Method and system for constructing electromagnetic transient IGBT real-time model
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-12-17
- Publication Date
- 2026-08-13
Smart Images
Figure CN2025143128_13082026_PF_FP_ABST
Abstract
Description
A method and system for constructing a real-time model of electromagnetic transient IGBTs
[0001] This application claims priority to Chinese Patent Application No. 202510132464.6, filed on February 6, 2025, entitled "A Method and System for Constructing a Real-Time Model of an Electromagnetic Transient IGBT", the entire contents of which are incorporated herein by reference. Technical Field
[0002] This invention relates to the field of low-frequency power transmission technology, and in particular to a method and system for constructing a real-time model of electromagnetic transient IGBTs. Background Technology
[0003] With the rapid development of new power electronics technologies and the introduction of flexible low-frequency power transmission technology, high-voltage, high-capacity, long-distance low-frequency power transmission has become technically feasible. The converter valve is a core device in the low-frequency power transmission field, used to achieve frequency conversion. The IGBT (Insulated Gate Bipolar Transistor) is the core component of the converter valve, and its operational stability directly affects the reliability of the converter valve. The IGBT turn-off transient process is a crucial factor in device application; IGBT turn-off overvoltage breakdown is one of the main causes of failure. Establishing an IGBT turn-off transient simulation model is an important method for analyzing IGBT turn-off characteristics.
[0004] Currently, existing IGBT transient simulation models include analytical models and numerical models. Analytical models offer high simulation accuracy, but the physical parameters are difficult to obtain from manufacturers, making modeling challenging. Numerical models approximate the IGBT's transient process using mathematical models, but because the turn-off processes of IGBTs from different manufacturers and models vary, the simulation accuracy of numerical models is often difficult to guarantee, and the modeling process is complex. Therefore, both analytical and numerical models consume significant computational resources and are not suitable for real-time simulation scenarios, resulting in low practicality of IGBT transient simulation models. Summary of the Invention
[0005] This invention provides a method and system for constructing a real-time electromagnetic transient IGBT model, which solves the technical problem that existing IGBT transient simulation models all require a large amount of computing resources and are not suitable for real-time simulation scenarios, resulting in low practicality of IGBT transient simulation models.
[0006] The first aspect of this invention provides a method for constructing a real-time model of an electromagnetic transient IGBT, comprising:
[0007] Obtain the structural data and training data of the IGBT to be constructed, perform data preprocessing on the training data, and generate corresponding training feature data;
[0008] The target gate switch delay model is obtained by simplifying and solving the preset gate switch delay model using the structural data.
[0009] The preset electromagnetic transient model is trained using the training feature data to generate the target electromagnetic transient model;
[0010] Using the target gate switch delay model and the target electromagnetic transient model, a real-time electromagnetic transient IGBT model corresponding to the IGBT to be constructed is built.
[0011] Optionally, the gate switch delay model includes an on-delay model and an off-delay model. The step of simplifying and solving the preset gate switch delay model using the structure data to obtain the target gate switch delay model includes:
[0012] Using the aforementioned structural data, corresponding composite bonds are generated;
[0013] According to the composite key in the preset device data key-value pair table, match the corresponding turn-on circuit parameters and turn-off circuit parameters.
[0014] Input the turn-on circuit parameters and the structural data into the turn-on delay model to obtain the target turn-on delay model and the first coefficient;
[0015] The first coefficient, the structural data, and the shutdown circuit parameters are input into the shutdown delay model to obtain the target shutdown delay model;
[0016] The target turn-on delay model and the target turn-off delay model are coupled to obtain the target gate switch delay model.
[0017] Optionally, the step of training a preset electromagnetic transient model using the training feature data to generate a target electromagnetic transient model includes:
[0018] The training feature data is input into a preset electromagnetic transient model for training, and the training working condition waveform is output. The electromagnetic transient model includes a standard convolutional layer, a convolutional module, a batch normalization layer and a Tanh activation layer connected in sequence.
[0019] Select a standard operating condition waveform diagram that is associated with the training operating condition waveform diagram from the training feature data;
[0020] Calculate the overlap between the training condition waveform and the standard condition waveform, and determine whether the overlap is not greater than a preset overlap threshold.
