Overvoltage control method and device for IGBT (Insulated Gate Bipolar Translator) module

Abstract

The invention relates to the technical field of overvoltage control, and particularly discloses an overvoltage control method and device for an IGBT module, and the method comprises the steps: carrying out the time sequence correlation analysis of collector current time sequence data through employing a deep learning algorithm after receiving a turn-off instruction, extracting the short-time fluctuation characteristics of collector current, and obtaining the short-time fluctuation characteristics of the collector current; and the initial gate resistance value is combined to dynamically predict the collector-emitter voltage peak amplitude possibly generated in the turn-off process. And based on comparison between the predicted voltage spike amplitude and a preset threshold value, an initial gate driving strategy is generated to guide the turn-off operation. In the turn-off execution stage, collector-emitter voltage data flow is monitored in real time and dynamically compared with a safety threshold value, and therefore closed-loop optimization adjustment is conducted on a driving strategy. According to the method, prospective evaluation and self-adaptive control of the overvoltage risk are realized through intelligent prediction of the turn-off transient voltage spike, the switching loss can be optimized while the overvoltage is suppressed, and the limitation of traditional fixed parameter control is broken through.

Description

Technical Field

[0001] The present application relates to the field of overvoltage control technology, and more specifically, to an overvoltage control method and device for an IGBT module. Background Art

[0002] Insulated-gate bipolar transistor (IGBT) modules, as key power semiconductor devices, are widely used in power electronic converters, motor drives, and grid-connected inverters for renewable energy due to their advantages such as high input impedance, low on-state voltage drop, and high switching speed. However, during the IGBT module's turn-off process, the rapidly changing collector current (di / dt) inevitably introduces stray inductance in the power circuit (e.g., inductance introduced by the DC bus layout and internal module connections). This induction voltage is superimposed on the DC bus voltage and applied between the IGBT's collector and emitter, resulting in the so-called turn-off voltage spike. If the magnitude of this voltage spike exceeds the rated breakdown voltage of the IGBT device, it can accelerate device aging and reduce reliability at best. In worse cases, it can cause permanent damage to the device and even lead to system failure, threatening equipment and personnel safety. Therefore, effectively controlling the IGBT module's turn-off overvoltage is critical to ensuring the safe and reliable operation of power electronic systems.

[0003] In power electronics systems, existing IGBT module overvoltage control solutions primarily rely on passive snubber circuit designs or fixed gate resistor adjustment. Passive snubber circuits (such as RCD snubber networks) utilize a clamping loop consisting of parallel capacitors, resistors, and diodes to absorb the energy released by parasitic inductance during IGBT turn-off, thereby suppressing voltage spikes. However, the capacitance of the snubber capacitor and the resistance of the resistor in these solutions must be pre-designed based on specific circuit parameters. When the system operating frequency, load range, or circuit layout change, fixed snubber parameters may not effectively cover the entire operating range, potentially leading to insufficient or excessive absorption. The additional components also increase system size and cost. Fixed gate resistor control strategies, on the other hand, adjust the IGBT turn-off speed by presetting a fixed resistance value, attempting to suppress voltage spikes by slowing the current drop rate. However, this solution treats the turn-off process as a static scenario, ignoring the dynamic fluctuations of load current in actual operation. When the system faces sudden load increases, unbalanced loads, or high-frequency start-stop cycles, fixed gate resistor values ​​cannot effectively match the energy released by parasitic inductance due to current fluctuations. For example, using a fixed, high-value gate resistor under light-load conditions can result in slow turn-off, causing the freewheeling diode to generate severe reverse recovery current due to prolonged reverse bias, further increasing device heating. Under heavy-load conditions, using a fixed, low-value gate resistor can shorten switching times, but the steep current drop rate can cause higher voltage spikes induced by the line's parasitic inductance, exceeding the IGBT's withstand voltage threshold. In other words, existing solutions generally lack the ability to perceive and predict the dynamic current characteristics of the turn-off process in real time and proactively optimize turn-off behavior accordingly.

[0004] Therefore, an optimized overvoltage control system and method for an IGBT module is desired. Summary of the Invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides an overvoltage control method and device for an IGBT module, which, after receiving a shutdown instruction, uses a deep learning algorithm to perform a time-series correlation analysis on the collector current timing data including the current moment, extracts the short-term fluctuation characteristics of the collector current, and dynamically predicts the collector-emitter voltage spike amplitude that may be generated during the shutdown process in combination with the initial gate resistance value. Furthermore, based on the comparison between the predicted voltage spike amplitude and the preset threshold, an initial gate drive strategy is generated to guide the shutdown operation. At the same time, in the shutdown execution phase, the collector-emitter voltage data stream is monitored in real time and dynamically compared with the safety threshold to perform closed-loop optimization and adjustment of the drive strategy. This method realizes forward-looking assessment and adaptive control of overvoltage risks through intelligent prediction of the shutdown transient voltage spike, can optimize switching losses while suppressing overvoltage, and break through the limitations of traditional fixed parameter control.

[0006] According to one aspect of the present application, a method for controlling an overvoltage of an IGBT module is provided, comprising:

[0007] Real-time acquisition of collector current and collector-emitter voltage of IGBT modules;

[0008] In response to receiving an IGBT shutdown instruction from a main controller, obtaining collector current values ​​at a plurality of predetermined time points including a current time point;

[0009] Based on the collector current values ​​at the plurality of predetermined time points, predicting a collector-emitter voltage spike amplitude that may be generated during the current shutdown, and generating an initial gate drive strategy based on the predicted collector-emitter voltage spike amplitude;

[0010] Controlling the gate driver to turn off the IGBT module according to the initial gate drive strategy, and monitoring the collector-emitter voltage value in real time to obtain a collector-emitter voltage data stream;

[0011] The collector-emitter voltage data stream is compared with a collector-emitter voltage safety threshold, and whether to adjust the gate drive strategy is determined based on the comparison result.

[0012] According to another aspect of the present application, an overvoltage control device for an IGBT module is provided, comprising:

[0013] Current and voltage data acquisition module, used to collect the collector current value and collector-emitter voltage value of the IGBT module in real time;

[0014] a collector current value acquisition module, configured to acquire collector current values ​​at a plurality of predetermined time points including the current time point in response to receiving an IGBT shutdown instruction from the main controller;

[0015] a spike prediction and strategy generation module, configured to predict the collector-emitter voltage spike amplitude that may be generated by the current shutdown based on the collector current values ​​at the plurality of predetermined time points, and to generate an initial gate drive strategy based on the predicted collector-emitter voltage spike amplitude;

[0016] a gate driver control and monitoring module, configured to control the gate driver to turn off the IGBT module according to the initial gate drive strategy, and to monitor the collector-emitter voltage value in real time to obtain a collector-emitter voltage data stream;

[0017] The safety threshold comparison and strategy adjustment module is used to compare the collector-emitter voltage data stream with the collector-emitter voltage safety threshold, and determine whether to adjust the gate drive strategy based on the comparison result.

