltage waveform data, the drain-source current waveform data and the drain-source voltage waveform data.

[0030] Exemplarily, a 1200V / 100A test condition is applied through a double pulse test platform, and a high-speed oscilloscope is used to capture the gate-source voltage Vgs(t), drain-source current I ds (t) and drain-source voltage V ds (t) waveform. Based on the waveform integration, the gate charge Qg=∫Vgs(t)·C iss (t)dt, where C iss (t) is obtained by real-time gate capacitance inversion algorithm.

[0031] S1013, determining a junction temperature-electrical parameter matrix according to the gate-source voltage waveform data, the drain-source current waveform data and the drain-source voltage waveform data, the junction temperature-electrical parameter matrix including: junction temperature sequence, on-resistance, switching time and switching energy loss.

[0032] Example, calculate the on-resistance Rds(on): Rds(on)=V ds_sat / I ds_avg , where V ds_sat By V ds (t) is obtained by mean filtering during the conduction steady-state phase.

[0033] Extract switching time tr: tr=argmax(dV ds / dt) rising , tf=argmax(dV ds / dt) falling .

[0034] Building the switching energy loss E sw :E sw =∫(V ds (t)·I ds (t))dt in [t 0 ,t 0 +tr] and [t 1 -tf,t 1 ]Numerical integration of time windows.

[0035] S1014, perform thermoelectric coupling compensation processing on the junction temperature-electrical parameter matrix to obtain a corrected parameter mapping relationship.

[0036] ​​​​​​Exemplary, the algorithm for calculating parasitic inductance compensation is: Rds(on) corr =Rds(on)+L par (dI ds / dt) avg , where Rds(on) corr is the on-resistance after compensation; Lpar is the equivalent parasitic inductance, which is equal to the sum of the parasitic inductances in the circuit. A sliding average filter is used to eliminate the temperature hysteresis effect of the gate charge Qg. The window width is related to the thermal time constant τ th =RC th Matching, where R is thermal resistance, C th is the heat capacity. The missing data points caused by measurement noise are filled by polynomial interpolation method to obtain the corrected parameter mapping relationship.

[0037] S1015, perform multi-physics field verification processing on the corrected parameter mapping relationship to generate a junction temperature-electrical parameter set.

[0038] For example, a finite element thermal-electrical-mechanical coupling simulation model is established, and the input parameters include chip thickness, solder layer thermal conductivity and package stress distribution. The steady-state temperature field and electric field coupling equations are solved by the Newton-Raphson iterative algorithm, and the output satisfies the verification data set of |experimental value-simulation value| / experimental value ≤ 3%.

[0039] In some embodiments, real-time parameter analysis and dynamic execution are performed on the three-dimensional temperature-efficiency characteristic data to obtain an optimized switching characteristic waveform, including: S1021-S1028.

[0040] S1021, implant the real-time junction temperature into the three-dimensional temperature-efficiency characteristic model data according to a preset interpolation method to obtain a dynamic parameter set under the current junction temperature.

[0041] Exemplarily, real-time temperature interpolation processing is performed on the three-dimensional temperature-efficiency characteristic model to obtain the current junction temperature T j The dynamic parameter set under {P(T j )}, where Rds(on)(T is calculated in the three-dimensional mapping matrix by the bicubic spline interpolation algorithm j )、Esw(T j ) and switching time parameter tr(T j ) / tf(T j ).

[0042] S1022, perform gate voltage compensation on the dynamic parameter set and generate a gate voltage adjustment instruction.

[0043] Exemplarily, establish the voltage compensation equation: ΔVgs=K v ·(Rds(on)(T​​​​​​​​j )-R 0 ) / R 0 +λ·(E sw (T j ) / E 0 ) 2 ; Among them, K v is the voltage compensation coefficient, λ is the nonlinear weight factor, the parameter value is determined by the least squares optimization algorithm, ΔVgs is the gate voltage supplement amount, and the gate voltage adjustment instruction is generated according to the gate voltage supplement amount ΔVgs.

[0044] S1023, optimize the gate resistance of the dynamic parameter set and generate a gate resistance adjustment instruction.

[0045] Exemplarily, construct the resistance optimization objective function: min{tr(T j )·tf(T j )+α·(E sw (T j ) / Vgs 2 )}, the golden section search algorithm is used to solve the gate resistance compensation value ΔRg in the interval Rg∈[1Ω, 20Ω], and the gate resistance adjustment instruction is generated according to the gate resistance compensation value ΔRg.

[0046] S1024: Perform discrete coding according to the gate voltage adjustment instruction and the gate resistance adjustment instruction to generate a first driving parameter.

[0047] Exemplarily, the gate voltage supplement amount ΔVgs is mapped to D by a 12-bit DAC resolution Vgs =INT[(ΔVgs+12) / 0.1], the gate resistance compensation ΔRg value is encoded as D through the digital potentiometer. Rg =ROUND(ΔRg / 0.5)×0.5Ω, generate the first driving parameter.