[0021] If not, stop training and generate the target electromagnetic transient model;
[0022] If so, the gradient descent method is used to adjust the model parameters of the electromagnetic transient model, and the process returns to the step of inputting the training feature data into the preset electromagnetic transient model for training and outputting the training condition waveform.
[0023] Optionally, the step of inputting the training feature data into a preset electromagnetic transient model for training and outputting a training operating condition waveform includes:
[0024] The training feature data is convolved using the standard convolutional layer to obtain a data feature map.
[0025] The convolution module performs residual processing on the data feature map to obtain a waveform feature map;
[0026] The waveform feature map is standardized by the batch normalization layer in the electromagnetic transient model to obtain a standard waveform feature map.
[0027] The standard waveform feature map is extracted by using the Tanh activation layer in the electromagnetic transient model to obtain the training condition waveform map.
[0028] Optionally, the convolutional module includes a first extraction branch and a feature fusion layer. The step of performing residual processing on the data feature map through the convolutional module to obtain a waveform feature map includes:
[0029] The first extraction branch is used to extract features from the data feature map to obtain a first feature map. The first extraction branch includes a first convolutional layer, a second convolutional layer, a batch normalization layer, a Tanh activation layer and a 1×1 standard convolutional layer connected in sequence.
[0030] The first feature map and the data feature map are fused by a feature fusion layer to obtain a waveform feature map.
[0031] Optionally, the step of constructing the electromagnetic transient IGBT real-time model corresponding to the IGBT to be constructed using the target gate switch delay model and the target electromagnetic transient model includes:
[0032] By combining the target gate switch delay model and the target electromagnetic transient model, an initial electromagnetic transient model is obtained.
[0033] The initial electromagnetic transient model is loaded into a preset on-chip system to obtain the electromagnetic transient IGBT real-time model corresponding to the IGBT to be constructed.
[0034] The second aspect of this invention provides a real-time model construction system for electromagnetic transient IGBTs, comprising:
[0035] The preprocessing module is used to acquire the structural data and training data of the IGBT to be constructed, perform data preprocessing on the training data, and generate corresponding training feature data.
[0036] The first analysis module is used to simplify and solve the preset gate switch delay model using the structural data to obtain the target gate switch delay model.
[0037] The second analysis module is used to train the preset electromagnetic transient model using the training feature data to generate the target electromagnetic transient model.
[0038] The construction module is used to construct the electromagnetic transient IGBT real-time model corresponding to the IGBT to be constructed, using the target gate switch delay model and the target electromagnetic transient model.
[0039] A third aspect of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the electromagnetic transient IGBT real-time model construction method as described in any of the preceding claims.
[0040] The fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, implements the method for constructing a real-time electromagnetic transient IGBT model as described in any of the preceding claims.
[0041] The fifth aspect of the present invention provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein, when the program instructions are executed by a computer, the computer performs the electromagnetic transient IGBT real-time model construction method as described in any of the preceding claims.
[0042] As can be seen from the above technical solutions, the present invention has the following advantages:
[0043] By simplifying the conditions for physical interpretability of the IGBT to be constructed using structural data, a real-time electromagnetic transient IGBT model is obtained. This real-time electromagnetic transient IGBT model reduces the computational load during IGBT turn-on and turn-off simulations and can output accurate simulation results based on the input operating conditions. This overcomes the technical problem that existing IGBT transient simulation models consume large amounts of computational resources and are not suitable for real-time simulation scenarios, resulting in low practicality. Compared with traditional IGBT modeling methods, this invention simplifies the pre-set gate switching delay model using structural data, reducing the computational load during IGBT simulation and improving the practicality of the IGBT transient simulation model. Attached Figure Description
[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0045] Figure 1 is a flowchart of the steps of a real-time model construction method for electromagnetic transient IGBTs provided in Embodiment 1 of the present invention;
[0046] Figure 2 is a flowchart of the steps of a real-time model construction method for electromagnetic transient IGBTs provided in Embodiment 2 of the present invention;
[0047] Figure 3 shows the gate emitter voltage and gate current waveforms at the moment of IGBT turn-on and turn-off provided in Embodiment 2 of the present invention.
