A method and system for collaborative control of power semiconductor devices in a converter valve

Abstract

The present invention discloses a method and system for collaborative control of power semiconductor devices in a converter valve, which is applied to the technical field of converter valve control. By real-time acquisition of voltage transient response data at both ends of an IGBT and an IGCT, the voltage transient response data during the switching process is captured to obtain a time series data set, and then a fast Fourier transform is performed on the adjusted time series data set to obtain a fundamental wave amplitude. Then, the switching state of the converter valve is determined based on the fundamental wave amplitude, and switching loss data is obtained. Based on the switching loss data, adjusted switching timing data is obtained; an RL series circuit model is simulated based on the adjusted timing data set input, and the insulation scheme is adjusted based on current uniformity data. Then, the control strategy is adjusted through dynamic impedance matching data, thereby achieving high-reliability and high-efficiency power semiconductor power electronic device control and system optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of converter valve control, and in particular to a method and system for collaboratively controlling power semiconductor devices in a converter valve. Background Art

[0002] In DC transmission projects, converter valves, as core components, face a series of technical challenges. First, when integrated gate-commutated thyristors (IGCTs) and insulated gate bipolar transistors (IGBTs) are used in series, their voltage transient response and current distribution uniformity vary. This not only leads to inconsistent switching time delays, compromising series coordination, but also causes uneven thermal stress distribution in the power devices. This uneven thermal stress further complicates the design of heat dissipation channels and the optimization of electrical insulation spacing. Furthermore, the layout of modular connection interfaces must consider material dielectric constants and assembly tolerances to ensure structural stability. Variations in surface charge density during operation can affect power factor correction, requiring real-time monitoring and adjustment. These intertwined issues, for example, can lead to uneven current distribution due to voltage transient response, which in turn leads to concentrated thermal stress. This further complicates the design of heat dissipation channels and electrical insulation, ultimately increasing the cost of converter station equipment and reducing system performance. Therefore, a comprehensive solution is urgently needed to optimize the overall efficiency and reliability of DC transmission systems. Summary of the Invention

[0003] In order to solve the above technical problems, an embodiment of the present invention provides a method and system for coordinated control of power semiconductor devices in a converter valve to solve the problem that the overall efficiency of the DC transmission system is reduced when IGCT and IGBT are used in series in the existing DC transmission system.

[0004] A first aspect of an embodiment of the present invention provides a method for cooperatively controlling power semiconductor devices in a converter valve, comprising:

[0005] Acquire voltage transient response data across the IGCT and the IGBT to obtain a time series data set, and adjust the time series data set to obtain an adjusted time series data set, wherein the voltage transient response data includes an instantaneous voltage value and an instantaneous current value;

[0006] Performing a fast Fourier transform on the adjusted time series data set to obtain a fundamental wave amplitude, determining the switching state of the converter valve based on the fundamental wave amplitude, obtaining switching loss data based on the switching state, and obtaining adjusted switching timing data based on the switching loss data;

[0007] Based on the adjusted time series data set, current distribution uniformity data of the converter valve is obtained, and a final electrical insulation optimization design scheme is obtained based on the current distribution uniformity data. Based on the bandwidth data in the final electrical insulation optimization design scheme, a preliminary layout sequence is obtained using switching state change characteristics and preset material dielectric constant constraints, and the preliminary layout sequence is adjusted to obtain a final optimization sequence.

[0008] According to the thermal expansion matching data in the final optimization sequence, the final structural rigidity coefficient is obtained. Based on the final structural rigidity coefficient, the surface charge density suppression scheme is obtained. According to the dynamic impedance matching data in the surface charge density suppression scheme, the adjusted switching threshold update rule is obtained, so that the power system can control the power electronic devices according to the adjusted switching threshold update rule.

[0009] In a possible implementation of the first aspect, adjusting the time series dataset to obtain the adjusted time series dataset includes:

[0010] Perform time domain analysis on the time series data set to obtain the switching time and delay time;

[0011] Input the switching time and the delay time into a linear weighting module for calculation to obtain a distribution feature parameter set. If any distribution feature parameter in the distribution feature parameter set is greater than a first preset threshold, feature extraction is performed on the time series data set to obtain an optimized feature vector.

[0012] Determine whether the optimized eigenvector is abnormal. If it is abnormal, use cubic spline interpolation to correct the time series data set to obtain the adjusted time series data set.

[0013] In a possible implementation of the first aspect, obtaining adjusted switching timing data according to the switching loss data includes:

[0014] The switching loss data is input into the three-dimensional thermoelectric coupling model for calculation to obtain the junction temperature fluctuation range;

[0015] An iterative solution is performed with the goal of minimizing the difference in the junction temperature fluctuation range to obtain the optimal delay compensation amount;

[0016] The PWM dead time of the driving circuit is adjusted according to the optimal delay compensation amount to obtain adjusted switching timing data.

[0017] In a possible implementation of the first aspect, obtaining a final electrical insulation optimization design solution based on current distribution uniformity data includes:

[0018] Analyze the current distribution uniformity data to obtain a distribution feature set;

[0019] Based on the distribution feature set, a geometric optimization scheme for the heat dissipation channel is obtained, the geometric optimization scheme is optimized according to the heat flux density to obtain a preliminary adjustment scheme, the spacing data in the preliminary adjustment scheme is optimized to obtain an optimized spacing distribution, and based on the optimized spacing distribution, it is determined whether the separation of the power electronic devices of the IGCT and the IGBT meets a second preset threshold value. If so, a final spacing adjustment scheme is obtained;

[0020] Based on the final spacing adjustment plan, the finite element analysis tool was used to verify the thermal stress equilibrium state and obtain the current distribution uniformity verification results;

[0021] Based on the current distribution uniformity verification results, the final electrical insulation optimization design scheme is obtained.

[0022] In a possible implementation of the first aspect, adjusting the preliminary layout sequence to obtain a final optimized sequence includes:

[0023] Based on the switching state change characteristics of IGCT and IGBT during the switching process, the bandwidth data distribution range is obtained according to the switching state change characteristics;

[0024] Extract features from the bandwidth data distribution range to obtain switching threshold features. Based on the switching threshold features, obtain dynamic adjustment requirements. Based on the dynamic adjustment requirements and preset material dielectric constant constraints, obtain the threshold range under dielectric constant constraints.

[0025] According to the threshold range under the dielectric constant constraint, the layout structure of the modular connection interface is optimized using the dynamic programming algorithm to obtain a preliminary layout sequence;

[0026] Based on the deviation data of the preliminary layout sequence and assembly tolerance, the final optimized sequence is obtained.

[0027] In a possible implementation of the first aspect, obtaining a final structural rigidity coefficient according to thermal expansion matching data in a final optimization sequence includes:

[0028] Perform dimensionality reduction on the thermal expansion matching data in the final optimization sequence to obtain the optimized matching feature vector;

[0029] According to the optimized matching degree feature vector, charge density data is obtained, the charge density data is filtered to obtain a smoothed charge density curve, and the frequency domain feature of the smoothed charge density curve is extracted to obtain the main frequency distribution range;

[0030] If the main frequency distribution range is within the preset frequency interval, the control signal is bandwidth-limited to obtain a bandwidth-limited control signal;

[0031] According to the control signal after bandwidth limitation, the structural rigidity coefficient is obtained, and the structural rigidity coefficient is smoothed to obtain a rigidity coefficient curve. If the slope of the rigidity coefficient curve is greater than the preset slope threshold, the optimized matching degree feature vector is adjusted until the slope of the rigidity coefficient curve is less than the preset slope threshold to obtain the final structural rigidity coefficient.