[0048] S1025, perform switching cycle synchronization processing on the driving parameters according to the preset switching cycle to obtain the second driving parameters.

[0049] For example, design hard real-time control logic based on Zynq SoC, in the dead time window [t dead_start ,t dead_end ] to complete parameter updates and eliminate asynchronous signal jitter through cross-clock domain synchronization circuits.

[0050] S1026, calculating the rising edge slope and the falling edge slope according to the second driving parameter, and generating a third driving parameter for adjusting the on and off state according to the rising edge slope and the falling edge slope.

[0051] ​​​​​​Exemplarily, based on the second driving parameter, the voltage change curves of the rising edge (tr) and the falling edge (tf) of the gate driving waveform are captured in real time by a high-speed sampling circuit. The rising edge slope k is extracted by differential operation. r and falling edge slope k f , where k r Reflects the gate capacitance charging rate during the conduction phase, k f Characterizes the gate charge Qg release speed in the turn-off phase. Based on the slope characteristics, a dynamic adjustment algorithm is constructed: when k r When it exceeds the preset threshold, it is determined that the switching speed is too fast, resulting in EMI exceeding the standard. At this time, the gate resistance Rg is increased according to the exponential decay law; when k f When it is lower than the critical value, it is identified as a risk of turn-off delay, and the gate voltage Vgs is increased through a nonlinear compensation mechanism. The temperature-slope coupling factor λ (T is introduced during the adjustment process j )=1+0.005(T j -25), so that the correction amount of Rg and Vgs changes adaptively with the junction temperature, generating the third drive parameter (Vgs', Rg'), whose value is dynamically adjusted within the range of Vgs±2V and Rg±3Ω, achieving a refined balance between switching speed and loss.

[0052] S1027, perform switching transient correction on the third driving parameter to generate a feedback correction coefficient set.

[0053] Exemplarily, after injecting the third driving parameter into the gate driving circuit, the switching transient waveforms of Vds(t) and Ids(t) are synchronously captured through nanosecond delay synchronous sampling technology. Three key characteristic quantities are extracted: Miller platform duration Δt plateau 、Voltage overshoot amplitude ΔV overshoot and current drop rate di / dt fall . Design a multi-dimensional deviation analysis model to compare the measured characteristic quantities with the ideal reference values. The weight coefficient is dynamically allocated according to the operating frequency (focusing on di / dt at high frequencies and ΔV at low frequencies). overshoot ). The wavelet packet decomposition algorithm is used to separate the original correction factor into frequency bands, filter out high-frequency measurement noise, and retain low-frequency components that are strongly related to the switching behavior. The mean μ and standard deviation σ of the correction factor are calculated by the sliding window statistics method (window width = 5 switching cycles). When the correction factor deviates from the range of μ±3σ, the abnormal value replacement mechanism is activated to replace the current value with the historical mean to generate a feedback correction coefficient set {C fb}, including voltage compensation coefficient Kv, resistance correction coefficient Kr and timing adjustment factor Δt.

[0054] ​S1028. Construct a spatial equation based on the feedback correction coefficient set, and analyze it using a Kalman filter to obtain an optimized switching characteristic waveform. The optimized switching characteristic waveform includes: a target gate voltage and a target gate resistance.

[0055] Exemplarily, a five-dimensional state space equation is constructed, and the state variables include Vgs, Rg, junction temperature T j 、Switching frequency f sw and dead time T dead . Set the feedback correction coefficient set {C fb} is used as the observation vector input to the Kalman filter, and the weights of the process noise Q and the observation noise R are dynamically adjusted through the covariance matrix update algorithm. In the prediction stage, the prior estimate of the state variable is inferred based on the three-dimensional temperature-efficiency model; in the correction stage, the prior value is Bayesian optimized by integrating the real-time observation data, focusing on solving the coupling interference problem between Vgs and Rg. In view of the nonlinear characteristics of SiC MOSFET, the extended Kalman filter (EKF) is introduced to linearize the Jacobian matrix of the state equation. After 3 to 5 iterations, the filter outputs the posterior estimate Vgs opt and Rg opt , with a resolution of 0.1V and 0.5Ω respectively, the generated optimized switching characteristic waveform has a variable slope tr / tf and a dynamic dead zone window, and the nanosecond refresh of the drive parameters is achieved through FPGA hardware logic to ensure that the closed-loop adjustment of the parameters is completed within each switching cycle.

[0056] In some embodiments, the third driving parameter is subjected to switching transient correction to generate a feedback correction coefficient set, including: S271-S274.

[0057] S271: Perform high-speed sampling and differential processing on the driving waveform corresponding to the third driving parameter to extract a set of switching transient feature quantities.