[0048] Figure 4 is a device data list provided in Embodiment 2 of the present invention;
[0049] Figure 5 is a waveform of the gate collector current at the moment of IGBT turn-off provided in Embodiment 2 of the present invention;
[0050] Figure 6 is an equivalent circuit diagram of the IGBT provided in Embodiment 2 of the present invention;
[0051] Figure 7 is a schematic diagram of the electromagnetic transient model provided in Embodiment 2 of the present invention;
[0052] Figure 8 is a structural block diagram of an electromagnetic transient IGBT real-time model construction system provided in Embodiment 3 of the present invention;
[0053] Figure 9 is a structural block diagram of an electronic device provided in Embodiment 4 of the present invention. Detailed Implementation
[0054] This invention provides a method and system for constructing a real-time electromagnetic transient IGBT model, which addresses the technical problem that existing IGBT transient simulation models consume a large amount of computing resources and are not suitable for real-time simulation scenarios, resulting in low practicality of IGBT transient simulation models.
[0055] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.
[0056] Please refer to Figure 1, which is a flowchart of the steps of a real-time model construction method for electromagnetic transient IGBTs provided in Embodiment 1 of the present invention.
[0057] This invention provides a method for constructing a real-time model of an electromagnetic transient IGBT, comprising:
[0058] Step 101: Obtain the structural data and training data of the IGBT to be constructed, perform data preprocessing on the training data, and generate corresponding training feature data;
[0059] Structural data refers to the parasitic resistance, circuit connection resistance, and gate-emitter capacitance of the IGBT to be constructed.
[0060] Training data refers to the historical operating condition waveform data of the IGBT to be constructed. The historical operating condition waveform data includes the IGBT resistance value, controllable current source and operating condition waveform under normal steady state conditions.
[0061] In this embodiment of the invention, structural data and training data of the IGBT to be constructed are obtained, and data cleaning is performed on the training data to obtain training feature data.
[0062] Step 102: Simplify and solve the preset gate switch delay model using structural data to obtain the target gate switch delay model;
[0063] In this embodiment of the invention, structural data is input into a preset device data table, and the corresponding turn-on delay time and turn-off delay time are matched. The gate switch delay model is solved using the structural data, turn-on delay time and turn-off delay time to obtain the target gate switch delay model.
[0064] Step 103: Train the preset electromagnetic transient model using training feature data to generate the target electromagnetic transient model;
[0065] In this embodiment of the invention, a preset electromagnetic transient model is trained using training feature data to obtain a trained target electromagnetic transient model.
[0066] It should be noted that the values of resistance and current source during IGBT switching transients are derived from the trained electromagnetic transient model (i.e., neural network). The rise time, fall time, and direct control of the current source to rise or fall linearly from the original value to the corresponding value are used as training feature data (or multiple actual operating condition waveforms are used, and the waveform points to be trained are extracted in the form of discretized sampling). The electromagnetic transient model is trained according to different operating conditions (such as different temperatures and humidity), so that the trained electromagnetic transient model can quickly infer the waveform under the input operating condition.
[0067] It is worth mentioning that the electromagnetic transient model training process is as follows: 1. Forward propagation: Input the training feature data into the preset electromagnetic transient model, and calculate the model's prediction results through each layer of the network. 2. Loss calculation: Calculate the loss value using a preset loss function based on the prediction results and the true labels of the training feature data. 3. Backpropagation: Calculate the gradient of the model parameters using the backpropagation algorithm based on the loss value. The gradient represents the rate of change of the loss function with respect to each parameter and is used to guide the direction of parameter updates. 4. Parameter update: Update the model parameters using the optimizer based on the calculated gradients. The updated parameters will gradually reduce the value of the loss function. 5. Repeat steps 1-4 until the predetermined number of training epochs is reached or the stopping condition is met (such as the loss function converging).
[0068] Step 104: Using the target gate switch delay model and the target electromagnetic transient model, construct the electromagnetic transient IGBT real-time model corresponding to the IGBT to be constructed.