[0032] In a possible implementation of the first aspect, obtaining an adjusted switching threshold update rule based on dynamic impedance matching data in a surface charge density suppression scheme includes:

[0033] According to the dynamic impedance matching data in the surface charge density suppression scheme, the final switching threshold update rule is obtained;

[0034] The switching state data and switching time data are obtained to obtain the dynamic characteristics of the delay change, the dynamic characteristics of the delay change are fitted to obtain a smoothed correction fluctuation result, and an adjustment range is obtained based on the smoothed correction fluctuation result. The final switching threshold update rule is adjusted based on the adjustment range to obtain an adjusted switching threshold update rule.

[0035] In a possible implementation of the first aspect, obtaining a final switching threshold update rule based on dynamic impedance matching data in a surface charge density suppression scheme includes:

[0036] Perform distributed calculations on the surface charge density data in the surface charge density suppression scheme to obtain the changing trend of dynamic impedance matching;

[0037] Based on the changing trend of dynamic impedance matching, the current distribution uniformity region is divided to obtain the initial division result. The initial division result is subjected to finite element analysis to obtain the initial threshold of the switching state. If the initial threshold of the switching state is not within the preset range, the geometric parameters of the heat dissipation channel are adjusted to obtain the optimized switching state.

[0038] According to the optimized switching state, the correction data of dynamic impedance matching is obtained, the current distribution adjustment parameters are extracted from the correction data, and the updated configuration scheme of the electrical insulation distance is obtained. According to the updated configuration scheme of the electrical insulation distance, the final switching threshold update rule is determined.

[0039] In order to solve the same technical problem, a second aspect of an embodiment of the present invention provides a coordinated control system for power semiconductor devices in a converter valve, comprising:

[0040] An acquisition module is used to acquire voltage transient response data across the IGCT and the IGBT to obtain a time series data set, and adjust the time series data set to obtain an adjusted time series data set; wherein the voltage transient response data includes an instantaneous voltage value and an instantaneous current value;

[0041] a first adjustment module, configured to perform a fast Fourier transform on the adjusted time series data set to obtain a fundamental wave amplitude, determine the switching state of the converter valve based on the fundamental wave amplitude, obtain switching loss data based on the switching state, and obtain adjusted switching timing data based on the switching loss data;

[0042] a second adjustment module, configured to obtain current distribution uniformity data of the converter valve based on the adjusted time series data set, obtain a final electrical insulation optimization design scheme based on the current distribution uniformity data, obtain a preliminary layout sequence based on the bandwidth data in the final electrical insulation optimization design scheme, utilize switching state variation characteristics and preset material dielectric constant constraints, and adjust the preliminary layout sequence to obtain a final optimization sequence;

[0043] The third adjustment module is used to obtain the final structural stiffness coefficient based on the thermal expansion matching data in the final optimization sequence, obtain the surface charge density suppression scheme based on the final structural stiffness coefficient, and obtain the adjusted switching threshold update rule based on the dynamic impedance matching data in the surface charge density suppression scheme, so that the power system can control the power electronic devices according to the adjusted switching threshold update rule.

[0044] In a possible implementation of the second aspect, the acquisition module further includes a time domain analysis unit, a judgment unit, and a correction unit, wherein:

[0045] A time domain analysis unit is used to perform time domain analysis on a time series data set to obtain switching time, and based on the switching time, obtain switching time and delay time;

[0046] a judgment unit, configured to input the switching time and the delay time into a linear weighting module for calculation to obtain a set of distribution feature parameters, and if any distribution feature parameter in the set of distribution feature parameters is greater than a first preset threshold, perform feature extraction on the time series data set to obtain an optimized feature vector;

[0047] The correction unit is used to determine whether the optimized feature vector is abnormal. If it is abnormal, the time series data set is corrected using cubic spline interpolation to obtain an adjusted time series data set.

[0048] The technical solution of the present invention has the following advantages:

[0049] The embodiment of the present invention provides a coordinated control method for power semiconductor devices in a converter valve. By real-time acquisition of voltage / current data across the IGBT / IGCT, the voltage transient response data during the switching process is captured to obtain a time series data set. The adjusted time series data set is then fast Fourier transformed to obtain a fundamental wave amplitude. The switching state of the converter valve is then determined based on the fundamental wave amplitude, and switching loss data is obtained. Adjusted switching timing data is obtained based on the switching loss data. The RL series circuit model is simulated based on the adjusted timing data set input, and the insulation scheme is adjusted based on current uniformity data. The heat dissipation layout is optimized in combination with thermal stress simulation to reduce mechanical stress concentration caused by thermal deformation. The control strategy is then adjusted through dynamic impedance matching data, thereby achieving high-reliability and high-efficiency power electronic device (IGBT / IGCT) control and system optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific implementation methods or the description of the prior art. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0051] Figure 1 This is a structural diagram of a method for cooperatively controlling power semiconductor devices in a converter valve according to an embodiment of the present invention;

[0052] Figure 2 This is a system block diagram of a coordinated control system for power semiconductor devices in a converter valve according to an embodiment of the present invention;

[0053] Figure numerals: 200, collaborative control system of power semiconductor devices in converter valves; 201, acquisition module; 202, first adjustment module; 203, second adjustment module; 204, third adjustment module. DETAILED DESCRIPTION

[0054] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0055] In the description of the present invention, it should be noted that the terms "first", "second" and "third" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance.

[0056] The embodiment of the present invention provides a method for cooperatively controlling power semiconductor devices in a converter valve, such as Figure 1 FIG. 1 is a flow chart of a method for cooperatively controlling power semiconductor devices in a converter valve, including steps S101 to S104. The details of each step are as follows:

[0057] S101: Obtain voltage transient response data across the IGCT and the IGBT to obtain a time series data set, and adjust the time series data set to obtain an adjusted time series data set, wherein the voltage transient response data includes an instantaneous voltage value and an instantaneous current value.

[0058] In this embodiment, the voltage transient response data of the power electronic devices in the converter valve are collected by current sensors and voltage sensors, where the power electronic devices refer to IGCT and IGBT. According to the Nyquist sampling theorem, the sampling frequency must be at least twice the upper limit of the switching frequency of the power electronic devices. The theorem states that in order to accurately obtain the voltage transient response data from the sampled signal, the sampling frequency must be at least twice the upper limit of the switching frequency of the power electronic devices. Restore the original signal , the sampling frequency should be at least twice the highest frequency of the original signal, that is, When actually used in data collection of power electronic devices of converter valves, the upper limit of the switching frequency of the power electronic devices can be regarded as the highest frequency of the original signal. For example, if the upper limit of the IGBT switching frequency is According to the Nyquist sampling theorem, the sampling frequency should not be less than Taking into account actual electromagnetic interference, signal attenuation and other factors, in order to ensure the accuracy and completeness of the collected data, the sampling frequency is usually appropriately increased based on the theoretical value.

[0059] It should be noted that the sampling frequency is preferably set to This choice can not only meet the requirements of accurate signal acquisition, but also resist interference to a certain extent, ensuring that the collected data can accurately reflect the operating status of power electronic devices and provide reliable data support for subsequent time domain analysis and feature calculation.

[0060] It is worth noting that the voltage transient response data refers to the instantaneous voltage and current values across the IGCT and IGBT.

[0061] After obtaining the voltage transient response data, the time-domain analysis method is used to adjust the time series data set to obtain an adjusted time series data set.