[0058] Exemplarily, the driving waveform corresponding to the third driving parameter is sampled and differentiated at high speed, and a 16-bit high-speed ADC is used to synchronously capture the Vds(t) and Ids(t) waveform signals at a sampling rate of 2GS / s. The sampling time window covers the first 10% of the on-phase and the last 15% of the off-phase of the switching cycle. The Vds(t) waveform is processed in real time by a digital differentiator, and the voltage change rate dv / dt=ΔVds / Δt is calculated. The sliding window width is set to 5ns, the step length is 1ns, and the dv / dt of the on-phase is extracted max and dv / dt during the turn-off phase min is used as the key characteristic quantity. For the Ids(t) waveform, after applying the third-order Savitzky-Golay filter to smooth the noise, the current change rate di / dt=ΔIds / Δt is calculated. rise_start , t​​​rise_end ]Extract di / dt in time window avg Mean. At the same time, the Miller platform area is identified, and the platform starting point t is determined by the second-order derivative extreme value detection method plat_start and the end point t plat_end , calculate the platform duration Δt plat =t plat_end - t plat_start . Change dv / dt max 、dv / dt min 、di / dt avg and Δt pla t forms a four-dimensional switch transient feature set, which is stored as a time series matrix F sw ∈R (N×4) , where N is the number of switching cycles collected consecutively.

[0059] S272, perform dynamic deviation analysis on the set of switch transient characteristic quantities to generate an original correction factor matrix.

[0060] Exemplarily, for the switch transient feature quantity set F sw Perform dynamic deviation analysis and establish characteristic reference vector F ref =[dv / dt max_ref , dv / dt min_ref , di / dt avg_ref , Δt plat_ref ], the reference value comes from the corresponding junction temperature T in the three-dimensional temperature-efficiency characteristic model j Theoretical calculation value. Construct dynamic deviation vector ΔF=F sw -F ref , apply nonlinear transformation to each characteristic component: for dv / dt max and di / dt avg Use the hyperbolic tangent function tanh (0.5ΔF) to compress the large deviation amplitude; for dv / dt min and Δt plat Use the sign function sgn(ΔF) and the absolute value square root operation sqrt(|ΔF|) to enhance the sensitivity to small deviations. Design the weight matrix W=diag([0.4, 0.3, 0.2, 0.1]) and calculate the original correction factor C by matrix multiplication raw =W·ΔF transformed , where ΔF transformed is the deviation vector after nonlinear transformation. Set the saturation limit condition: when C raw_i >1, take 1, C raw_i <-1 is taken as -1, generating the standardized original correction factor matrix C raw ∈[-1, 1] (N×4) , whose row vectors correspond to the four-way correction parameters of each switching cycle.​​​​​​

[0061] S273, modify the original correction factor matrix through wavelet packet denoising to obtain the filtered correction factor.

[0062] The original correction factor matrix C is denoised by wavelet packet denoising raw To make corrections, select the db6 wavelet basis function to decompose each correction factor sequence into 4 layers to generate low-frequency approximate coefficients and high-frequency detail coefficients. Design an adaptive threshold , where σ is the median absolute deviation of the detail coefficients of each layer, M is the coefficient length, and soft threshold processing is performed on the high-frequency coefficients: d new =sgn(d)(|d|-T), when |d|>T otherwise 0, where d is the high frequency coefficient before update, d new is the updated high frequency coefficient. For the Miller platform time correction factor (column 4), an additional frequency band energy constraint is added: the 1-25MHz frequency band coefficient is retained to match the switching noise spectrum characteristics. After reconstructing the signal, the four columns of correction factors are phase aligned respectively, and the cross-correlation algorithm is used to calculate the delay of each channel, using dv / dt max The correction factor is used as the benchmark for time offset compensation, and the correction factor matrix C after filtering is obtained filt ∈R (N×4) , its signal-to-noise ratio is increased to more than 18dB of the original data, and the standard deviation of time domain fluctuation is reduced to 32% before filtering.

[0063] S274, according to the preset correction factor validity criterion, the confidence verification of the filtered correction factor is performed to generate a feedback correction coefficient set.

[0064] Exemplarily, according to the preset correction factor validity criterion, the filtered correction factor matrix C is calculated filt Statistical characteristics of : The sliding mean μ of each column correction factor i (window width = 20 switching cycles) and standard deviation σi. Establishing dynamic validity threshold μ i ±3σ i , when the correction factor exceeds the threshold range for three consecutive cycles, it is determined to be an abnormal data segment. Two-level processing is implemented for abnormal data: if the deviation direction is consistent with the temperature change trend (T j C filt Forward deviation), the exponential weighted moving average method is used for smoothing, and the attenuation factor α=0.7; if the deviation direction deviates from the temperature trend, it is replaced by the mean of the corresponding phase of the previous 5 cycles. The distribution characteristics of the correction factor are verified by the chi-square test, and the normal distribution assumption is accepted when the significance level p<0.01, and the feedback correction coefficient set C is generated fb ={K v ,K r , K​​​​t ,K d}, where K v is the voltage compensation gain coefficient, K r is the resistance correction proportional factor, K t is the time offset compensation, K d is the dead zone fine-tuning parameter, and the accuracy of each coefficient reaches 0.01, 0.1Ω, 2ns, and 5ns respectively.

[0065] In some embodiments, extracting the frequency domain characteristics of the optimized switching characteristic waveform and generating dynamic frequency modulation parameters include: S1031-S1035.