[0018] Compared with the prior art, the overvoltage control method and device for the IGBT module provided in this application, after receiving the shutdown command, uses a deep learning algorithm to perform time-series correlation analysis on the collector current time series data including the current moment, extracts the short-term fluctuation characteristics of the collector current, and dynamically predicts the collector-emitter voltage spike amplitude that may be generated during the shutdown process in combination with the initial gate resistance value. Furthermore, based on the comparison between the predicted voltage spike amplitude and the preset threshold, an initial gate drive strategy is generated to guide the shutdown operation. At the same time, during the shutdown execution phase, the collector-emitter voltage data stream is monitored in real time and dynamically compared with the safety threshold to perform closed-loop optimization and adjustment of the drive strategy. This method realizes forward-looking assessment and adaptive control of overvoltage risks through intelligent prediction of the shutdown transient voltage spike, can optimize switching losses while suppressing overvoltage, and breaks through the limitations of traditional fixed parameter control. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0020] Figure 1 Flowchart of an overvoltage control method for an IGBT module according to an embodiment of the present application.

[0021] Figure 2 4 is a flowchart of sub-step S3 of the overvoltage control method of the IGBT module according to an embodiment of the present application.

[0022] Figure 3 4 is a data flow diagram of sub-step S3 of the overvoltage control method of the IGBT module according to an embodiment of the present application.

[0023] Figure 4 4 is a flowchart of sub-step S31 of the overvoltage control method of the IGBT module according to an embodiment of the present application.

[0024] Figure 5 4 is a flowchart of sub-step S313 of the overvoltage control method of the IGBT module according to an embodiment of the present application.

[0025] Figure 6 4 is a flowchart of sub-step S33 of the overvoltage control method of the IGBT module according to an embodiment of the present application.

[0026] Figure 7 4 is a block diagram of an overvoltage control device for an IGBT module according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] As used in this application and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0028] Although the present application makes various references to certain modules in the system according to embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are illustrative only, and different aspects of the system and method can use different modules.

[0029] Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the various steps may be processed in reverse order or simultaneously, as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0030] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0031] It is worth noting that in this application, all actions to obtain data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0032] In response to the technical problems described in the above background technology, this application proposes an overvoltage control method for an IGBT module. After receiving a shutdown command, the method uses a deep learning algorithm to perform a time-series correlation analysis on the collector current time series data including the current moment, extracts the short-term fluctuation characteristics of the collector current, and dynamically predicts the collector-emitter voltage spike amplitude that may be generated during the shutdown process in combination with the initial gate resistance value. Furthermore, based on the comparison between the predicted voltage spike amplitude and the preset threshold, an initial gate drive strategy is generated to guide the shutdown operation. At the same time, during the shutdown execution phase, the collector-emitter voltage data stream is monitored in real time and dynamically compared with the safety threshold to perform closed-loop optimization and adjustment of the drive strategy. This method realizes forward-looking assessment and adaptive control of overvoltage risks through intelligent prediction of the shutdown transient voltage spike. It can optimize switching losses while suppressing overvoltage, breaking through the limitations of traditional fixed parameter control.

[0033] Figure 1 FIG. 1 is a flow chart of an overvoltage control method for an IGBT module according to an embodiment of the present application. Figure 1 As shown, the overvoltage control method of the IGBT module includes the following steps: S1, real-time acquisition of the collector current value and collector-emitter voltage value of the IGBT module; S2, in response to receiving an IGBT shutdown instruction from a main controller, obtaining the collector current values ​​at multiple predetermined time points including the current moment; S3, based on the collector current values ​​at the multiple predetermined time points, predicting the collector-emitter voltage spike amplitude that may be generated by this shutdown, and generating an initial gate drive strategy based on the predicted collector-emitter voltage spike amplitude; S4, controlling the gate driver to shut down the IGBT module according to the initial gate drive strategy, and monitoring the collector-emitter voltage value in real time to obtain a collector-emitter voltage data stream; S5, comparing the collector-emitter voltage data stream with the collector-emitter voltage safety threshold, and determining whether to adjust the gate drive strategy based on the comparison result.

[0034] In the above-mentioned overvoltage control method of the IGBT module, the step S1 collects the collector current value and collector-emitter voltage value of the IGBT module in real time. It should be understood that during the shutdown process of the IGBT module, the interaction between the loop stray inductance and the rapidly changing current (di / dt) can easily produce destructive voltage spikes, and the amplitude of the spike is closely related to the real-time working state of the IGBT. Therefore, the present application collects the collector current value (Ic) and collector-emitter voltage value (Vce) of the IGBT module in real time to achieve accurate grasp of the working state of the device, so that the overvoltage control can be supported by analyzing the shutdown transient characteristics of the IGBT module. Specifically, a high-speed current sensor (such as a Hall sensor or a sampling resistor) can be connected in series in the power circuit and a high-bandwidth voltage sensor (such as a voltage divider resistor network or an optocoupler isolation voltage sensor) can be connected in parallel at both ends of the collector-emitter, and the sensor signal can be converted into a digital quantity through a high-speed analog-to-digital converter (ADC) for subsequent processing. In this way, the instantaneous electrical state of the IGBT before the shutdown command is issued can be understood in real time, laying the foundation for subsequent accurate prediction and control.

[0035] During implementation, high-speed current sensors are essential for collector current measurement. Hall effect sensors are a preferred option due to their high sensitivity and wide dynamic range. These sensors utilize the Hall effect to detect current without requiring direct contact with the circuit being measured, thus preventing any interference with the original circuit and providing good isolation. A sampling resistor is another viable option. By connecting a small resistor in series with the circuit, the current flowing through it is calculated using Ohm's law. This method is simple and straightforward, but under high current conditions, the sampling resistor itself generates heat, potentially affecting measurement accuracy. Therefore, comprehensive thermal design considerations are necessary when implementing this method.

[0036] At the same time, in order to obtain the collector-emitter voltage value, a high-bandwidth voltage sensor can be connected in parallel between the collector and emitter of the IGBT. A voltage divider resistor network is a common method. By selecting the appropriate resistance ratio, a high voltage signal can be converted into a small voltage signal suitable for processing by an analog-to-digital converter. However, this method requires the resistors to have high precision and stability to ensure the accuracy of the measurement results. Optocoupler isolated voltage sensors are another effective solution. They can not only provide high-precision voltage measurement, but also have electrical isolation functions, which help improve the safety and anti-interference capabilities of the system. The optocoupler device contains a light-emitting diode and a photodetector. When the input voltage changes, the light intensity emitted by the light-emitting diode changes accordingly. The photodetector receives the light signal and converts it into an electrical signal output, thereby completing the voltage detection.