[0069] In this embodiment of the invention, the target gate switch delay model and the target electromagnetic transient model are combined to obtain the initial electromagnetic transient IGBT real-time model, and the initial electromagnetic transient IGBT real-time model is loaded into a preset on-chip system to obtain the electromagnetic transient IGBT real-time model corresponding to the IGBT to be constructed.
[0070] In this embodiment of the invention, by acquiring structural data and training feature data, the conditions for physical interpretability of the IGBT to be constructed are simplified to obtain a real-time electromagnetic transient IGBT model. This real-time electromagnetic transient IGBT model reduces the computational load during IGBT turn-on and turn-off simulations and can output accurate simulation results based on the input operating conditions. This overcomes the technical problem that existing IGBT transient simulation models consume large amounts of computational resources and are not suitable for real-time simulation scenarios, resulting in low practicality. Compared with traditional IGBT modeling methods, this invention simplifies the pre-set gate switch delay model using structural data, reducing the computational load for delay turn-on and delay turn-off times during IGBT simulations and improving the practicality of the IGBT transient simulation model.
[0071] Please refer to Figure 2, which is a flowchart of the steps of a real-time model construction method for electromagnetic transient IGBTs provided in Embodiment 2 of the present invention.
[0072] This invention provides a method for constructing a real-time model of an electromagnetic transient IGBT, comprising:
[0073] Step 201: Obtain the structural data and training data of the IGBT to be constructed, perform data preprocessing on the training data, and generate corresponding training feature data;
[0074] In this embodiment of the invention, when the structural data and training data of the IGBT to be constructed are received, the training data is preprocessed to generate corresponding training feature data.
[0075] Step 202: Simplify and solve the preset gate switch delay model using structural data to obtain the target gate switch delay model;
[0076] Furthermore, the gate switch delay model includes an on-delay model and an off-delay model, and step 202 includes the following sub-steps:
[0077] S11. Using structural data, generate the corresponding composite bonds;
[0078] In this embodiment of the invention, the structural data of the IGBT to be constructed is used to generate the corresponding composite bond.
[0079] S12. Match the corresponding turn-on circuit parameters and turn-off circuit parameters according to the preset device data key-value pair table of the composite key.
[0080] The turn-on circuit parameters refer to the turn-on delay time of the IGBT to be constructed under normal steady-state conditions.
[0081] The turn-off circuit parameters refer to the turn-off delay time of the IGBT to be constructed under normal steady-state conditions.
[0082] In this embodiment of the invention, referring to Figure 4, the composite key is input into a preset device data key-value pair table to match the corresponding turn-on circuit parameters and turn-off circuit parameters.
[0083] S13. Input the turn-on circuit parameters and structural data into the turn-on delay model to obtain the target turn-on delay model and the first coefficient;
[0084] In this embodiment of the invention, the turn-on circuit parameters and structural data are input into the turn-on delay model for simplified solution to obtain the target turn-on delay model and the first coefficient.
[0085] It should be noted that, referring to Figure 3, at the instant of switching on, the gate-emitter voltage changes from a negative to a positive value. Due to the parasitic resistance inside the circuit and the external buffer circuit, the voltage change cannot be realized abruptly, resulting in a first-order response of the circuit. This response is the full response circuit of the resistance and capacitance between the gate and emitter. The resistance value needs to consider the internal parasitic resistance of the switch and the external connection resistance. The formula for calculating the resistance value is R. 门极回路 =R 寄生电阻 +R 电路接入 When the circuit is not turned on, the voltage between the collector and emitter is high. Under this characteristic, the capacitance values between the three electrodes can be considered as almost constant characteristic curves. Therefore, the turn-on delay model of the IGBT needs to be constructed.
[0086] The specific activation delay model is as follows: T 开通延迟时间 =A×R 门极回路 ×C GE ;
[0087] Among them, T 开通延迟时间 The activation delay time is represented by A, which is the first coefficient, and R is the activation delay time. 门极回路 C is the gate circuit resistance. GE This is the gate-emitter capacitor.
[0088] S14. Input the first coefficient, structural data and turn-off circuit parameters into the turn-off delay model to obtain the target turn-off delay model;
[0089] In this embodiment of the invention, the first coefficient, structural data, and turn-off circuit parameters are input into the turn-off delay model for simplified solution to obtain the target turn-off delay model.