[0066] S1031, segment the optimized switching characteristic waveform to obtain switching cycle timing data.

[0067] Exemplarily, the optimized switching characteristic waveform is segmented to obtain the switching cycle timing data. Exemplarily, an adaptive threshold detection algorithm is used to accurately segment the switching waveform, and the threshold is 10% of the waveform peak as the segmentation point. An 8-bit high-speed analog-to-digital converter is used to capture the complete switching process with a sampling interval of 50ns. Each switching cycle is divided into four key stages: the turn-on rising stage, the turn-on stable stage, the turn-off falling stage, and the cut-off stage. The corresponding time windows are set to 100ns, (ton-100)ns, 100ns, and (toff-100)ns, respectively. For each time window, the instantaneous values ​​of the voltage v(t) and the current i(t) are recorded to form 4×N-dimensional switching cycle timing data, where N is the number of sampling points in each stage.

[0068] S1032, perform short-time Fourier transform on the switching cycle timing data to generate a time-frequency distribution matrix.

[0069] Exemplarily, a short-time Fourier transform is performed on the switching cycle timing data to generate a time-frequency distribution matrix. Exemplarily, a Hamming window function is applied to the timing feature matrix for windowing, and the window length is set to 128 points with an overlap rate of 50% to ensure that spectrum leakage is minimized. The frequency resolution of the short-time Fourier transform is set to 200kHz, the time resolution is 25ns, and the calculation process uses the radix 2-FFT algorithm to reduce the computational complexity. The transformation result forms a complex matrix F(t, ω), where t represents the time axis, ω represents the frequency axis, and the matrix elements represent the complex amplitude of the frequency ω at time t. The power spectrum density matrix P(t, ω) is obtained by calculating |F(t, ω)|², which reflects the frequency energy distribution of the switching characteristic waveform at different time points. The dimension of the power spectrum density matrix is ​​T×Ω, where T is the number of time sampling points and Ω is the number of frequency sampling points, and the typical value is 256×128.

[0070] S1033. Perform harmonic energy spectrum analysis on the time-frequency distribution matrix to extract high-order harmonic feature vectors.

[0071] Exemplarily, the harmonic energy spectrum analysis is performed on the time-frequency distribution matrix to extract the high-order harmonic feature vector. Exemplarily, based on the power spectrum density matrix P(t, ω), the energy distribution of the first 15 harmonic components is calculated with the switching frequency fs as the base frequency. The energy of each harmonic Ei is obtained by the integral formula E i =∫P(t, i·fs)dt, i∈[1,15]. Harmonic distortion is calculated by the formula , i∈[2,15] calculation, reflecting the nonlinear characteristics of the waveform. Perform principal component analysis on high-order harmonics (above the third order) and extract the eigenvalue λ j and the eigenvector v j , j∈[1,5], construct a 5-dimensional orthogonal feature space. In this feature space, high-order harmonics are represented as h=∑α j ·v j , where α j is the projection coefficient, through the matrix operation h·v j Calculated. The feature vector contains the core features of high-frequency oscillation, parasitic ringing and electromagnetic interference in the switching waveform, which is used as the input of the subsequent frequency modulation algorithm.

[0072] S1034. Based on the real-time junction temperature, multivariate regression analysis is performed on the high-order harmonic feature vector to generate the temperature-related frequency attenuation coefficient.

[0073] Exemplarily, establish the harmonic eigenvector α j The functional relationship model with junction temperature T uses a generalized additive model (GAM) for nonlinear fitting. The fitting function is set to f(T)=∑β ij ·s(T,d f =4), where s is a cubic spline function and the degree of freedom d f =4,β ij is the fitting coefficient. Solve β by iterative weighted least squares method ij value, the iteration stop condition is that the residual square sum changes less than 0.001 or reaches the maximum number of iterations of 50 times. Based on the fitting results, the frequency attenuation coefficient matrix A(T)=[a 1 (T), a 2 (T), ..., a 5 (T)], where a i (T)=exp(-γ i ·(T-T0) / 100), γ i is the temperature sensitivity factor, which is determined by the model parameter β ij Through nonlinear transformation, T 0 ​​The reference temperature is 25°C. The frequency attenuation coefficient matrix associates the high-order harmonic characteristics with the junction temperature to achieve temperature-adaptive frequency control.

[0074] S1035, construct a segmented frequency control function according to the frequency attenuation coefficient to generate dynamic frequency modulation parameters.