[0062] In one embodiment, adjusting the time series data set to obtain the adjusted time series data set includes:

[0063] Perform time domain analysis on the time series data set to obtain the switching time and delay time;

[0064] Input the switching time and the delay time into a linear weighting module for calculation to obtain a distribution feature parameter set. If any distribution feature parameter in the distribution feature parameter set is greater than a first preset threshold, feature extraction is performed on the time series data set to obtain an optimized feature vector.

[0065] Determine whether the optimized eigenvector is abnormal. If it is abnormal, use cubic spline interpolation to correct the time series data set to obtain the adjusted time series data set.

[0066] In this embodiment, time-domain analysis is first performed on the time series data set. Voltage and current waveforms are obtained from the time series data set. The voltage waveform is analyzed, and the 10% to 90% intervals between the rising and falling edges are detected to calculate the switching time. For example, if the voltage takes 200 ns to rise from 10% to 90% in a given sampling period, the switching time can be considered to be 200 ns. Simultaneously, the zero-crossing point of the current waveform is detected to calculate the delay time, for example, 50 ns.

[0067] The switching time and delay time are then input into a pre-established linear weighted model for processing to obtain a set of distribution feature parameters. The distribution feature parameter set includes statistics such as mean and variance. If any statistic exceeds the first preset threshold, the anomaly detection process is entered. The specific process of the anomaly detection process is as follows: the time series data set is input into a convolutional neural network for processing, the optimized feature vector is output, and the optimized feature vector is classified using a radial basis kernel support vector machine to obtain a classification result. Based on the classification result, it is determined whether the optimized feature vector is abnormal. If it is determined to be an abnormal state, the time series data set is corrected using cubic spline interpolation to obtain an adjusted time series data set. This method can effectively remove abnormal fluctuations while maintaining data continuity, thereby improving the data quality of the time series data set.

[0068] It is worth noting that the weights in the linear weighted model are determined using the analytic hierarchy process (AHP) based on the rated voltage and rated current of the IGCT and IGBT power electronic devices, and their importance in influencing the switching characteristics. The first preset threshold for each statistic can be set based on actual needs; for example, the first preset threshold for the variance is 5%. The convolutional neural network is a three-layer convolutional neural network, each followed by a maximum pooling layer to extract optimized feature vectors. This structure effectively captures the local characteristics and global trends of the voltage and current waveforms. The radial basis function support vector machine (RBF-SVM) is a variant of the support vector machine (SVM). It uses the radial basis function kernel (RBFKernel) to map raw data into a high-dimensional space, solving nonlinear classification problems.

[0069] If uneven current or voltage distribution is detected during a sampling event, interpolation correction can smooth the current or voltage waveform, facilitating subsequent analysis. This multi-level data processing and analysis method can promptly detect abnormalities in power electronic components within converter valves, providing a crucial basis for preventive maintenance and improving the reliability and safety of power systems.

[0070] S102: Perform a fast Fourier transform on the adjusted time series data set to obtain a fundamental wave amplitude, perform a determination based on the fundamental wave amplitude, obtain the switching state of the converter valve, obtain switching loss data based on the switching state, and obtain adjusted switching timing data based on the switching loss data.

[0071] In this embodiment, the adjusted time series data set is processed by fast Fourier transform, which can convert the time domain signal into the frequency domain, extract the key frequency components, and obtain the fundamental amplitude. The adjusted time series data set collected in the converter valve may contain the fundamental and multiple harmonics. For example, after decomposition by fast Fourier transform, the fundamental frequency is usually 50Hz or 60Hz, while the third harmonic is 150Hz or 180Hz. Then, by setting the Hanning window function to suppress spectrum leakage, the spectrum aliasing caused by non-periodic signals can be effectively reduced. For example, assuming that the original signal sampling frequency is 10kHz and 1024 points are collected, after adding the Hanning window, the spectrum resolution is improved, and the fundamental amplitude and the third harmonic amplitude are clearer and more discernible, which is helpful for subsequent state judgment.

[0072] The switching status of the converter valve is then determined based on the fundamental wave amplitude. For example, if the fundamental wave amplitude exceeds a preset per-unit threshold, such as 0.9 pu, the switching status can be determined to be active. During the switching process of the converter valve, the voltage transient may experience significant amplitude jumps. This increase in fundamental wave amplitude reflects the switching behavior of the power electronic device. This determination method allows for rapid identification of the operating status and facilitates real-time monitoring.

[0073] After obtaining the switching state, the switching loss data is collected, and the adjusted switching timing data is obtained based on the switching loss data.

[0074] It should be noted that switching loss data refers to the energy loss caused by switching operations (turning on / off) of power electronic devices (such as IGBTs and IGCTs) during switching operations.

[0075] Furthermore, calculating the switching delay using the transmission line characteristic impedance Z0 and the load impedance ZL can be understood as analyzing signal propagation characteristics using impedance matching. For example, if Z0 is 50Ω, the imaginary part of the load impedance ZL is small, and the fundamental frequency f is 50Hz, the switching delay calculated using the inverse tangent function can be in the microsecond range, such as 10μs. This method considers the physical characteristics of the circuit and more accurately reflects the switching behavior of IGBTs and IGCTs.

[0076] In one embodiment, obtaining adjusted switching timing data based on switching loss data includes:

[0077] The switching loss data is input into the three-dimensional thermoelectric coupling model for calculation to obtain the junction temperature fluctuation range;

[0078] An iterative solution is performed with the goal of minimizing the difference in the junction temperature fluctuation range to obtain the optimal delay compensation amount;

[0079] The PWM dead time of the driving circuit is adjusted according to the optimal delay compensation amount to obtain adjusted switching timing data.

[0080] In this implementation, a three-dimensional thermoelectric coupling model of IGBTs and IGCTs was established. Switching loss data was input into the model, and transient thermal simulations were performed to obtain the junction temperature fluctuation range as a thermal stress indicator. For example, assuming an IGBT switching loss of 2mJ / cycle, the simulation results showed a junction temperature fluctuation range of 15°C, while that of the IGCT was 10°C. This difference reflects the heterogeneity of the thermal characteristics of power electronic devices and provides a basis for subsequent optimization of switching timing data.

[0081] Next, with the goal of minimizing the difference in junction temperature fluctuations, a particle swarm optimization algorithm was used to iteratively determine the optimal delay compensation. The driver circuit's PWM dead time was then adjusted based on the optimal delay compensation. Timing was then adjusted using a preset controller, resulting in an adjusted timing data set. For example, adjusting the PWM dead time from the default 3μs to 8μs enhanced the switching synchronization between the IGBT and IGCT, reducing current spikes and improving the life of the power electronic components.

[0082] Furthermore, the adjusted timing data set can be input into a pre-established RL series circuit model for use in the next cycle judgment. Specifically, the RL series circuit model can be simplified to an equivalent circuit with a 1Ω resistor and a 0.1mH inductor. When using the RL series circuit model for judgment, the voltage and current phase differences are analyzed to determine whether the switching timing data still needs fine-tuning. If necessary, fine-tuning is continued based on the switching loss data. The fine-tuning process has been described in detail above and will not be repeated here. This closed-loop feedback mechanism continuously optimizes system performance.

[0083] It is worth noting that when using the particle swarm algorithm to iteratively solve for the optimal delay compensation, the particle swarm size was set to 30 and the number of iterations was 80. The particle swarm size was chosen to ensure sufficient coverage of the algorithm's search space while avoiding excessive particles that would waste computing resources and reduce efficiency. The number of iterations was determined through multiple trials to strike a balance between algorithm convergence accuracy and computational time. A larger particle swarm size and an appropriately increased number of iterations help increase the probability of the algorithm finding the global optimal solution and minimize differences in junction temperature fluctuations.