[0090] It should be noted that, as shown in Figure 3, the turn-off instant is divided into two stages. In the first stage, the emitter-collector voltage remains unchanged, and the gate-emitter capacitance C... GEA first-order discharge response will occur. In the second stage, the emitter-collector voltage changes, and the three capacitance values will undergo drastic nonlinear changes. The specific transient changes can be derived through complex nonlinear calculations. However, the time in this stage remains essentially constant, or it can be considered a process with a particularly large time constant, and the time is relatively fixed. Referring to Figure 5, the time constant D≈t. d(off) -t d(on) Since the slopes of both are basically the same, their growth rates can be considered to be the same. Therefore, by simplifying the calculation process at the moment of shutdown, the shutdown delay model can be obtained as follows: T 关断延迟时间1 =B*R 门极回路 *C GE T 关断延迟时间2 =C*R 门极回路 *C GE ≈DT 总关断延迟时间 =T 关断延迟时间1 +T 关断延迟时间2 =B*R 门极回路 *C GE +DB≈A
[0091] Among them, T 关断延迟时间1 T is the first shutdown delay time. 关断延迟时间2 Where D is the second turn-off delay time, B is the time constant, and T is the second coefficient. 总关断延迟时间 C is the turn-off delay time, and C is the third coefficient.
[0092] S15. Couple the target turn-on delay model and the target turn-off delay model to obtain the target gate switching delay model.
[0093] In this embodiment of the invention, the target turn-on delay model and the target turn-off delay model are combined to obtain the target gate switch delay model.
[0094] Step 203: Input the training feature data into the preset electromagnetic transient model for training and output the training working condition waveform. The electromagnetic transient model includes a standard convolutional layer, a convolutional module, a batch normalization layer and a Tanh activation layer connected in sequence.
[0095] It should be noted that the standard convolutional layer is a 10×10 standard convolutional layer.
[0096] It is worth mentioning that the IGBT will be connected in a two-node configuration to reduce the number of nodes in the system. The specific circuit representation is a structure of a variable resistor connected in parallel with a controllable current source, as shown in Figure 6. The IGBT resistance value and the controllable current source under normal steady-state conditions are derived from its steady-state resistance value and bias voltage.
[0097] Further, referring to Figure 7, step 203 includes the following sub-steps:
[0098] S21. Perform convolution operations on the training feature data using standard convolutional layers to obtain the data feature map;
[0099] In this embodiment of the invention, the training feature data is processed by a 10×10 standard convolutional layer to obtain a data feature map.
[0100] S22. Perform residual processing on the data feature map using the convolution module to obtain the waveform feature map;
[0101] Furthermore, the convolutional module includes a first extraction branch and a feature fusion layer, and S22 includes the following sub-steps:
[0102] S221. The first extraction branch is used to extract features from the data feature map to obtain the first feature map. The first extraction branch includes a first convolutional layer, a second convolutional layer, a batch normalization layer, a Tanh activation layer and a 1×1 standard convolutional layer connected in sequence.
[0103] In this embodiment of the invention, a first feature map is obtained by extracting features from the data feature map through a 1×1 convolutional layer, a 3×3 convolutional layer, a batch normalization layer, a Tanh activation layer, and a 1×1 standard convolutional layer connected in sequence.
[0104] S222. The first feature map and the data feature map are fused by the feature fusion layer to obtain the waveform feature map.
[0105] In this embodiment of the invention, a feature fusion layer is used to fuse the first feature map and the data feature map to obtain a waveform feature map.
[0106] S23. Standardize the waveform feature map by performing a batch normalization layer in the electromagnetic transient model to obtain a standard waveform feature map;
[0107] S24. The standard waveform feature map is extracted by using the Tanh activation layer in the electromagnetic transient model to obtain the training condition waveform map.
[0108] In this embodiment of the invention, the waveform feature map is extracted sequentially through a batch normalization layer and a Tanh activation layer to obtain the training condition waveform map.
[0109] It's worth noting that, as a non-linear activation function, Tanh enables neural networks to learn non-linear relationships in input data. For example, in image recognition tasks, the relationships between features such as object shape and texture in an image are often non-linear. Tanh activation layers can help neural networks model these complex non-linear relationships, thereby improving the model's ability to recognize images.