[0037] During actual deployment, electromagnetic compatibility (EMC) must also be fully considered. Because power electronics systems typically operate in high-frequency and high-voltage environments, they are susceptible to external electromagnetic interference. To ensure the reliability of sensors and ADCs, appropriate protective measures must be implemented, such as proper wiring layout, additional shielding layers, and optimized grounding. For sensitive components like ADCs, effective power supply filtering and decoupling are crucial to effectively reduce noise interference and improve signal quality.

[0038] Furthermore, to ensure efficient and accurate real-time data acquisition, a comprehensive calibration mechanism is also necessary. Regular sensor calibration can correct errors caused by long-term use or environmental changes, ensuring data consistency and reliability. For example, Hall effect sensors or sampling resistors can be calibrated using a standard source to adjust their gain and offset. For ADCs, known standard voltage signals can be used for testing to verify that the conversion results meet expectations.

[0039] In the above-mentioned overvoltage control method for an IGBT module, step S2, in response to receiving an IGBT shutdown command from a main controller, obtains collector current values ​​at multiple predetermined time points, including the current moment. It should be understood that since the magnitude of the shutdown voltage spike is not only related to the current value at the moment of shutdown, but also closely related to the dynamic change trend of the current before shutdown (i.e., short-term fluctuation characteristics), a simple single-point current value is insufficient to fully reflect the impending di / dt situation. Therefore, in order to obtain sufficient information for predicting the shutdown voltage spike, the present application further retrieves collector current values ​​at multiple predetermined time points, including the current moment, from real-time monitoring data upon receiving an IGBT shutdown command from the main controller, to construct a short-term series reflecting recent current dynamics. Specifically, when the main controller issues a shutdown signal, the control unit immediately extracts the collector current values ​​of the most recent N sampling points (N is a predetermined number, and the time point corresponds to the sampling moment) from the current data stream collected in real time. In this way, by using the current history data in a short period of time before shutdown to capture the trend and characteristics of current changes, it can better predict the size of di / dt during the shutdown process than the current value at a single moment, and thus more accurately predict the voltage spike.

[0040] In the above-mentioned overvoltage control method of the IGBT module, the step S3 predicts the collector-emitter voltage spike amplitude that may be generated by this shutdown based on the collector current values ​​at the multiple predetermined time points, and generates an initial gate drive strategy based on the predicted collector-emitter voltage spike amplitude. Specifically, since the fixed strategy in the prior art cannot adapt to changing working conditions, and the passive response strategy has a delay, it is difficult to achieve optimal control. In this regard, the present application further predicts the collector-emitter voltage spike amplitude that may be generated by this shutdown based on the collector current values ​​at the multiple predetermined time points and uses this to actively optimize the shutdown behavior, thereby estimating the potential overvoltage risk before the shutdown action is executed and formulating targeted countermeasures. Among them, Figure 2 4 is a flowchart of sub-step S3 of the overvoltage control method of the IGBT module according to an embodiment of the present application. Figure 3 FIG. 1 is a data flow diagram of sub-step S3 of the overvoltage control method of the IGBT module according to an embodiment of the present application. Figure 2 and Figure 3As shown, the step S3 includes the following steps: S31, extracting the collector current short-time timing fluctuation characteristics from the collector current values ​​at the multiple predetermined time points; S32, obtaining the initial turn-off gate resistance value; S33, calculating the predicted collector-emitter voltage rise based on the initial turn-off gate resistance value and the collector current short-time timing fluctuation characteristics; S34, adding the collector-emitter voltage value at the current moment to the predicted collector-emitter voltage rise to obtain the predicted collector-emitter voltage peak amplitude; S35, generating the initial gate drive strategy based on the comparison between the predicted collector-emitter voltage peak amplitude and the collector-emitter voltage preset threshold.

[0041] Specifically, the step S31 extracts the collector current short-term timing fluctuation characteristics from the collector current values ​​at the multiple predetermined time points. Specifically, the collector-emitter voltage spike amplitude caused by the IGBT shutdown is not only related to the absolute value of the current at the shutdown moment, but also depends on the dynamic change rate and fluctuation shape of the current before shutdown at a deeper level. Therefore, in order to more accurately capture the key dynamic factors that lead to voltage spikes, this application introduces a deep learning algorithm to extract the timing fluctuation characteristics of the collector current values ​​at the multiple predetermined time points to achieve in-depth characterization of the current dynamic behavior. Among them, Figure 4 FIG. 1 is a flow chart of sub-step S31 of the overvoltage control method of the IGBT module according to an embodiment of the present application. Figure 4 As shown, the step S31 includes the steps of: S311, arranging the collector current values ​​of the multiple predetermined time points into a collector current short-time input vector according to the time dimension; S312, performing time series multi-scale encoding on the collector current short-time input vector to obtain a collector current short-time time series fluctuation feature multi-scale encoding vector; S313, performing time series fine-grained expression optimization on the collector current short-time time series fluctuation feature multi-scale encoding vector to obtain a collector current short-time time series fluctuation feature optimized encoding vector as the collector current short-time time series fluctuation feature.

[0042] More specifically, the step S311 arranges the collector current values ​​at the multiple predetermined time points into collector current short-time input vectors according to the time dimension. It should be understood that since the original collected collector current data is a discrete time point sample, in order to effectively learn the timing dependency and fluctuation pattern in the collector current data, these discrete samples need to be ordered first. Therefore, the present application arranges the collector current values ​​at the multiple predetermined time points into collector current short-time input vectors according to the time dimension to retain the time sequence relationship between the collector current data, revealing the continuity and trend of the current over time, thereby achieving a normalized representation of the data and providing a vectorized data structure containing timing information for subsequent timing feature analysis.