[0084] S103: Based on the adjusted time series data set, obtain the current distribution uniformity data of the converter valve, and obtain the final electrical insulation optimization design scheme based on the current distribution uniformity data. Based on the bandwidth data in the final electrical insulation optimization design scheme, use the switching state change characteristics and preset material dielectric constant constraints to obtain a preliminary layout sequence, and adjust the preliminary layout sequence to obtain the final optimization sequence.

[0085] In this embodiment, adjusting the switching timing alters the synchronous switching characteristics of power electronic devices (IGBTs / IGCTs), leading to differences in transient current distribution across parallel branches and, in turn, current unevenness (e.g., current in a particular branch exceeding the mean by 20%). This localized concentration of current can exacerbate heat accumulation in the corresponding area, causing the local temperature rise of the insulation material to exceed safety thresholds. Based on this, real-time monitoring of current distribution uniformity data enables the generation of targeted electrical insulation optimization design solutions, such as increasing the thickness of local heat dissipation channels or using high-dielectric-strength materials.

[0086] Specifically, current distribution uniformity data is collected by monitoring the current density distribution during operation of power electronic devices. For example, in a power module, a thermal imager is used to measure temperature changes in different areas and infer the uniformity of the current distribution. If the temperature in one area is significantly higher than in other areas, this may indicate current concentration and poor uniformity. This method is intuitive and easy to implement, making it particularly suitable for the initial stages of thermal stress analysis.

[0087] A support vector machine algorithm was then used to analyze the current distribution uniformity data, using current density values as input features. The model was trained to identify regularities in the distribution. Based on this current distribution uniformity data, a final electrical insulation optimization design was derived. Based on the bandwidth data from this final electrical insulation optimization design, the switching state variation characteristics and the preset material dielectric constant constraints were used to determine a preliminary layout sequence. This preliminary layout sequence was then adjusted to obtain the final optimized sequence.

[0088] In one embodiment, a final electrical insulation optimization design solution is obtained based on the current distribution uniformity data, including:

[0089] Analyze the current distribution uniformity data to obtain a distribution feature set;

[0090] Based on the distribution feature set, a geometric optimization scheme for the heat dissipation channel is obtained, the geometric optimization scheme is optimized according to the heat flux density to obtain a preliminary adjustment scheme, the spacing data in the preliminary adjustment scheme is optimized to obtain an optimized spacing distribution, and based on the optimized spacing distribution, it is determined whether the separation of the power electronic devices of the IGCT and the IGBT meets a second preset threshold value. If so, a final spacing adjustment scheme is obtained;

[0091] Based on the final spacing adjustment plan, the finite element analysis tool was used to verify the thermal stress equilibrium state and obtain the current distribution uniformity verification results;

[0092] Based on the current distribution uniformity verification results, the final electrical insulation optimization design scheme is obtained.

[0093] In this embodiment, a support vector machine algorithm is used to analyze current distribution uniformity data to obtain a set of distribution features. For example, within an IGBT module, multiple sets of operating data are collected, with uniform and non-uniform states labeled. After analysis using the support vector machine algorithm, a set of distribution features, such as concentration or dispersion, is output. This approach effectively distinguishes current distribution characteristics under complex operating conditions, providing data support for subsequent optimization.

[0094] Then, when designing a geometric optimization scheme for the heat dissipation channel based on the distribution feature set, the channel width or depth is adjusted. For example, if the distribution feature set indicates current concentration in a certain area, the heat dissipation channel in that area can be widened to increase the heat flow removal capacity. For example, during the design, the channel width is increased from 2mm to 3mm for hot spots near the pin area to improve local heat dissipation efficiency. The geometric optimization scheme is then optimized based on the heat flux density to obtain a preliminary adjustment scheme. Specifically, assuming that the heat flux density gradually decreases from the center to the edge, the separation effect parameter can be defined by measuring the heat flux difference between the edge and the center. Numerically, if the difference is less than a preset value, such as 5W / cm², the separation effect is considered insufficient. Then, using the heat flux density difference as the objective function, a gradient descent algorithm is used to iteratively calculate the separation effect parameter, gradually adjusting the channel layout to obtain a preliminary adjustment scheme. For example, in the initial scheme, the channel spacing was 4mm, and through iteration, 3.5mm was found to be more optimal. This method uses data-driven approaching to the optimal solution, ensuring computational efficiency.

[0095] After obtaining the spacing data in the preliminary adjustment plan, a secondary iterative calculation is performed based on the changing trend of the thermal coupling relationship to determine the optimized spacing distribution. For example: if the spacing between two power electronic devices is reduced from 5mm to 4mm, the thermal coupling is enhanced. The temperature rise trend can be observed through simulation and 3.8mm is determined to be the optimized spacing. This method can balance heat dissipation and space utilization. Then, based on the optimized spacing distribution, it is determined whether the separation of power electronic devices meets the second preset threshold. If not, the iterative calculation is repeated until the condition is met to obtain the final spacing adjustment plan. For example, the second preset threshold can be set to 10°C. If the simulation shows that the temperature difference under a certain plan is 12°C, it is necessary to repeat the iteration until the condition is met. This repeated verification ensures the reliability of the design.

[0096] Furthermore, finite element analysis tools were used to verify the thermal stress equilibrium state based on the final spacing adjustment plan, obtaining verification results for the uniformity of current distribution. Based on these uniformity verification results, the design plan was integrated to determine the final optimized electrical insulation design. For example, by integrating this data, the insulation layer thickness was adjusted to 1.2mm, improving insulation performance while maintaining heat dissipation capacity. This comprehensive design can significantly improve the lifespan and stability of power electronic devices. Data integration specifically refers to the aggregation of data such as temperature and current distribution.

[0097] In one embodiment, adjusting the preliminary layout sequence to obtain a final optimized sequence includes:

[0098] Based on the switching state change characteristics of IGCT and IGBT during the switching process, the bandwidth data distribution range is obtained;

[0099] Extract features from the bandwidth data distribution range to obtain switching threshold features. Based on the switching threshold features, obtain dynamic adjustment requirements. Based on the dynamic adjustment requirements and preset material dielectric constant constraints, obtain the threshold range under dielectric constant constraints.

[0100] According to the threshold range under the dielectric constant constraint, the layout structure of the modular connection interface is optimized using the dynamic programming algorithm to obtain a preliminary layout sequence;

[0101] Based on the deviation data of the preliminary layout sequence and assembly tolerance, the final optimized sequence is obtained.

[0102] In this embodiment, obtaining control signal bandwidth data is a key step in electrical insulation optimization design. Bandwidth data generally reflects the frequency range of signal transmission. Specifically, bandwidth data can be extracted by measuring the effect of adjusted electrical insulation spacing on signal attenuation. Assuming an initial spacing of 2 mm and an adjusted spacing of 3 mm, the bandwidth may expand from 500 MHz to 700 MHz, providing a basis for subsequent analysis.

[0103] Quantifying the characteristics of switching state changes can be understood as converting the dynamic response of power electronic devices during switching into analyzable data, thereby determining the bandwidth distribution range. Specifically, this can be quantified by recording the amplitude of current or voltage fluctuations during the switching process. For example, at a switching frequency of 100 times per second, the bandwidth distribution range might be 600MHz to 800MHz. This quantification helps clarify the operating boundaries of the system.