[0110] Step 204: Select standard operating condition waveforms that are associated with the training operating condition waveforms from the training feature data;
[0111] In this embodiment of the invention, a standard operating condition waveform diagram corresponding to the training operating condition waveform diagram is selected from the training feature data.
[0112] Step 205: Calculate the overlap between the training condition waveform and the standard condition waveform, and determine whether the overlap is not greater than the preset overlap threshold.
[0113] In this embodiment of the invention, the overlap between the training condition waveform and the standard condition waveform is calculated, and it is determined whether the overlap exceeds a preset overlap threshold.
[0114] Step 206: If not, stop training and generate the target electromagnetic transient model;
[0115] In this embodiment of the invention, when the overlap degree is greater than the overlap degree threshold, training is stopped and the target electromagnetic transient model is generated.
[0116] Step 207: If yes, then use the gradient descent method to adjust the model parameters of the electromagnetic transient model, return to the step of inputting the training feature data into the preset electromagnetic transient model for training, and outputting the training condition waveform diagram.
[0117] In this embodiment of the invention, when the overlap degree is less than or equal to the overlap degree threshold, the model parameters of the electromagnetic transient model are adjusted by gradient descent, grid search or random search, and the process jumps to step 203 until the overlap degree is greater than the overlap degree threshold.
[0118] Step 208: Using the target gate switch delay model and the target electromagnetic transient model, construct the electromagnetic transient IGBT real-time model corresponding to the IGBT to be constructed.
[0119] Furthermore, step 208 includes the following sub-steps:
[0120] S31. Combine the target gate switch delay model and the target electromagnetic transient model to obtain the initial electromagnetic transient model;
[0121] In this embodiment of the invention, the target gate switch delay model and the target electromagnetic transient model are coupled to obtain the initial electromagnetic transient model.
[0122] S32. Load the initial electromagnetic transient model into the preset on-chip system to obtain the electromagnetic transient IGBT real-time model corresponding to the IGBT to be constructed.
[0123] In this embodiment of the invention, the initial electromagnetic transient model is decomposed into multiple transient analysis sub-modules and loaded into each core of a preset on-chip system to obtain the electromagnetic transient IGBT real-time model corresponding to the IGBT to be constructed.
[0124] In this embodiment of the invention, by acquiring structural data and training feature data, the conditions for physical interpretability of the IGBT to be constructed are simplified to obtain a real-time electromagnetic transient IGBT model. This real-time electromagnetic transient IGBT model reduces the computational load during IGBT turn-on and turn-off simulations and can output accurate simulation results based on the input operating conditions. This overcomes the technical problem that existing IGBT transient simulation models consume large amounts of computational resources and are not suitable for real-time simulation scenarios, resulting in low practicality. Compared with traditional IGBT modeling methods, this invention simplifies the pre-set gate switch delay model using structural data, reducing the computational load for delay turn-on and delay turn-off times during IGBT simulations and improving the practicality of the IGBT transient simulation model.
[0125] Please refer to Figure 8, which is a structural block diagram of an electromagnetic transient IGBT real-time model construction system provided in Embodiment 3 of the present invention.
[0126] This invention provides a real-time model building system for electromagnetic transient IGBTs, comprising:
[0127] Preprocessing module 301 is used to acquire the structural data and training data of the IGBT to be constructed, perform data preprocessing on the training data, and generate corresponding training feature data.
[0128] The first analysis module 302 is used to simplify and solve the preset gate switch delay model using structural data to obtain the target gate switch delay model.
[0129] The second analysis module 303 is used to train a preset electromagnetic transient model using training feature data to generate a target electromagnetic transient model.
[0130] Module 304 is used to construct the real-time electromagnetic transient IGBT model corresponding to the IGBT to be constructed, using the target gate switching delay model and the target electromagnetic transient model.
[0131] Furthermore, the gate switch delay model includes an on-delay model and an off-delay model. The first analysis module 302 includes:
[0132] The first construction submodule is used to generate corresponding composite keys from structural data;
[0133] The matching submodule is used to match the corresponding turn-on circuit parameters and turn-off circuit parameters according to the composite key in the preset device data key-value pair table.