[0043] More specifically, in a specific example of the present application, the step S312 includes: inputting the collector current short-time input vector into a time-series multi-scale encoder based on a CNN-BiGRU hybrid architecture to obtain a multi-scale encoding vector of the collector current short-time time-series fluctuation feature. It should be understood that since the short-time behavior of the collector current before shutdown contains complex local patterns (such as rapid fluctuations, oscillations) and time-series dependencies (such as changing trends), in order to effectively extract multi-level, multi-scale current time-series dynamic features from the collector current short-time input vector, the present application adopts a CNN-BiGRU hybrid architecture to perform time-series multi-scale encoding on the collector current short-time input vector to capture the fine local features and long-distance time-series dependencies in the current data. Specifically, the collector current short-term input vector is first fed into the CNN layer, which uses a one-dimensional convolution kernel to slide in the time dimension to identify local change patterns in the current waveform, such as rapid fluctuations and spike shapes; subsequently, the sequence of local time domain features of the collector current extracted by CNN is passed to the bidirectional gated recurrent unit (BiGRU). BiGRU processes sequence information in both the forward and reverse time dimensions, and can effectively learn the long-term trend of current changes, periodic fluctuations, and the correlation between the previous and next time points. It finally outputs hidden state features that combine local fine features and global temporal dependencies, namely, the multi-scale encoding vector of collector current short-term temporal fluctuation features. Based on this, the present application achieves a comprehensive and in-depth understanding of the dynamic behavior of collector current timing by comprehensively utilizing the powerful extraction capability of CNN (convolutional neural network) for local patterns and the bidirectional capture capability of BiGRU (bidirectional gated recurrent unit) for long-distance temporal dependencies.

[0044] More specifically, the step S313 performs timing fine-grained expression optimization on the multi-scale coding vector of the collector current short-time temporal fluctuation characteristics to obtain the optimized coding vector of the collector current short-time temporal fluctuation characteristics as the collector current short-time temporal fluctuation characteristics. Specifically, considering that the multi-scale coding vector of the collector current short-time temporal fluctuation characteristics may have feature redundancy or loose temporal correlation, its internal structure fails to optimally highlight the key information directly related to the voltage spike prediction. Therefore, in order to further improve the effectiveness and pertinence of the features and make it more accurately serve the voltage prediction task, the present application further performs timing fine-grained expression optimization on the multi-scale coding vector of the collector current short-time temporal fluctuation characteristics, and aims to amplify the fine-grained temporal information with high discrimination for predicting voltage spikes and suppress irrelevant noise or redundant components by adjusting and optimizing the temporal relationship or structure between the internal elements of the multi-scale coding vector of the collector current short-time temporal fluctuation characteristics. Among them, Figure 5FIG. 1 is a flow chart of sub-step S313 of the overvoltage control method of the IGBT module according to an embodiment of the present application. Figure 5 As shown, the step S313 includes the steps of: S3131, performing time-series local fluctuation modal reconfiguration on the multi-scale coding vector of the collector current short-time temporal fluctuation characteristics to obtain a set of collector current short-time temporal fluctuation local modal extraction feature coding vectors; S3132, performing feature reshaping gain calculation based on the energy density of the fluctuation component on each collector current short-time temporal fluctuation local modal extraction feature coding vector in the set of collector current short-time temporal fluctuation local modal extraction feature coding vectors to obtain a set of collector current short-time temporal fluctuation local modal feature reshaping gain operators; S3133, based on the set of collector current short-time temporal fluctuation local modal feature reshaping gain operators, performing significance reshaping on the multi-scale coding vector of the collector current short-time temporal fluctuation characteristics to obtain the collector current short-time temporal fluctuation feature optimization coding vector.

[0045] In a specific example of the present application, step S3131 includes: first, performing time series local fluctuation mode reconstruction based on one-dimensional convolution coding on the multi-scale coding vector of the collector current short-term time series fluctuation feature to obtain a set of collector current short-term time series fluctuation local modal feature coding vectors, which is expressed as follows:

[0046] Conv l×1 (X)={x1,x2,...,x i ,...,x n}

[0047] Among them, X represents the multi-scale encoding vector of the collector current short-term timing fluctuation characteristics, Conv l×1 (·) represents the one-dimensional convolutional coding operation, l is the scale of the one-dimensional convolution kernel, x1, x2, x i and x n They respectively represent the 1st, 2nd, i-th and n-th collector current short-time timing fluctuation local modal feature coding vectors in the set of collector current short-time timing fluctuation local modal feature coding vectors, and n is the number of vectors in the set of collector current short-time timing fluctuation local modal feature coding vectors.

[0048] That is, by reconstructing the time-series local fluctuation mode of the multi-scale coding vector of the collector current short-time timing fluctuation characteristics based on one-dimensional convolution coding, the local characteristic patterns at different positions in the multi-scale coding vector of the collector current short-time timing fluctuation characteristics are effectively mapped to a new feature space, so as to capture the local characteristic patterns in the current fluctuation from different angles and obtain a set of local modal characteristic coding vectors of the collector current short-time timing fluctuation, thereby realizing the diverse expression of structural information and refining the local structure modeling.

[0049] Then, information extraction is performed on each collector current short-term time series fluctuation local modal feature coding vector in the set of the collector current short-term time series fluctuation local modal feature coding vectors to obtain the set of the collector current short-term time series fluctuation local modal extraction feature coding vectors, which is expressed as follows:

[0050]

[0051] Among them, x i The ith collector current short-term time series fluctuation local modal feature encoding vector in the set of collector current short-term time series fluctuation local modal feature encoding vectors is represented by ||·|| 2 Indicates the calculation of the square of the vector norm, v i The ith collector current short-term timing fluctuation local modal extraction feature encoding vector in the set of collector current short-term timing fluctuation local modal extraction feature encoding vectors, that is, x i The corresponding collector current short-term timing fluctuation local mode extraction feature encoding vector.

[0052] That is, through information extraction, the components closely related to the key dynamic information in the local modal feature encoding vector of the collector current short-term timing fluctuation are selectively retained, redundant dimensions and mixed noise and irrelevant information are suppressed or eliminated, and the efficiency and effectiveness of feature representation are improved to form a simpler and more generalizable set of local modal extraction feature encoding vectors of the collector current short-term timing fluctuation, providing high-quality feature input for subsequent feature reshaping gain calculation, ensuring that subsequent processing steps can more accurately focus on the current fluctuation characteristics that play a key role in overvoltage control.

[0053] In a specific example of the present application, step S3132 includes: first, calculating the effective fluctuation component statistical coefficient of each collector current short-term time series fluctuation local modal extraction feature coding vector in the set of the collector current short-term time series fluctuation local modal extraction feature coding vector, which is expressed as follows:

[0054]

[0055] in, Indicates the eigenvalue of the jth position in the local modal extraction feature encoding vector of the collector current short-term timing fluctuation, count i (·) represents the counting function, ε represents the threshold parameter, |·| represents the absolute value, and They represent the eigenvalues ​​of the j-1th and j+1th positions in the local modal extraction feature encoding vector of the collector current short-term timing fluctuation, respectively. i Indicates vi The corresponding statistical coefficient of the effective volatility component.