[0104] Fourier transforms are used to extract switching threshold characteristics from the bandwidth data distribution range. For example, if bandwidth data fluctuates significantly around 650MHz, Fourier transform analysis can reveal the need for dynamic switching state adjustment at this frequency point, such as adjusting the switching threshold from 0.5V to 0.7V to accommodate fast switching scenarios.

[0105] Based on the switching threshold characteristics, the dynamic adjustment requirements for the switching state corresponding to the frequency point are analyzed. Based on this dynamic adjustment requirement and the preset material dielectric constant constraint, the threshold range under the dielectric constant constraint is obtained. Specifically, since the dielectric constant directly affects signal transmission speed and insulation performance, if a material with a dielectric constant of 4.5 is selected, the switching threshold range may be limited to 0.6V to 0.8V. However, if the dielectric constant is reduced to 3.5, the range can be expanded to 0.5V to 0.9V. This constraint provides a theoretical basis for subsequent optimization.

[0106] When using dynamic programming to optimize the layout structure of modular connection interfaces, the interface placement can be considered a multi-stage decision-making problem. Specifically, assuming there are five connection interfaces, the initial layout is linear. After dynamic programming is used to adjust it to a circular layout, the signal transmission delay can be reduced from 10ns to 8ns, making the initial layout sequence more efficient.

[0107] When adjusting the distribution of connection interfaces using a preliminary layout sequence, assembly tolerance deviation data is crucial. For example, after adjusting the interface spacing from 5mm to 5.2mm, the tolerance deviation distribution set may show a reduction in the deviation range from ±0.1mm to ±0.05mm, indicating a more stable layout after the adjustment. Therefore, by adjusting the distribution of connection interfaces using a preliminary layout sequence, assembly tolerance deviation data, namely the tolerance deviation distribution set, is obtained. Feature parameters of the optimization sequence are extracted from the tolerance deviation distribution set and classified using a support vector machine to obtain an assembly tolerance range optimization sequence. A linear regression algorithm is used to predict the stability of the switching state within the assembly tolerance range optimization sequence, resulting in the final optimized sequence.

[0108] It's important to note that support vector machine classification can categorize deviation ranges into "high precision" and "low precision." Deviations less than ±0.06mm are classified as high precision, and this classification helps identify optimal assembly solutions. Linear regression is used to predict the stability of switching states within the assembly tolerance range. For example, by analyzing 100 switching cycles, the predicted stability increased from 90% to 95%, making the final optimized sequence more reliable. This prediction can effectively guide design improvements.

[0109] S104: According to the thermal expansion matching data in the final optimization sequence, a final structural stiffness coefficient is obtained. Based on the final structural stiffness coefficient, a surface charge density suppression scheme is obtained. According to the dynamic impedance matching data in the surface charge density suppression scheme, an adjusted switching threshold update rule is obtained, so that the power system controls the power electronic devices according to the adjusted switching threshold update rule.

[0110] In this embodiment, the final structural rigidity coefficient is obtained based on the thermal expansion matching data in the final optimization sequence. Specifically, the thermal expansion matching data including temperature, displacement, and material parameters are first collected by sensors. In the manufacture of electrical insulation modules, the temperature may vary from 20°C to 80°C, the displacement records the small deformation of the assembly, such as 0.1mm, and the material parameters include the thermal expansion coefficient, such as . These data form a multi-dimensional time series, reflecting the dynamic process of thermal expansion matching. Then, the thermal expansion matching data and charge density data are processed to obtain the surface charge density suppression scheme. Specifically, the stable range data of the final structural rigidity coefficient is obtained, and the assembly is modeled and analyzed using finite element analysis software to obtain the maximum stress and deformation under different working conditions. For example, for a certain mechanical connector, the material is set to aluminum alloy, and different loads such as 10kN and 20kN are applied to simulate static and dynamic load conditions, respectively. The maximum stress is calculated to be approximately 150MPa and 200MPa, and the deformation is 0.2mm and 0.35mm, respectively. This method can intuitively reflect the range of change of the rigidity coefficient under different working conditions, which is helpful for subsequent optimization of the design.

[0111] Next, we analyzed the geometry of the heat dissipation channel. As can be seen, boundary conditions were first set in the multiphysics simulation software. The heat dissipation channel was designed with a rectangular cross-section, 50 mm long and 10 mm wide. When setting the convection heat transfer boundary conditions, the wall temperature was fixed at 80°C, and the air convection coefficient was set at 25 W / m²·K. Adjusting the channel width to 15 mm revealed a more uniform heat flux distribution. This boundary condition setting effectively simulates the actual heat dissipation environment and provides reliable data support for thermal expansion matching calculations.

[0112] The formula for calculating thermal expansion matching is: ΔL=αLΔT. Where L is the characteristic length, ΔT is the temperature difference, α is the expansion coefficient, and ΔL is the expansion matching. Assuming the material is steel, the thermal expansion coefficient α is 12×10 -6 / ℃, the characteristic length L is 100mm, the temperature difference ΔT is 50℃, then the thermal expansion matching degree ΔL is about 0.06mm. If the preset threshold is 0.08mm, then this value meets the requirement. For example, it can be verified from different material aspects, such as copper’s α is 17×10 -6 / ℃, ΔL increases to 0.085mm, exceeding the threshold, indicating that material selection has a significant impact on matching.

[0113] Then build the evaluation function , weight coefficient and Can be adjusted according to design requirements. =0.6, =0.4, ΔL is 0.06mm, σ is 1.5, then F=0.6×0.06+0.4×1.5=0.636. If σ rises to 2.0, F increases to 0.836, indicating that stress concentration needs to be optimized. Then, the distribution density of modular connection interfaces is optimized by genetic algorithm to obtain a surface charge density suppression scheme. For example, the initial distribution density is 4 interfaces per square centimeter, which is optimized to 6 after 50 iterations. Thermal-mechanical coupling simulation shows that the maximum stress drops from 180MPa to 160MPa, and the deformation is reduced from 0.3mm to 0.25mm. This optimization can significantly improve the stability of the structure. Furthermore, after adjusting the fitness function weight, the density converges to 5, reflecting the flexibility of the algorithm. For the calculation of surface charge density distribution, when using the electrostatic field module, the grid size can be set to 1mm. For example, the simulation results show that the charge density in a certain area 5μC / m², which is higher than the threshold of 4μC / m². At this time, adjust the genetic algorithm weight and re-optimize This method can effectively control the charge distribution and ensure the reliability of system operation.

[0114] In one embodiment, the final structural rigidity coefficient is obtained based on the thermal expansion matching data in the final optimization sequence, including:

[0115] Perform dimensionality reduction on the thermal expansion matching data in the final optimization sequence to obtain the optimized matching feature vector;

[0116] Acquire charge density data, filter the charge density data to obtain a smoothed charge density curve, perform frequency domain feature extraction on the smoothed charge density curve, and obtain a main frequency distribution range;

[0117] If the main frequency distribution range is within the preset frequency interval, the control signal is bandwidth-limited to obtain a bandwidth-limited control signal;

[0118] According to the control signal after bandwidth limitation, the structural rigidity coefficient is obtained, and the structural rigidity coefficient is smoothed to obtain a rigidity coefficient curve. If the slope of the rigidity coefficient curve is greater than the preset slope threshold, the optimized matching degree feature vector is adjusted until the slope of the rigidity coefficient curve is less than the preset slope threshold to obtain the final structural rigidity coefficient.