[0134] The first parsing submodule is used to input the turn-on circuit parameters and structural data into the turn-on delay model to obtain the target turn-on delay model and the first coefficient;
[0135] The second analysis submodule is used to input the first coefficient, structural data and turn-off circuit parameters into the turn-off delay model to obtain the target turn-off delay model;
[0136] The coupling submodule is used to couple the target turn-on delay model and the target turn-off delay model to obtain the target gate switching delay model.
[0137] Furthermore, the second analysis module 303 includes:
[0138] The training submodule is used to input training feature data into a preset electromagnetic transient model for training and output a training waveform diagram. The electromagnetic transient model includes a standard convolutional layer, a convolutional module, a batch normalization layer and a Tanh activation layer connected in sequence.
[0139] The filtering submodule is used to select standard operating condition waveforms that are associated with the training operating condition waveforms from the training feature data;
[0140] The first analysis submodule is used to calculate the overlap between the training condition waveform and the standard condition waveform, and to determine whether the overlap is not greater than the preset overlap threshold.
[0141] If not, stop training and generate the target electromagnetic transient model;
[0142] If so, the gradient descent method is used to adjust the model parameters of the electromagnetic transient model, and the process of inputting the training feature data into the preset electromagnetic transient model for training and outputting the training condition waveform is returned.
[0143] Furthermore, the training submodule includes:
[0144] The first convolutional unit is used to perform convolution operations on the training feature data using standard convolutional layers to obtain a data feature map.
[0145] The residual unit is used to perform residual processing on the data feature map through the convolution module to obtain the waveform feature map;
[0146] The standardization unit is used to perform a standard waveform feature map by applying the batch normalization layer in the electromagnetic transient model to the waveform feature map.
[0147] The second convolutional unit is used to extract features from the standard waveform feature map through the Tanh activation layer in the electromagnetic transient model to obtain the training condition waveform map.
[0148] Furthermore, the convolutional module includes a first extraction branch and a feature fusion layer, and residual units, including:
[0149] The feature extraction submodule is used to extract features from the data feature map using the first extraction branch to obtain the first feature map. The first extraction branch includes a first convolutional layer, a second convolutional layer, a batch normalization layer, a Tanh activation layer and a 1×1 standard convolutional layer connected in sequence.
[0150] The feature fusion submodule is used to perform feature fusion on the first feature map and the data feature map through the feature fusion layer to obtain the waveform feature map.
[0151] Furthermore, module 304 is constructed, including:
[0152] The simultaneous equation submodule is used to combine the target gate switch delay model and the target electromagnetic transient model to obtain the initial electromagnetic transient model;
[0153] The loading submodule is used to load the initial electromagnetic transient model into the preset on-chip system to obtain the electromagnetic transient IGBT real-time model corresponding to the IGBT to be constructed.
[0154] Please refer to Figure 9, which is a structural block diagram of an electronic device provided in Embodiment 4 of the present invention.
[0155] An electronic device according to an embodiment of the present invention includes: a memory 401 and a processor 402. The memory 402 stores a computer program. When the computer program is executed by the processor 402, the processor 402 executes the electromagnetic transient IGBT real-time model construction method as described in any of the above embodiments.
[0156] Memory 401 may be an electronic memory such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM. Memory 401 has storage space 403 for program code 413 for performing any of the method steps described above. For example, storage space 403 for program code may include individual program codes 413 for implementing the various steps in the methods described above. This program code may be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact discs (CDs), memory cards, or floppy disks. The program code may be compressed, for example, in a suitable form. When run by a computing processing device, this code causes the computing processing device to perform the various steps in the methods described above.
[0157] Embodiment 5 of the present invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the electromagnetic transient IGBT real-time model construction method as described in any of the above embodiments.
[0158] Embodiment 6 of the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer performs the electromagnetic transient IGBT real-time model construction method as described in any of the above embodiments.
[0159] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0160] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.