[0075] Exemplarily, a segmented frequency control function is constructed according to the frequency attenuation coefficient to generate dynamic frequency modulation parameters. Exemplarily, based on the frequency attenuation coefficient matrix A(T), a three-segment frequency modulation function F(T) is designed. For the junction temperature T<120℃ range, the frequency is set to F 1 (T)=fs0·[1-0.1·∑a i (T)·w i ], fs0 is the nominal switching frequency 100kHz, w i is the weighting coefficient; for the range of 120℃≤T<150℃, the frequency is set to F 2 (T)=F 1 (120)-(T-120)·(F 1 (120)-F 3 (150)) / 30; for T≥150℃ range, the frequency is set to F 3 (T)=F 1 (120)·0.6·[1-∑a i (T)·w i' ],w i' is the weighting coefficient for the high temperature zone. The frequency calculation accuracy is 0.1kHz and the update period is 10μs, ensuring smooth frequency transition when the temperature fluctuates. The dynamic frequency modulation parameters include the frequency setting value F(T), the slope limiting factor k (10kHz / ℃) and the dead zone compensation coefficient λ (linearly changes with frequency, λ=0.2+0.8·F(T) / fs0). The oscillator frequency is adjusted in real time through the PWM generator of the microcontroller to achieve temperature-adaptive switching frequency dynamic modulation of the SiC MOSFET.

[0076] In some embodiments, based on the real-time junction temperature, a multivariate regression analysis is performed on the high-order harmonic feature vector to generate a temperature-related frequency attenuation coefficient, including: S341-S344.

[0077] S341, standardize the high-order harmonic feature vector and real-time junction temperature to generate a normalized feature matrix.

[0078] Exemplarily, the Z-score standardization method is used to process the data of each dimension in the high-order harmonic feature vector, and the mean and standard deviation of each dimension are calculated. The feature value is subtracted from the mean and divided by the standard deviation, so that the mean of each feature component is 0 and the standard deviation is 1. Real-time junction temperature data is collected by the sliding window method, with a window width of 200ms and a step size of 50ms to obtain multiple groups of temperature sampling points. The temperature data is also processed by minimum-maximum standardization and mapped to the interval [0, 1]. The normalized high-order harmonic feature vector and temperature data are combined to form an m×n-dimensional normalized feature matrix, where m is the number of samples, usually 1000-5000 sample points, and n is the number of feature dimensions, including 33 dimensions of harmonic features and 1 dimension of temperature features, for a total of 34-dimensional feature space. Normalization eliminates the differences between different dimensions, avoids large-value features dominating the regression model, and improves model stability and generalization ability.

[0079] S342. Based on the principal component analysis method, the normalized feature matrix is ​​reduced in dimension to obtain a low-dimensional feature space projection.

[0080] For example, the covariance matrix of the normalized feature matrix is ​​calculated, and the matrix dimension is 34×34, which reflects the correlation between the features. Solve the eigenvalues ​​and eigenvectors of the covariance matrix, sort them in descending order by the size of the eigenvalues, and select the first k eigenvectors with a cumulative contribution rate of 95%. The k value is usually between 5-8. Construct an eigenvector matrix W with a dimension of 34×k, and multiply the normalized feature matrix by the W matrix on the left to achieve a linear mapping transformation from 34 dimensions to k dimensions. The dimensionality reduction process retains the main variation information of the data, eliminates redundant dimensions and noise, and the low-dimensional feature space projection significantly improves the computational efficiency while retaining the original data structure. Principal component analysis also visualizes the distribution of high-dimensional data, displays sample clustering through two-dimensional or three-dimensional scatter plots, intuitively judges the correlation pattern between temperature and harmonic features, and assists in the parameter selection and optimization of subsequent regression models.

[0081] S343, perform elastic network regression modeling on the low-dimensional feature space projection to generate the temperature-harmonic coupling weight vector.

[0082] For example, an elastic network regression model is constructed. The loss function includes a mean square error term, an L1 regularization term, and an L2 regularization term. The L1 term coefficient α is set to 0.5, and the L2 term coefficient λ is set to 0.3 to achieve a compromise between sparsity and smoothness. The coordinate descent method is used to optimize the model parameters. The upper limit of the number of iterations is set to 1000 times, and the convergence threshold is 10 -6. The data set is divided into a training set and a test set in a ratio of 8:2. The model parameters are fitted on the training set, and the model performance is evaluated on the test set. The cross-validation root mean square error (RMSE) is less than 0.05, which is considered to be model convergence. The elastic network regression model outputs a temperature-harmonic coupling weight vector with a dimension of k×1, which represents the contribution of each principal component in the low-dimensional feature space to the temperature change. The positive and negative signs of the weight vector elements reflect the positive or negative correlation between the corresponding principal component and the temperature, and the absolute value indicates the intensity of the influence, which lays the foundation for the subsequent frequency modulation parameter calculation.

[0083] S344, perform cross-validation based on the temperature-harmonic coupling weight vector to generate a set of frequency attenuation coefficients.

[0084] For example, the K-fold cross-validation method is used, with the K value set to 10. The data set is evenly divided into 10 subsets, and 9 subsets are selected as training data each time, and the remaining 1 subset is used as validation data. The training-validation process is repeated 10 times, using a different subset combination each time. The average performance indicators of the 10 validations are calculated, including the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R²). A sensitivity analysis is performed on the validation results to identify the response curve of the model to temperature changes. Based on the response curve, the slope and intercept at the key temperature points are extracted to construct a piecewise linear function. The piecewise linear function is parameterized as a set of frequency attenuation coefficients, and the coefficient set includes the reference frequency f 0 , temperature coefficient kT, threshold temperature Tth, saturation attenuation ratio rsat and other core parameters. The frequency attenuation coefficient set is stored in a lookup table. In practical applications, the corresponding frequency modulation factor can be directly indexed according to the measured temperature to achieve fast switching and smooth transition of dynamic frequency modulation strategies.