[0056] That is, by calculating the statistical coefficient of the effective fluctuation component, the number or proportion of the effective fluctuation components that play a key role in the local modal extraction feature coding vector of each collector current short-term timing fluctuation is quantitatively evaluated, and the statistical coefficient of the effective fluctuation component of the local modal extraction feature coding vector of each collector current short-term timing fluctuation is obtained, which provides a clear quantitative indicator for the subsequent further processing of the features, helps to more accurately adjust the feature representation, so that it can better reflect the intrinsic relationship between current fluctuation and overvoltage, thereby improving the accuracy and effectiveness of the entire overvoltage control strategy.

[0057] Then, based on the statistical coefficient of the effective fluctuation component of each collector current short-term time series fluctuation local modal extraction feature coding vector, the feature reconstruction gain operator of each collector current short-term time series fluctuation local modal extraction feature coding vector is calculated to obtain a set of collector current short-term time series fluctuation local modal feature reconstruction gain operators, which is expressed as follows:

[0058]

[0059] Among them, λ i Represents x i The corresponding feature reshaping gain calculation intermediate parameters, arctan (·) represents the inverse tangent function, e i Represents x i The corresponding collector current short-term timing fluctuation local modal characteristics reshape the gain operator.

[0060] That is, by calculating the feature reshaping gain operator associated with the statistical coefficient of the effective fluctuation component, the differentiated enhancement or suppression of different fluctuation modes in the local modal extraction feature coding vector of the collector current short-term timing fluctuation is achieved, and an adaptive feature control mechanism that adapts to the dynamic characteristics of current fluctuation is constructed. The set of generated collector current short-term timing fluctuation local modal feature reshaping gain operators can dynamically characterize the relative importance of the effective fluctuation components in each collector current short-term timing fluctuation local modal extraction feature coding vector, and through the refined control of the feature space, the characteristic modes such as high-frequency fluctuations and sudden changes in edges that play a key role in voltage spike prediction are enhanced, and the influence of noise and secondary fluctuation components is weakened, forming a feature enhancement representation that is highly correlated with the IGBT turn-off overvoltage risk, providing a structured control basis for the subsequent feature aggregation based on gain weights, thereby improving the adaptability of the overvoltage prediction model to complex working conditions.

[0061] In particular, in a preferred example of the present application, step S3133 includes: first, performing a time-invariant gain convergence stabilization process on the feature reshaping gain operator of each collector current short-term timing fluctuation local modal extraction feature encoding vector to obtain a set of optimized collector current short-term timing fluctuation local modal feature reshaping gain operators, which is expressed as:

[0062]

[0063] Among them, e (·) represents an exponential function with a natural constant as the base, K i represents the global analytical flatness, T i represents the standard configuration field strength, e′ i Indicates e i The corresponding optimization of the collector current short-term timing fluctuation local modal characteristics reshapes the gain operator.

[0064] Here, when calculating the intermediate parameters of the feature reshaping gain corresponding to the local modal extraction feature coding vector of the collector current short-term temporal fluctuation, the present application regards each collector current short-term temporal fluctuation local modal extraction feature coding vector as a generating vector in the set space composed of effective components, so that the polar angle representation satisfies the directional symmetry requirements, thereby ensuring the translation invariance characteristics of the effective components in the set space. That is, the set space is decomposed based on the metric of the holomorphic structure, and the effectiveness enhancement processing of the local modal extraction feature coding vector of the collector current short-term temporal fluctuation is represented as the global analytical flatness representation under the holomorphic structure, and then the single-mode coupling is constructed as a standard configuration field representation. Through the effectiveness enhancement representation of the local modal feature reshaping gain operator of the collector current short-term temporal fluctuation, the characteristics of its single mode as the highest weighted state of flatness excitation in the spatial holomorphic structure are reflected. Specifically, the effectiveness representation of the local modal feature reshaping gain operator of the collector current short-term temporal fluctuation is used to reflect its dominant mode in the feature space, and then the time-invariant gain convergence stabilization processing is performed on it. During this process, it is necessary to ensure that the symmetry of the phase direction is maintained under the condition of stable feature space, so as to avoid the loss of characteristic phase information in the subsequent coupling enhancement process based on the reshaping of local modal feature significance, and improve the representation ability of the local modal extraction feature coding vector of the collector current short-term timing fluctuations.

[0065] Then, the set of the optimization collector current short-term temporal fluctuation local modal feature reshaping gain operators is normalized to obtain a set of collector current short-term temporal fluctuation local modal feature reshaping gain weights, which is expressed as follows:

[0066]

[0067] Among them, exp(·) represents the exponential function operation with e as the base, a i represents e′ i The corresponding collector current short-term timing fluctuation local modal characteristics reshape the gain weight.

[0068] That is, through normalization processing, the operator for optimizing the local modal feature reshaping gain of the collector current short-time temporal fluctuation is converted into a set of relatively comparable collector current short-time temporal fluctuation local modal feature reshaping gain weights, so that the enhancement amplitude of each collector current short-time temporal fluctuation local modal feature encoding vector can be quantified in a unified metric space, ensuring that reasonable weights can be allocated according to the energy density differences of the effective fluctuation components during weighted aggregation, avoiding the weakening of key features or excessive amplification of secondary features due to scale deviation.

[0069] Finally, based on the set of gain weights reshaped by the local modal characteristics of the collector current short-term timing fluctuation, the set of encoding vectors of the local modal characteristics of the collector current short-term timing fluctuation is weightedly aggregated to obtain the optimized encoding vector of the collector current short-term timing fluctuation characteristics, which can be expressed as follows:

[0070]

[0071] Among them, v enhanced Represents the optimized encoding vector of the collector current short-term timing fluctuation characteristics.

[0072] That is, the gain weights are reshaped based on the local modal characteristics of each collector current short-time timing fluctuation, and the encoding vectors of the local modal characteristics of the collector current short-time timing fluctuation are weightedly fused. While retaining the spatial diversity of the collector current short-time timing fluctuation characteristics, the key fluctuation modes directly related to the shutdown overvoltage are enhanced to form an optimized encoding vector of the collector current short-time timing fluctuation characteristics containing hierarchical importance information, which provides more accurate input features for the subsequent voltage spike prediction model and enhances the ability to assess overvoltage risks under dynamic working conditions.

[0073] Specifically, the step S32 obtains the initial turn-off gate resistance value. It should be understood that the generation of voltage spikes is the result of the combined effect of device characteristics, external circuits (including stray inductance) and control strategies (determined by gate resistance). Among them, the gate drive resistance is the core parameter for controlling the turn-off speed of the IGBT. It controls the turn-off time by adjusting the gate charge discharge speed. Different resistance values ​​will lead to different gate charge extraction rates, which in turn significantly affect the collector current drop slope (di / dt) and the final voltage spike. Therefore, in order to enable the voltage spike prediction to be associated with a specific control action, this application obtains the initial turn-off gate resistance value to incorporate this key control parameter into the prediction model as an important basis for optimizing the gate drive strategy.