[0119] In this embodiment, a mapping relationship is established between the three-dimensional feature vector after dimensionality reduction and the surface charge density (such as through a regression model or a lookup table method) to dynamically predict high-risk areas of charge density. Charge density data is collected in real time by a surface strain charge density sensor. Specifically, the sensor is arranged on the surface of the insulating part, and the collected charge density may fluctuate due to environmental interference. For example, the raw data changes by 0.02μC / cm² per second. After noise suppression using a linear Gaussian Kalman filter with a preset noise covariance matrix, the data is smoothed to a fluctuation range of less than 0.01μC / cm². The variance matrix can be preset according to historical noise characteristics to ensure that the filter effectively separates the signal and noise.

[0120] For the smoothed charge density curve, a Hanning window-weighted discrete Fourier transform is used to extract frequency domain features, resulting in the dominant frequency distribution range. Using a Hanning window can reduce spectral leakage. For example, the transformed amplitude spectrum shows peaks at 5Hz and 10Hz, and the dominant frequency distribution range is determined to be 4-12Hz. This frequency domain analysis helps identify periodic variations in the signal. If the dominant frequency distribution range falls within a preset frequency range, such as 3-15Hz, the cutoff frequency of the adaptive filter is set to a predetermined value, such as 15Hz, and the control signal is bandwidth-limited to obtain a bandwidth-limited control signal. The amplitude of the bandwidth-limited signal is reduced from 2V to 1.5V to avoid high-frequency interference. This adjustment ensures that the signal is suitable for subsequent control steps.

[0121] Based on the bandwidth-limited control signal, calculate the ratio of its amplitude to the assembly's resonant frequency as the structural stiffness coefficient. For example, if the signal amplitude is 1.5V and the resonant frequency is 20Hz, the ratio is 0.075, and the structural stiffness coefficient is 0.075.

[0122] The structural rigidity coefficient is smoothed by the moving average method to obtain a rigidity coefficient curve. For example, 10 sampling points are processed by the moving average method to obtain a smooth rigidity coefficient curve, and the fluctuation range is reduced to 0.01. This smoothing process improves the stability of the rigidity assessment. If the slope of the rigidity coefficient curve is greater than the preset slope threshold, such as 0.05, the principal component analysis weight matrix iterative update algorithm is used to adjust the thermal expansion matching parameters, that is, the optimized matching eigenvector, until the slope of the rigidity coefficient curve is less than the preset slope threshold to obtain the final structural rigidity coefficient. For example, the initial weight is biased towards temperature influence, and the slope is reduced to below 0.03 by iteratively increasing the displacement weight. This iterative optimization enables the system to better adapt to thermal expansion changes and ensure the long-term stability of the assembly structure.

[0123] In one embodiment, based on the dynamic impedance matching data in the surface charge density suppression scheme, an adjusted switching threshold update rule is obtained, including:

[0124] According to the dynamic impedance matching data in the surface charge density suppression scheme, the final switching threshold update rule is obtained;

[0125] The switching state data and switching time data are obtained to obtain the dynamic characteristics of the delay change, the dynamic characteristics of the delay change are fitted to obtain a smoothed correction fluctuation result, and an adjustment range is obtained based on the smoothed correction fluctuation result. The final switching threshold update rule is adjusted based on the adjustment range to obtain an adjusted switching threshold update rule.

[0126] In this embodiment, dynamic impedance matching data is obtained for the surface charge density suppression scheme. In this scheme, dynamic impedance matching data refers to key parameters that optimize electric field distribution and suppress surface charge accumulation by real-time monitoring and adjustment of the electrical impedance characteristics of the IGBT module during high-frequency switching or variable operating conditions. Based on this dynamic impedance matching data, the final switching threshold update rule is derived.

[0127] Capture switching state data and switching time data to verify or optimize the final switching threshold update rules. For example, in a power system, the switching state may involve the switching of capacitor banks, while the switching time data records the precise moment when the power electronic devices operate. High-frequency sampling equipment is used to collect time series data on switching actions, such as recording delay values every 0.1 seconds. This generates a set of curves reflecting dynamic changes and identifies the dynamic characteristics of delay changes. This approach can intuitively demonstrate the trend of delay fluctuations over time, providing data support for subsequent analysis.

[0128] The least squares method is then used to fit the dynamic characteristics of the delayed change to the power factor, resulting in a smoothed, corrected fluctuation. For example, assuming the power factor fluctuates between 0.85 and 0.95, the least squares method can produce a smooth curve through linear or polynomial fitting. This smoothed result helps more accurately determine the system's operating status.

[0129] Based on the smoothed corrected fluctuation results, the adjustment range for the switching threshold is determined. The final switching threshold update rule is adjusted based on the adjustment range, resulting in the adjusted switching threshold update rule. Specifically, if the power factor is below 0.9, the switching threshold needs to be adjusted, potentially within a ±5% range. This results in the adjusted switching threshold update rule. For example, if the original threshold is 100 amperes, the adjustment range is 95 to 105 amperes. This range provides clear boundaries for subsequent optimization, preventing blind adjustments that could lead to system instability.

[0130] Since the distribution of thermal stress may vary depending on the design of the heat dissipation channel or the material properties, if the adjustment range exceeds the preset range, the thermal stress distribution is calculated using a finite element analysis tool to obtain the preliminary value of the equilibrium parameter. Based on the preliminary value of the equilibrium parameter, a linear regression analysis is used to analyze the mapping relationship between the thermal stress and the final distribution to obtain the optimized parameter set. For example, assuming that thermal stress is positively correlated with temperature, a regression analysis can be used to derive a mapping relationship where the thermal stress increases by 2 Pa for every increase of 1 degree Celsius. For example, in multiple experiments, after adjusting the air volume, the temperature stabilized at 60 degrees Celsius, and the thermal stress distribution became more uniform. The optimized parameter set includes the adjusted wind speed and channel width.

[0131] It should be noted that the final distribution refers to the spatial distribution state of thermal stress after the system reaches a steady state or is optimized.

[0132] Using the optimized parameter set, the control logic for the switching state is updated, and a stability criterion is used to determine the stability data of the system operation. The control logic is adjusted to reduce the switching frequency when the temperature is below 65 degrees Celsius. The stability criterion is verified by observing whether the system response converges. For example, after adjustment, the delay fluctuation is reduced from ±10% to ±2%, indicating that the system is more stable. The stability data of the system operation is then obtained, and the final distribution is corrected using the Kalman filter algorithm to obtain the output result of thermal stress balance. It is understandable that in one operation, the deviation between the predicted and actual thermal stress distribution values was 5 Pa. After the Kalman filter correction, the deviation was reduced to 1 Pa. The final output result shows that the thermal stress balance has improved. This correction can provide a more reliable operating basis for the system and optimize overall performance.

[0133] In one embodiment, the final switching threshold update rule is obtained based on the dynamic impedance matching data in the surface charge density suppression scheme, including:

[0134] Perform distributed calculations on the surface charge density data in the surface charge density suppression scheme to obtain the changing trend of dynamic impedance matching;

[0135] Based on the changing trend of dynamic impedance matching, the current distribution uniformity region is divided to obtain the initial division result. The initial division result is subjected to finite element analysis to obtain the initial threshold of the switching state. If the initial threshold of the switching state is not within the preset range, the geometric parameters of the heat dissipation channel are adjusted to obtain the optimized switching state.

[0136] According to the optimized switching state, the correction data of dynamic impedance matching is obtained, the current distribution adjustment parameters are extracted from the correction data, and the updated configuration scheme of the electrical insulation distance is obtained. According to the updated configuration scheme of the electrical insulation distance, the final switching threshold update rule is determined.