[0085] In some embodiments, the dynamic frequency modulation parameters are iteratively optimized for dead time to obtain a control parameter combination of silicon carbide MOS, including: S1041-S1046.

[0086] S1041, generating a delay-frequency correlation function according to the dynamic frequency modulation parameter and the preset switch delay parameter.

[0087] Exemplarily, a polynomial fitting method is used to construct a delay-frequency mapping function, and the dynamic frequency modulation parameter space is divided into 8 sub-intervals, each of which is described by a third-order polynomial. A total of 16 sets of key parameter points, including the rising edge delay, falling edge delay, and turn-on delay time at different switching frequencies, are recorded, and the precise correlation curve between the delay parameter and the frequency is obtained by least squares fitting. The model introduces a temperature correction factor, and a set of calibration data is collected every 25°C in the range of -40°C to 150°C to correct the effect of temperature on the switching delay. The accuracy of the delay-frequency correlation function directly affects the accuracy of the subsequent dead time calculation, and the fitting error is controlled within 5ns.

[0088] S1042, calculate the initial dead time according to the delay-frequency correlation function to generate a reference dead time.

[0089] For example, the delay-frequency correlation function is applied to calculate the turn-off delay time and the turn-on delay time at each frequency operating point, and the difference is multiplied by a safety factor of 1.2 to obtain the reference dead time at different frequencies. The calculation process takes into account the dispersion of device characteristic parameters, introduces the Monte Carlo analysis method to simulate 1000 random working conditions, and takes the upper limit of the 95% confidence interval as the reference dead time. For the high-frequency working area (greater than 200kHz), an additional 25ns redundancy is added to cope with the high-frequency parasitic oscillation of silicon carbide MOS. The generated reference dead time shows a nonlinear curve with frequency. The reference dead time in the low-frequency area (less than 50kHz) is about 320ns, and the reference dead time in the high-frequency area (greater than 500kHz) is reduced to 140ns.

[0090] S1043, perform zero-crossing detection and differential extraction on the real-time drain-source voltage waveform and the real-time drain-source current waveform to obtain a transient feature quantity set.

[0091] For example, a high-speed sampling front-end circuit is used to collect the drain-source voltage signal at a rate of 2GS / s and the drain-source current at a rate of 1GS / s. Wavelet transform is applied for noise suppression, and the signal-to-noise ratio is improved to more than 58dB. The waveform differential value is calculated based on the improved three-point central difference method, and the threshold is set to the voltage change rate of 15V / μs and the current change rate of 20A / μs. A total of 14 transient feature quantities are extracted, including the zero crossing moment, the maximum differential value and its occurrence time, the voltage rise / fall time, the current rise / fall time and the ringing amplitude. The sliding window technology is used to update the feature quantity set in real time, with a window width of 8 switching cycles and a sliding step of 1 cycle to ensure the timeliness of the feature quantity.

[0092] S1044, perform dead time error analysis based on transient characteristic quantity set analysis to generate dynamic correction quantity.

[0093] Exemplarily, a dead time error prediction model based on a neural network is constructed. The input layer receives 14 transient feature quantities, and the hidden layer adopts a two-layer structure containing 32 and 16 neurons respectively. The activation function uses LeakyReLU to improve the model's ability to fit nonlinear characteristics. The model reversely calculates the error value between the optimal dead time and the current benchmark dead time by analyzing the deviation between the energy loss in the voltage-current overlap area and the ideal minimum loss. An exponential smoothing filter is applied to the prediction results, and the smoothing coefficient is set to 0.7 to suppress sudden prediction deviations. The dead time error analysis is calibrated under different load current and junction temperature conditions, and the generated dynamic correction value ranges from -50ns to +30ns, with an accuracy controlled within 2ns.

[0094] S1045, iteratively correct the reference dead time according to the dynamic correction amount to generate an optimized dead time sequence.

[0095] Exemplarily, an adaptive proportional-integral algorithm is used to dynamically adjust the reference dead time, with the proportional coefficient Kp set to 0.8, the integral coefficient Ki set to 0.05, and the iteration step length being one cycle of the drive signal. A gradient descent constraint is introduced during the iteration process, and the single correction amount does not exceed 15% of the reference dead time. The dead time adjustment history for 32 consecutive cycles is recorded, and its change trend is analyzed. When the dead time change amplitude for 8 consecutive cycles is less than 5ns, it is determined to be in a convergence state. For load mutation conditions, a fast response mechanism is activated. When the load current changes by more than 20% of the rated value, the proportional coefficient is temporarily increased to 1.2 to accelerate the dead time adjustment process. The optimized dead time sequence is stored in the controller RAM in the form of a data table, containing the corresponding dead time values ​​for 256 frequency points and 128 temperature points.