[0074] Specifically, the step S33 is to calculate the predicted collector-emitter voltage rise based on the initial turn-off gate resistance and the collector current short-term timing fluctuation characteristics. Figure 6 FIG. 1 is a flow chart of sub-step S33 of the overvoltage control method of the IGBT module according to an embodiment of the present application. Figure 6 As shown, the step S33 includes the steps of: S331, decoding the short-time timing fluctuation characteristics of the collector current to obtain a real-time collector current correction factor; S332, correcting the collector current value at the current moment based on the real-time collector current correction factor to obtain a corrected real-time collector current value; S333, inputting the corrected real-time collector current value and the initial turn-off gate resistance value into a circuit characteristic model to obtain a predicted collector-emitter voltage rise.

[0075] More specifically, step S331 decodes the collector current short-term fluctuation characteristics to obtain a real-time collector current correction factor. It should be understood that the collector current short-term fluctuation characteristics reflect the dynamic behavior pattern of the current before shutdown, which has a certain predictive effect on the actual state of the collector current at the moment of shutdown. For example, if the current is in a rapid rising phase before shutdown, even if the instantaneous value of the collector current at the current moment is not high, its inherent "momentum" may cause a large di / dt at shutdown; conversely, if the current has high-frequency oscillations, even if the average value is stable, the rapid change at its peak may cause a severe di / dt at shutdown. In other words, relying solely on the static current snapshot value at the time of the shutdown instruction (i.e., the collector current value at the current moment) to predict the voltage spike may not be enough to accurately reflect the true dynamic effect of the current at the moment of shutdown. Therefore, to better reflect actual physical processes, the present application further decodes the collector current short-term fluctuation characteristics to quantify the short-term current change trend and fluctuation shape before shutdown into a correction factor for the current state at the current moment. In an embodiment of the present application, a multi-layer perceptron (MLP) is used to decode the collector current short-term fluctuation characteristics. As a feedforward neural network, the MLP can learn and map the complex nonlinear relationship between input features and target output. By performing multi-level nonlinear transformations on the collector current short-term fluctuation characteristics, the MLP learns the dynamic current fluctuation characteristics and maps them into a real-time collector current correction factor that can characterize the dynamic effects of the current. The real-time collector current correction factor is equivalent to assigning a weight to the current static current value based on the current's "dynamic momentum." For example, if the collector current short-term fluctuation characteristics indicate a sharp current fluctuation or a rapid upward trend, the decoded correction factor will increase the corrected current value; conversely, if the collector current short-term fluctuation characteristics indicate a stable or declining current, the correction factor may decrease or maintain its original value.

[0076] More specifically, step S332 corrects the current collector current value based on the real-time collector current correction factor to obtain a corrected real-time collector current value. Specifically, the real-time collector current correction factor is multiplied by the current collector current value to dynamically adjust the current value. This is done to account for fluctuations and changing trends in the current before shutdown, ensuring that the corrected collector current value better represents the actual current state at shutdown, thereby providing a more reliable basis for subsequent voltage spike prediction.

[0077] More specifically, step S333 inputs the corrected real-time collector current value and the initial turn-off gate resistance value into a circuit characteristic model to obtain a predicted collector-emitter voltage rise. It should be understood that the collector-emitter voltage rise (ΔVce) is physically determined primarily by the current change rate (di / dt) and the loop stray inductance (L_stray) during turn-off (ΔVce ≈ L_stray * di / dt). Di / dt itself is influenced by both the turn-off current magnitude (represented here by the corrected current value) and the turn-off speed (determined by the gate resistance Rg). Therefore, to estimate the primary components of the voltage spike by combining the corrected current information with the selected control parameters, the present application utilizes a circuit characteristic model to process the corrected real-time collector current value (Ic_corrected) and the initial turn-off gate resistance value (Rg_initial) to calculate the expected voltage rise resulting from their combined effect. Specifically, the circuit characteristic model is a lookup table calibrated with experimental data. Through offline simulation (using an accurate SPIC model) or experimental testing, the resulting Vee peak (Vce_peak) or overvoltage amplitude (Avce = Vce_peak - Vae) and dv / dt are recorded under different combinations of Ie, Vac, and possible gate drive parameters (especially different turn-off resistors Rg_off or different turn-off negative voltages Vge_off). This results in a series of (Ie, Rg_off, ΔVce) data points describing the expected collector-emitter voltage rise under different collector current and gate resistor combinations. In the actual prediction process, the predicted collector-emitter voltage rise can be quickly obtained by using the corrected collector current value and the initial turn-off gate resistance value as input and looking up the corresponding ΔVce value in the table.

[0078] Specifically, in step S34, the collector-emitter voltage value at the current moment is added to the predicted voltage rise of the collector-emitter voltage to obtain the predicted collector-emitter voltage peak amplitude. Specifically, since the total voltage stress that the IGBT bears when it is turned off is the sum of its steady-state voltage before shutdown (usually close to the DC bus voltage) and the dynamic voltage spike generated during the shutdown process (i.e., the predicted voltage rise). Therefore, in order to obtain a complete estimate of the highest voltage stress that may occur during this shutdown, the present application adds the collector-emitter voltage value at the current moment (Vce_current) to the predicted voltage rise of the collector-emitter voltage (ΔVce_predicted), so as to superimpose the starting voltage platform before shutdown and the voltage increment dynamically generated by shutdown, thereby estimating the peak value that Vce may reach during the entire shutdown process, which serves as the key decision basis for subsequent adjustment of the gate drive strategy.

[0079] Specifically, step S35 generates the initial gate drive strategy based on a comparison between the predicted collector-emitter voltage peak amplitude and a preset collector-emitter voltage threshold. Specifically, because using the initially assumed gate resistor for shutdown may result in a predicted voltage spike exceeding the device's safe operating range or optimization target range, in order to proactively take measures to ensure safety and optimize performance, the present application further makes forward-looking decisions based on a comparison between the predicted collector-emitter voltage peak amplitude (Vce_peak_predicted) and the preset collector-emitter voltage threshold. For example, if it is predicted that using the currently assumed gate resistor will result in an excessively high voltage spike, a safer drive strategy (typically meaning a slower shutdown) should be selected before shutdown begins. In this case, the controller will select a gate resistor larger than the initially assumed value as the actual initial gate drive strategy. Conversely, if the predicted voltage margin is sufficient, a faster shutdown strategy can be maintained or selected to reduce switching losses. In this case, the controller will maintain the original resistance or select an initial gate drive strategy with a smaller resistance value. In this way, control parameters can be proactively optimized before the shutdown operation is initiated, effectively preventing overvoltage risks and balancing safety and efficiency based on actual conditions.