[0137] In this embodiment, dynamic impedance matching data is obtained in the surface charge density suppression scheme. In the surface charge density suppression scheme, dynamic impedance matching data refers to the key parameters for optimizing the electric field distribution and suppressing surface charge accumulation by real-time monitoring and adjustment of the electrical impedance characteristics of the IGBT module during high-frequency switching or variable operating conditions. Specifically, surface charge density data is measured in an experimental environment using a high-precision charge sensor. An electrostatic probe is used to scan along the surface of the heat dissipation channel to collect charge distribution data at different locations. The measurement range can be set to 0.1 to 10μC / m².

[0138] Using Apache Spark for distributed computing, the surface charge density data is sharded to analyze the dynamic impedance matching trends. Specifically, assuming a 1TB data volume, it can be divided into 100 subsets, and Spark's parallel computing capabilities can be leveraged to analyze the dynamic impedance matching trends. For example, when a heat dissipation channel is operating, the impedance may gradually change from 50Ω to 75Ω, revealing the influence of charge distribution on impedance. This distributed approach significantly improves processing efficiency.

[0139] Then, a PID controller is used to perform hierarchical control according to the changing trend of dynamic impedance matching. For example, the proportional coefficient is set is 0.5, integral coefficient is 0.1, differential coefficient The value of 0.05 is used. By adjusting the control parameters in real time, the voltage fluctuation is reduced from ±10 V to ±2 V. This hierarchical control can effectively decompose the fluctuation characteristics, preliminarily divide the current distribution uniformity area, and obtain the initial division result.

[0140] Finite element analysis was performed using ANSYS on the initial partitioning results to construct a three-dimensional model of the heat dissipation channel. Based on the three-dimensional model, the initial threshold of the switching state was analyzed. If the initial threshold exceeded the preset range, the genetic algorithm was used to adjust the geometric parameters of the heat dissipation channel to obtain the optimized switching state. For example, the electrical insulation spacing was set to 2mm and the channel width was set to 5mm. Through electro-thermal coupling simulation, the temperature change from 25°C to 80°C under the switching state was analyzed, and the initial threshold was determined to be 90°C. This collaborative logical analysis ensures a balance between electrical and thermal performance. If the initial threshold exceeds the preset range, for example, exceeding 100°C, the channel width can be optimized from 5mm to 7mm. After 50 iterations, the optimized switching state is obtained.

[0141] For the optimized switching state, a gradient descent method is used to calculate the update rule for the switching threshold to obtain the corrected dynamic impedance matching data. For example, a learning rate of 0.01 can be set. Through multiple iterations, the dynamic impedance matching data is corrected from the initial 60Ω to 55Ω, obtaining more accurate corrected data.

[0142] Principal component analysis (PCA) is then used to extract the current distribution adjustment parameters from the corrected data, resulting in an updated configuration for the electrical insulation spacing. For example, using PCA to extract adjustment parameters from the corrected data can identify the primary factors influencing current distribution. For example, the analysis may reveal that insulation spacing contributes 60% to current uniformity, leading to an updated spacing configuration of 2.5mm. This extraction method helps streamline design variables.

[0143] Based on the updated configuration of adjustment parameters and electrical insulation spacing, Lyapunov stability theory is used to determine the stability of the voltage transient response. Specifically, an energy function is constructed to analyze the system's convergence and determine the final switching threshold update rule. For example, when voltage fluctuations are controlled within ±1V, the switching threshold update rule is determined to be adjusted every 10ms. This theoretical analysis ensures the long-term stable operation of the system.

[0144] The embodiment of the present invention provides a coordinated control system for power semiconductor devices in a converter valve, such as Figure 2 As shown, Figure 2 The system block diagram of the power semiconductor device coordinated control system 200 in the converter valve includes:

[0145] An acquisition module 201 is configured to acquire voltage transient response data across the IGCT and the IGBT to obtain a time series data set, and adjust the time series data set to obtain an adjusted time series data set, wherein the voltage transient response data includes an instantaneous voltage value and an instantaneous current value;

[0146] A first adjustment module 202 is configured to perform a fast Fourier transform on the adjusted time series data set to obtain a fundamental wave amplitude, determine the switching state of the converter valve based on the fundamental wave amplitude, obtain switching loss data based on the switching state, and obtain adjusted switching timing data based on the switching loss data;

[0147] The second adjustment module 203 is configured to obtain current distribution uniformity data of the converter valve based on the adjusted time series data set, obtain a final electrical insulation optimization design scheme based on the current distribution uniformity data, obtain a preliminary layout sequence based on the bandwidth data in the final electrical insulation optimization design scheme, utilize switching state variation characteristics and preset material dielectric constant constraints, and adjust the preliminary layout sequence to obtain a final optimization sequence;

[0148] The third adjustment module 204 is used to obtain a final structural stiffness coefficient based on the thermal expansion matching data in the final optimization sequence, obtain a surface charge density suppression scheme based on the final structural stiffness coefficient, and obtain an adjusted switching threshold update rule based on the dynamic impedance matching data in the surface charge density suppression scheme, so that the power system controls the power electronic devices according to the adjusted switching threshold update rule.

[0149] In one embodiment, the acquisition module 201 further includes a time domain analysis unit, a judgment unit, and a correction unit, wherein:

[0150] A time domain analysis unit is used to perform time domain analysis on a time series data set to obtain switching time, and based on the switching time, obtain switching time and delay time;

[0151] a judgment unit, configured to input the switching time and the delay time into a linear weighting module for calculation to obtain a set of distribution feature parameters, and if any distribution feature parameter in the set of distribution feature parameters is greater than a first preset threshold, perform feature extraction on the time series data set to obtain an optimized feature vector;

[0152] The correction unit is used to determine whether the optimized feature vector is abnormal. If it is abnormal, the time series data set is corrected using cubic spline interpolation to obtain an adjusted time series data set.

[0153] The specific implementation of the collaborative control system for power semiconductor devices in a converter valve is basically the same as the specific embodiment of the method for collaborative control system for power semiconductor devices in a converter valve described above, and will not be repeated here.

[0154] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0155] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for cooperatively controlling power semiconductor devices in a converter valve, characterized in that: include: Acquiring voltage transient response data across the IGCT and the IGBT to obtain a time series data set, and adjusting the time series data set to obtain an adjusted time series data set, wherein the voltage transient response data includes an instantaneous voltage value and an instantaneous current value; Performing a fast Fourier transform on the adjusted time series data set to obtain a fundamental wave amplitude, performing a determination based on the fundamental wave amplitude to obtain a switching state of the converter valve, obtaining switching loss data based on the switching state, and obtaining adjusted switching timing data based on the switching loss data; obtaining current distribution uniformity data of the converter valve based on the adjusted time series data set, obtaining a final electrical insulation optimization design scheme based on the current distribution uniformity data, obtaining a preliminary layout sequence based on the bandwidth data in the final electrical insulation optimization design scheme and utilizing switching state variation characteristics and preset material dielectric constant constraints, and adjusting the preliminary layout sequence to obtain a final optimization sequence; According to the thermal expansion matching data in the final optimization sequence, a final structural stiffness coefficient is obtained. Based on the final structural stiffness coefficient, a surface charge density suppression scheme is obtained. According to the dynamic impedance matching data in the surface charge density suppression scheme, an adjusted switching threshold update rule is obtained, so that the power system controls the power electronic device according to the adjusted switching threshold update rule.