[0096] S1046, optimize the optimized dead time sequence according to the preset safety constraints to generate a control parameter combination of silicon carbide MOS.

[0097] For example, the safety constraints are set to include a minimum dead time lower limit of 100ns, a breakdown protection margin of 15V, a junction temperature protection threshold of 150°C, and a switch overcurrent protection value of 1.8 times the rated current. A global traversal check is performed on the optimized dead time sequence to ensure that the dead time of all operating points meets the minimum safety margin. For critical operating conditions, the thermoelectric coupling simulation model is called to predict the junction temperature change rate. When the predicted junction temperature change rate exceeds 5°C / s, the additional margin of the dead time is automatically increased. The control parameter combination not only includes a frequency-dead time correspondence table, but also generates the optimal configuration of eight key parameters such as drive voltage, gate drive resistance, and Miller clamping voltage. The parameter combination is called in real time through a lookup table, and the switching loss and reliability are effectively balanced by controlling the response delay within one PWM cycle.

[0098] Please refer to Figure 2 , Figure 2 This is a schematic block diagram of a temperature compensation device for a silicon carbide MOS drive control chip provided in an embodiment of the present application. The temperature compensation device 200 for the silicon carbide MOS drive control chip is used to perform the temperature compensation method for the silicon carbide MOS drive control chip described above. The temperature compensation device 200 for the silicon carbide MOS drive control chip can be configured in a server.

[0099] Among them, the server can be an independent server or a server cluster, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0100] If Figure 2 As shown in FIG. 1 , the temperature compensation device 200 for the silicon carbide MOS driving control chip includes: a data acquisition module 201, a data optimization module 202, a parameter generation module 203 and a result output module 204.

[0101] Data acquisition module 201 is used to obtain the junction temperature-electrical parameter set of silicon carbide MOS, perform thermoelectric coupling simulation modeling based on the junction temperature-electrical parameter mapping data set, and obtain three-dimensional temperature-efficiency characteristic data.

[0102] Data optimization module 202 is used to perform real-time parameter analysis and dynamic execution on the three-dimensional temperature-efficiency characteristic data to obtain an optimized switching characteristic waveform.

[0103] Parameter generation module 203 is used to extract the frequency domain characteristics of the optimized switching characteristic waveform and generate dynamic frequency modulation parameters.

[0104] Result output module 204 is used to perform dead time iterative optimization on dynamic frequency modulation parameters to obtain a control parameter combination of silicon carbide MOS.

[0105] The embodiment of the present application provides an electronic device, the electronic device includes a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and implement the temperature compensation method of the silicon carbide MOS drive control chip as any one of the embodiments of the present application when executing the computer program.

[0106] The present application embodiment provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the processor implements a temperature compensation method for a silicon carbide MOS drive control chip as described in any one of the embodiments of the present application. ​​

[0107] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A temperature compensation method for a silicon carbide MOS drive control chip, characterized in that: The silicon carbide MOS drive control chip is equipped with a temperature-efficiency characteristic model, and the method includes: Acquire a junction temperature-electrical parameter set of the silicon carbide MOS, perform thermoelectric coupling simulation modeling based on the junction temperature-electrical parameter mapping data set, and obtain three-dimensional temperature-efficiency characteristic data; Perform real-time parameter analysis and dynamic execution on the three-dimensional temperature-efficiency characteristic data to obtain the optimized switching characteristic waveform; Extracting the frequency domain characteristics of the optimized switching characteristic waveform to generate dynamic frequency modulation parameters; The dead time is iteratively optimized for the dynamic frequency modulation parameters to obtain a control parameter combination of the silicon carbide MOS.

2. The temperature compensation method of the silicon carbide MOS drive control chip according to claim 1, characterized in that: The step of obtaining a junction temperature-electrical parameter set of the silicon carbide MOS includes: The junction temperature test interval to be tested is divided nonlinearly to determine the junction temperature sampling point; Through a preset double pulse test platform and preset test conditions, switch transient data collection is performed on each of the junction temperature sampling points to obtain gate-source voltage waveform data, drain-source current waveform data, and drain-source voltage waveform data; Determine a junction temperature-electrical parameter matrix according to the gate-source voltage waveform data, the drain-source current waveform data and the drain-source voltage waveform data, wherein the junction temperature-electrical parameter matrix includes: junction temperature sequence, on-resistance, switching time and switching energy loss; Perform thermoelectric coupling compensation processing on the junction temperature-electrical parameter matrix to obtain a corrected parameter mapping relationship; The modified parameter mapping relationship is subjected to multi-physics field verification processing to generate a junction temperature-electrical parameter set.