[0080] In the above-described IGBT module overvoltage control method, step S4 controls the gate driver to shut down the IGBT module according to the initial gate drive strategy and monitors the collector-emitter voltage in real time to obtain a collector-emitter voltage data stream. It should be understood that, to verify the effectiveness of the initial gate drive strategy and provide a basis for possible online adjustments, the present application controls the gate driver to shut down the IGBT module according to the initial gate drive strategy while simultaneously capturing the actual evolution of Vce during the shutdown process through continuous high-speed monitoring. This provides true feedback data on voltage stress during the shutdown process for subsequent closed-loop correction, thus constituting a monitoring link.

[0081] In the above-mentioned overvoltage control method for the IGBT module, step S5 compares the collector-emitter voltage data stream with the collector-emitter voltage safety threshold, and determines whether to adjust the gate drive strategy based on the comparison result. Specifically, during the actual shutdown process, although a predictive initial strategy has been adopted, Vce may still rise rapidly and approach or even exceed the safety limit due to various reasons (such as sudden changes in operating conditions). Therefore, in order to ensure the absolute safety of the IGBT and provide the ultimate protection barrier, the present application compares the collector-emitter voltage data stream with the collector-emitter voltage safety threshold to implement a real-time closed-loop feedback protection mechanism. That is, the actual voltage is continuously monitored during the shutdown process. Once it is found that the voltage has a tendency to exceed the safety limit, the drive strategy adjustment program is immediately triggered to forcibly suppress the voltage rise. For example, quickly switch to a larger gate resistor, or activate the active clamping function (if supported by the hardware) to forcibly slow down the di / dt or directly clamp the voltage to prevent Vce from rising further. In this way, even if the initial prediction is not completely accurate or encounters extreme operating conditions, real-time closed-loop adjustments can be made to ensure that the IGBT module always operates within a safe voltage range.

[0082] In summary, the overvoltage control method of the IGBT module based on the embodiment of the present application is explained. After receiving the shutdown command, the method uses a deep learning algorithm to perform time-series correlation analysis on the collector current time series data including the current moment, extracts the short-term fluctuation characteristics of the collector current, and dynamically predicts the collector-emitter voltage spike amplitude that may be generated during the shutdown process in combination with the initial gate resistance value. Furthermore, based on the comparison between the predicted voltage spike amplitude and the preset threshold, the initial gate drive strategy is generated to guide the shutdown operation. At the same time, in the shutdown execution phase, the collector-emitter voltage data stream is monitored in real time and dynamically compared with the safety threshold to perform closed-loop optimization and adjustment of the drive strategy. This method realizes forward-looking assessment and adaptive control of overvoltage risks through intelligent prediction of the shutdown transient voltage spike, and can optimize switching losses while suppressing overvoltage, breaking through the limitations of traditional fixed parameter control.

[0083] Furthermore, an overvoltage control device for an IGBT module is provided.

[0084] Figure 7 FIG. 1 is a block diagram of an overvoltage control device for an IGBT module according to an embodiment of the present application. Figure 7As shown, the overvoltage control device 100 of the IGBT module according to the embodiment of the present application includes: a current and voltage data acquisition module 110, which is used to collect the collector current value and collector-emitter voltage value of the IGBT module in real time; a collector current value acquisition module 120, which is used to obtain the collector current values ​​at multiple predetermined time points including the current moment in response to receiving the IGBT shutdown instruction from the main controller; a spike prediction and strategy generation module 130, which is used to predict the collector-emitter voltage that may be generated by this shutdown based on the collector current values ​​at the multiple predetermined time points. The voltage spike amplitude is predicted, and an initial gate drive strategy is generated based on the predicted collector-emitter voltage spike amplitude; a gate driver control and monitoring module 140 is used to control the gate driver to turn off the IGBT module according to the initial gate drive strategy, and monitor the collector-emitter voltage value in real time to obtain a collector-emitter voltage data stream; a safety threshold comparison and strategy adjustment module 150 is used to compare the collector-emitter voltage data stream with the collector-emitter voltage safety threshold, and determine whether to adjust the gate drive strategy based on the comparison result.

[0085] Here, those skilled in the art will appreciate that the specific operations of each module in the overvoltage control device of the IGBT module have been described in detail above. Figures 1 to 6 The overvoltage control method of the IGBT module has been described in detail, and therefore, its repeated description will be omitted.

[0086] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in the present invention are merely illustrative and non-limiting, and should not be construed as necessarily possessed by each embodiment of the present invention. Furthermore, the specific details of the above embodiments are provided for illustrative purposes and to facilitate understanding, and are not intended to be limiting. These details do not necessarily limit the present invention to being implemented using these specific details.

[0087] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0088] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be encompassed therein. Any reference to a figure in a claim should not be construed as limiting the claim to which it relates.

[0089] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0090] Finally, it should be noted that the above description has been provided for the purpose of illustration and description. In addition, the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to be limiting. Although the technical solutions may be modified or replaced with equivalents with reference to the preferred embodiments, they do not depart from the spirit and scope of the technical solutions of the present invention.

Claims

1. An overvoltage control method for an IGBT module, characterized in that: include: Real-time acquisition of collector current and collector-emitter voltage of IGBT modules; In response to receiving an IGBT shutdown instruction from a main controller, obtaining collector current values ​​at a plurality of predetermined time points including a current time point; Based on the collector current values ​​at the plurality of predetermined time points, predicting a collector-emitter voltage spike amplitude that may be generated during the current shutdown, and generating an initial gate drive strategy based on the predicted collector-emitter voltage spike amplitude; Controlling the gate driver to turn off the IGBT module according to the initial gate drive strategy, and monitoring the collector-emitter voltage value in real time to obtain a collector-emitter voltage data stream; The collector-emitter voltage data stream is compared with a collector-emitter voltage safety threshold, and whether to adjust the gate drive strategy is determined based on the comparison result.

2. The overvoltage control method of the IGBT module according to claim 1, characterized in that: Based on the collector current values ​​at the plurality of predetermined time points, a collector-emitter voltage spike amplitude that may be generated during the current shutdown is predicted, and based on the predicted collector-emitter voltage spike amplitude, an initial gate drive strategy is generated, including: extracting collector current short-term timing fluctuation characteristics from the collector current values ​​at the plurality of predetermined time points; Get the initial turn-off gate resistance value; Calculating a predicted collector-emitter voltage rise based on the initial turn-off gate resistance value and the collector current short-term timing fluctuation characteristics; Adding the current collector-emitter voltage value to the predicted collector-emitter voltage rise to obtain the predicted collector-emitter voltage peak amplitude; The initial gate drive strategy is generated based on a comparison between the predicted collector-emitter voltage peak amplitude and a preset collector-emitter voltage threshold.