2. The method for coordinated control of power semiconductor devices in a converter valve according to claim 1, wherein: The adjusting the time series data set to obtain an adjusted time series data set includes: Performing time domain analysis on the time series data set to obtain switching time and delay time; Inputting the switching time and the delay time into a linear weighting module for calculation to obtain a distribution feature parameter set, and if any distribution feature parameter in the distribution feature parameter set is greater than a first preset threshold, performing feature extraction on the time series data set to obtain an optimized feature vector; It is determined whether the optimized feature vector is abnormal. If abnormal, the time series data set is corrected by using cubic spline interpolation to obtain an adjusted time series data set.

3. The method for coordinated control of power semiconductor devices in a converter valve according to claim 1, wherein: The step of obtaining adjusted switching timing data according to the switching loss data includes: Inputting the switching loss data into a three-dimensional thermoelectric coupling model for calculation to obtain a junction temperature fluctuation range; An iterative solution is performed with the goal of minimizing the difference in the junction temperature fluctuation range to obtain an optimal delay compensation amount; The PWM dead time of the driving circuit is adjusted according to the optimal delay compensation amount to obtain adjusted switching timing data.

4. The method for coordinated control of power semiconductor devices in a converter valve according to claim 1, wherein: The final electrical insulation optimization design scheme is obtained based on the current distribution uniformity data, including: Analyzing the current distribution uniformity data to obtain a distribution feature set; Obtaining a geometric optimization scheme for the heat dissipation channel based on the set of distribution features, optimizing the geometric optimization scheme based on the heat flux density to obtain a preliminary adjustment scheme, optimizing the spacing data in the preliminary adjustment scheme to obtain an optimized spacing distribution, and determining whether the separation of the power electronic devices of the IGCT and the IGBT meets a second preset threshold based on the optimized spacing distribution; if so, obtaining a final spacing adjustment scheme; Based on the final spacing adjustment scheme, a finite element analysis tool is used to verify the thermal stress equilibrium state and obtain a current distribution uniformity verification result; According to the current distribution uniformity verification result, the final electrical insulation optimization design scheme is obtained.

5. The method for coordinated control of power semiconductor devices in a converter valve according to claim 1, wherein: The adjusting of the preliminary layout sequence to obtain a final optimized sequence includes: Obtaining a bandwidth data distribution range based on switching state change characteristics of the IGCT and the IGBT during the switching process; Extracting features from the bandwidth data distribution range to obtain a switching threshold feature, obtaining a dynamic adjustment requirement based on the switching threshold feature, and obtaining a threshold range under the dielectric constant constraint based on the dynamic adjustment requirement and the preset material dielectric constant constraint condition; According to the threshold range under the dielectric constant constraint, the layout structure of the modular connection interface is optimized using the dynamic programming algorithm to obtain a preliminary layout sequence; A final optimized sequence is obtained based on the preliminary layout sequence and the deviation data of the assembly tolerance.

6. The method for coordinated control of power semiconductor devices in a converter valve according to claim 1, wherein: The final structural rigidity coefficient is obtained according to the thermal expansion matching data in the final optimization sequence, including: Performing dimensionality reduction on the thermal expansion matching data in the final optimization sequence to obtain an optimized matching feature vector; Acquiring charge density data according to the optimized matching degree feature vector, filtering the charge density data to obtain a smoothed charge density curve, and performing frequency domain feature extraction on the smoothed charge density curve to obtain a main frequency distribution range; If the main frequency distribution range is within the preset frequency interval, bandwidth limiting is performed on the control signal to obtain a bandwidth-limited control signal; A structural rigidity coefficient is obtained based on the bandwidth-limited control signal, and the structural rigidity coefficient is smoothed to obtain a rigidity coefficient curve. If the slope of the rigidity coefficient curve is greater than a preset slope threshold, the optimized matching degree feature vector is adjusted until the slope of the rigidity coefficient curve is less than the preset slope threshold, thereby obtaining a final structural rigidity coefficient.

7. The method for coordinated control of power semiconductor devices in a converter valve according to claim 1, wherein: The step of obtaining an adjusted switching threshold update rule based on the dynamic impedance matching data in the surface charge density suppression scheme includes: Obtaining a final switching threshold update rule based on the dynamic impedance matching data in the surface charge density suppression scheme; Obtain switching state data and switching time data to obtain dynamic characteristics of delay changes, fit the dynamic characteristics of delay changes to obtain smoothed correction fluctuation results, obtain an adjustment range based on the smoothed correction fluctuation results, adjust the final switching threshold update rule based on the adjustment range, and obtain an adjusted switching threshold update rule.

8. The method for coordinated control of power semiconductor devices in a converter valve according to claim 7, wherein: The method of obtaining a final switching threshold update rule based on the dynamic impedance matching data in the surface charge density suppression scheme includes: Performing distributed calculations on the surface charge density data in the surface charge density suppression scheme to obtain a change trend of dynamic impedance matching; Based on the changing trend of dynamic impedance matching, current distribution uniformity is divided into regions to obtain an initial division result. Finite element analysis is performed on the initial division result to obtain an initial threshold value of the switching state. If the initial threshold value of the switching state is not within a preset range, the geometric parameters of the heat dissipation channel are adjusted to obtain an optimized switching state. According to the optimized switching state, correction data of dynamic impedance matching is obtained, current distribution adjustment parameters are extracted from the correction data, and an updated configuration scheme of the electrical insulation distance is obtained. According to the updated configuration scheme of the electrical insulation distance, a final switching threshold update rule is determined.

9. A coordinated control system for power semiconductor devices in a converter valve, characterized in that: include: an acquisition module, configured to acquire voltage transient response data across the IGCT and the IGBT to obtain a time series data set, and adjust the time series data set to obtain an adjusted time series data set; wherein the voltage transient response data includes an instantaneous voltage value and an instantaneous current value; a first adjustment module, configured to perform a fast Fourier transform on the adjusted time series data set to obtain a fundamental wave amplitude, determine based on the fundamental wave amplitude to obtain a switching state of the converter valve, obtain switching loss data based on the switching state, and obtain adjusted switching timing data based on the switching loss data; a second adjustment module, configured to obtain current distribution uniformity data of the converter valve based on the adjusted time series data set, obtain a final electrical insulation optimization design scheme based on the current distribution uniformity data, obtain a preliminary layout sequence based on the bandwidth data in the final electrical insulation optimization design scheme, utilize switching state variation characteristics and preset material dielectric constant constraints, and adjust the preliminary layout sequence to obtain a final optimization sequence; A third adjustment module is configured to obtain a final structural stiffness coefficient based on the thermal expansion matching data in the final optimization sequence, obtain a surface charge density suppression scheme based on the final structural stiffness coefficient, and obtain an adjusted switching threshold update rule based on the dynamic impedance matching data in the surface charge density suppression scheme, so that the power system controls the power electronic device according to the adjusted switching threshold update rule.

10. The coordinated control system for power semiconductor devices in a converter valve according to claim 9, characterized in that: The acquisition module also includes a time domain analysis unit, a judgment unit and a correction unit, wherein: The time domain analysis unit is configured to perform time domain analysis on the time series data set to obtain switching time, and obtain switching time and delay time based on the switching time; The judgment unit is configured to input the switching time and the delay time into a linear weighting module for calculation to obtain a distribution feature parameter set, and if any distribution feature parameter in the distribution feature parameter set is greater than a first preset threshold, perform feature extraction on the time series data set to obtain an optimized feature vector; The correction unit is used to determine whether the optimized feature vector is abnormal. If abnormal, the time series data set is corrected by using cubic spline interpolation to obtain an adjusted time series data set.

Images