3. The temperature compensation method of the silicon carbide MOS drive control chip according to claim 1, characterized in that: The real-time parameter analysis and dynamic execution of the three-dimensional temperature-efficiency characteristic data to obtain an optimized switching characteristic waveform includes: The real-time junction temperature is implanted into the three-dimensional temperature-efficiency characteristic model data according to a preset interpolation method to obtain a dynamic parameter set at the current junction temperature; Performing gate voltage compensation on the dynamic parameter set to generate a gate voltage adjustment instruction; Optimize the gate resistance of the dynamic parameter set and generate a gate resistance adjustment instruction; Performing discrete coding according to the gate voltage adjustment instruction and the gate resistance adjustment instruction to generate a first driving parameter; Performing switching cycle synchronization processing on the driving parameter according to a preset switching cycle to obtain a second driving parameter; Calculate a rising edge slope and a falling edge slope according to the second driving parameter, and generate a third driving parameter for adjusting on and off according to the rising edge slope and the falling edge slope; Performing switching transient correction on the third driving parameter to generate a feedback correction coefficient set; A space equation is constructed according to the feedback correction coefficient set, and is analyzed using a Kalman filter to obtain the optimized switching characteristic waveform, which includes: a target gate voltage and a target gate resistance.

4. The temperature compensation method of the silicon carbide MOS drive control chip according to claim 3, characterized in that: The performing switching transient correction on the third driving parameter to generate a feedback correction coefficient set includes: Performing high-speed sampling and differentiation processing on the driving waveform corresponding to the third driving parameter to extract a set of switching transient feature quantities; Performing dynamic deviation analysis on the set of switching transient characteristic quantities to generate an original correction factor matrix; The original correction factor matrix is ​​corrected by wavelet packet denoising to obtain a filtered correction factor; According to the preset correction factor validity criterion, the confidence verification is performed on the filtered correction factor to generate the feedback correction coefficient set.

5. The temperature compensation method of the silicon carbide MOS drive control chip according to claim 1, characterized in that: The step of extracting the optimized switching characteristic waveform frequency domain features and generating dynamic frequency modulation parameters comprises: Segmenting the optimized switching characteristic waveform to obtain switching cycle timing data; Performing short-time Fourier transform on the switching cycle timing data to generate a time-frequency distribution matrix; Perform harmonic energy spectrum analysis on the time-frequency distribution matrix to extract high-order harmonic feature vectors; Based on the real-time junction temperature, a multivariate regression analysis is performed on the high-order harmonic characteristic vector to generate a frequency attenuation coefficient related to temperature; A segmented frequency control function is constructed according to the frequency attenuation coefficient to generate the dynamic frequency modulation parameter.

6. The temperature compensation method of the silicon carbide MOS drive control chip according to claim 5, characterized in that: The method of performing multivariate regression analysis on the high-order harmonic characteristic vector based on the real-time junction temperature to generate a frequency attenuation coefficient related to temperature includes: Normalizing the high-order harmonic characteristic vector and the real-time junction temperature to generate a normalized characteristic matrix; Based on the principal component analysis method, the normalized feature matrix is ​​reduced in dimension to obtain a low-dimensional feature space projection; Performing elastic network regression modeling on the low-dimensional feature space projection to generate a temperature-harmonic coupling weight vector; A cross validation is performed based on the temperature-harmonic coupling weight vector to generate a set of frequency attenuation coefficients.

7. The temperature compensation method of the silicon carbide MOS drive control chip according to claim 1, characterized in that: The dead time iterative optimization of the dynamic frequency modulation parameters is performed to obtain the control parameter combination of the silicon carbide MOS, including: generating a delay-frequency correlation function according to the dynamic frequency modulation parameter and a preset switch delay parameter; Calculating an initial dead time according to the delay-frequency correlation function to generate a reference dead time; Perform zero-crossing detection and differential extraction on the real-time drain-source voltage waveform and the real-time drain-source current waveform to obtain a transient feature quantity set; Perform dead time error analysis based on the transient characteristic quantity set analysis to generate a dynamic correction quantity; Iteratively correct the reference dead time according to the dynamic correction amount to generate an optimized dead time sequence; The optimized dead time sequence is optimized according to preset safety constraints to generate a control parameter combination of the silicon carbide MOS.

8. A temperature compensation device for a silicon carbide MOS drive control chip, characterized in that: The temperature compensation device of the silicon carbide MOS drive control chip is used to perform the temperature compensation method of the silicon carbide MOS drive control chip according to any one of claims 1 to 7, and the temperature compensation device of the silicon carbide MOS drive control chip comprises: A data acquisition module is used to acquire a junction temperature-electrical parameter set of the silicon carbide MOS, and perform thermoelectric coupling simulation modeling based on the junction temperature-electrical parameter mapping data set to obtain three-dimensional temperature-efficiency characteristic data; Data optimization module, used to perform real-time parameter analysis and dynamic execution of three-dimensional temperature-efficiency characteristic data to obtain optimized switching characteristic waveform; A parameter generation module, used to extract the frequency domain characteristics of the optimized switching characteristic waveform and generate dynamic frequency modulation parameters; The result output module is used to perform dead time iterative optimization on the dynamic frequency modulation parameters to obtain a control parameter combination of the silicon carbide MOS.

Images