3. The overvoltage control method of the IGBT module according to claim 2, characterized in that: Extracting collector current short-term timing fluctuation characteristics from the collector current values ​​at the plurality of predetermined time points includes: Arranging the collector current values ​​at the plurality of predetermined time points into a collector current short-time input vector according to a time dimension; Performing time series multi-scale coding on the collector current short-time input vector to obtain a multi-scale coding vector of the collector current short-time time series fluctuation characteristics; The multi-scale encoding vector of the collector current short-time timing fluctuation feature is optimized in a time-series fine-grained manner to obtain an optimized encoding vector of the collector current short-time timing fluctuation feature as the collector current short-time timing fluctuation feature.

4. The overvoltage control method of the IGBT module according to claim 3, characterized in that: Performing time series multi-scale encoding on the collector current short-time input vector to obtain a multi-scale encoding vector of the collector current short-time time series fluctuation characteristics includes: The collector current short-time input vector is input into a temporal multi-scale encoder based on a CNN-BiGRU hybrid architecture to obtain a multi-scale encoding vector of the collector current short-time temporal fluctuation characteristics.

5. The overvoltage control method of the IGBT module according to claim 4, characterized in that: Performing time series fine-grained expression optimization on the multi-scale encoding vector of the collector current short-time timing fluctuation feature to obtain an optimized encoding vector of the collector current short-time timing fluctuation feature as the collector current short-time timing fluctuation feature, including: Performing time series local fluctuation modal reconfiguration on the multi-scale encoding vector of the collector current short-time temporal fluctuation feature to obtain a set of collector current short-time temporal fluctuation local modal extraction feature encoding vectors; Performing a feature reconstruction gain calculation based on the energy density of the fluctuation component on each collector current short-time time series fluctuation local modal extraction feature coding vector in the set of the collector current short-time time series fluctuation local modal extraction feature coding vectors to obtain a set of collector current short-time time series fluctuation local modal feature reconstruction gain operators; Based on a set of reshaping gain operators for the local modal characteristics of the collector current short-term temporal fluctuation, the multi-scale encoding vector of the collector current short-term temporal fluctuation characteristics is significantly reshaped to obtain the optimized encoding vector of the collector current short-term temporal fluctuation characteristics.

6. The overvoltage control method of the IGBT module according to claim 5, characterized in that: Performing time series local fluctuation modal reconfiguration on the multi-scale encoding vector of the collector current short-time temporal fluctuation feature to obtain a set of collector current short-time temporal fluctuation local modal extraction feature encoding vectors, including: Performing time series local fluctuation modal reconstruction based on one-dimensional convolution coding on the multi-scale coding vector of the collector current short-time time series fluctuation feature to obtain a set of collector current short-time time series fluctuation local modal feature coding vectors; Information extraction is performed on each collector current short-term timing fluctuation local modal feature coding vector in the set of the collector current short-term timing fluctuation local modal feature coding vectors to obtain the set of the collector current short-term timing fluctuation local modal extraction feature coding vectors.

7. The overvoltage control method of the IGBT module according to claim 6, characterized in that: Performing a feature reconstruction gain calculation based on the energy density of the fluctuation component on each collector current short-time time series fluctuation local modal extraction feature coding vector in the set of the collector current short-time time series fluctuation local modal extraction feature coding vectors to obtain a set of collector current short-time time series fluctuation local modal feature reconstruction gain operators, including: Calculating the effective fluctuation component statistical coefficient of each collector current short-time time series fluctuation local modal extraction feature coding vector in the set of the collector current short-time time series fluctuation local modal extraction feature coding vector; Based on the statistical coefficients of the effective fluctuation components of the local modal extraction feature coding vectors of the short-term temporal fluctuations of each collector current, the feature reshaping gain operators of the local modal extraction feature coding vectors of the short-term temporal fluctuations of each collector current are calculated to obtain a set of the local modal feature reshaping gain operators of the short-term temporal fluctuations of the collector current.

8. The overvoltage control method of the IGBT module according to claim 7, characterized in that: Based on a set of local modal feature reshaping gain operators of the collector current short-term temporal fluctuation, a multi-scale encoding vector of the collector current short-term temporal fluctuation feature is significantly reshaped to obtain an optimized encoding vector of the collector current short-term temporal fluctuation feature, including: Performing a time-invariant gain convergence stabilization process on the feature reshaping gain operators of the local modal extraction feature coding vectors of the respective collector current short-term temporal fluctuations to obtain a set of optimized collector current short-term temporal fluctuation local modal feature reshaping gain operators; Normalizing the set of operators for optimizing the reshaping gain of the local modal characteristics of the collector current short-term temporal fluctuation to obtain a set of weights for reshaping the gain of the local modal characteristics of the collector current short-term temporal fluctuation; Based on the set of reshaping gain weights of the local modal characteristics of the collector current short-term timing fluctuations, the set of the local modal characteristics encoding vectors of the collector current short-term timing fluctuations is weightedly aggregated to obtain the optimized encoding vector of the collector current short-term timing fluctuation characteristics.

9. The overvoltage control method of the IGBT module according to claim 2, characterized in that: Calculating a predicted collector-emitter voltage rise based on the initial turn-off gate resistance value and the collector current short-term timing fluctuation characteristics includes: Decoding the collector current short-term timing fluctuation characteristics to obtain a real-time collector current correction factor; Correcting the collector current value at the current moment based on the real-time collector current correction factor to obtain a corrected real-time collector current value; The corrected real-time collector current value and the initial turn-off gate resistance value are input into a circuit characteristic model to obtain a predicted collector-emitter voltage rise.

10. An overvoltage control device for an IGBT module, characterized in that: include: Current and voltage data acquisition module, used to collect the collector current value and collector-emitter voltage value of the IGBT module in real time; a collector current value acquisition module, configured to acquire collector current values ​​at a plurality of predetermined time points including the current time point in response to receiving an IGBT shutdown instruction from the main controller; a spike prediction and strategy generation module, configured to predict the collector-emitter voltage spike amplitude that may be generated by the current shutdown based on the collector current values ​​at the plurality of predetermined time points, and to generate an initial gate drive strategy based on the predicted collector-emitter voltage spike amplitude; a gate driver control and monitoring module, configured to control the gate driver to turn off the IGBT module according to the initial gate drive strategy, and to monitor the collector-emitter voltage value in real time to obtain a collector-emitter voltage data stream; The safety threshold comparison and strategy adjustment module is used to compare the collector-emitter voltage data stream with the collector-emitter voltage safety threshold, and determine whether to adjust the gate drive strategy based on the comparison result.

Images