[0161] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0162] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0163] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0164] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for constructing an electromagnetic transient IGBT real-time model, characterized in that, include: Obtain the structural data and training data of the IGBT to be constructed, perform data preprocessing on the training data, and generate corresponding training feature data; The target gate switch delay model is obtained by simplifying and solving the preset gate switch delay model using the structural data. The preset electromagnetic transient model is trained using the training feature data to generate the target electromagnetic transient model; Using the target gate switch delay model and the target electromagnetic transient model, a real-time electromagnetic transient IGBT model corresponding to the IGBT to be constructed is built.
2. The electromagnetic transient IGBT real-time model construction method according to claim 1, characterized in that, The gate switch delay model includes an on-delay model and an off-delay model. The step of simplifying and solving the preset gate switch delay model using the structured data to obtain the target gate switch delay model includes: Using the aforementioned structural data, corresponding composite bonds are generated; According to the composite key in the preset device data key-value pair table, match the corresponding turn-on circuit parameters and turn-off circuit parameters. Input the turn-on circuit parameters and the structural data into the turn-on delay model to obtain the target turn-on delay model and the first coefficient; The first coefficient, the structural data, and the shutdown circuit parameters are input into the shutdown delay model to obtain the target shutdown delay model; The target turn-on delay model and the target turn-off delay model are coupled to obtain the target gate switch delay model.
3. The electromagnetic transient IGBT real-time model construction method according to claim 1, characterized by, The step of training a preset electromagnetic transient model using the training feature data to generate a target electromagnetic transient model includes: The training feature data is input into a preset electromagnetic transient model for training, and the training working condition waveform is output. The electromagnetic transient model includes a standard convolutional layer, a convolutional module, a batch normalization layer and a Tanh activation layer connected in sequence. Select a standard operating condition waveform diagram that is associated with the training operating condition waveform diagram from the training feature data; Calculate the overlap between the training condition waveform and the standard condition waveform, and determine whether the overlap is not greater than a preset overlap threshold. If not, stop training and generate the target electromagnetic transient model; If so, the gradient descent method is used to adjust the model parameters of the electromagnetic transient model, and the process returns to the step of inputting the training feature data into the preset electromagnetic transient model for training and outputting the training condition waveform.
4. The electromagnetic transient IGBT real-time model construction method according to claim 3, characterized by, The step of inputting the training feature data into a preset electromagnetic transient model for training and outputting a training condition waveform includes: The training feature data is convolved using the standard convolutional layer to obtain a data feature map. The convolution module performs residual processing on the data feature map to obtain a waveform feature map; The waveform feature map is standardized by the batch normalization layer in the electromagnetic transient model to obtain a standard waveform feature map. The standard waveform feature map is extracted by using the Tanh activation layer in the electromagnetic transient model to obtain the training condition waveform map.
5. The electromagnetic transient IGBT real-time model construction method according to claim 4, characterized by, The convolutional module includes a first extraction branch and a feature fusion layer. The step of performing residual processing on the data feature map through the convolutional module to obtain a waveform feature map includes: The first extraction branch is used to extract features from the data feature map to obtain a first feature map. The first extraction branch includes a first convolutional layer, a second convolutional layer, a batch normalization layer, a Tanh activation layer and a 1×1 standard convolutional layer connected in sequence. The first feature map and the data feature map are fused by a feature fusion layer to obtain a waveform feature map.
6. The electromagnetic transient IGBT real-time model construction method of claim 1, wherein, The step of constructing the electromagnetic transient IGBT real-time model corresponding to the IGBT to be constructed using the target gate switch delay model and the target electromagnetic transient model includes: The initial electromagnetic transient model is obtained by combining the target gate switch delay model and the target electromagnetic transient model. The initial electromagnetic transient model is loaded into a preset on-chip system to obtain the electromagnetic transient IGBT real-time model corresponding to the IGBT to be constructed.
7. An electromagnetic transient IGBT real-time model construction system, characterized by, include: The preprocessing module is used to acquire the structural data and training data of the IGBT to be constructed, perform data preprocessing on the training data, and generate corresponding training feature data. The first analysis module is used to simplify and solve the preset gate switch delay model using the structural data to obtain the target gate switch delay model. The second analysis module is used to train the preset electromagnetic transient model using the training feature data to generate the target electromagnetic transient model. The construction module is used to construct the electromagnetic transient IGBT real-time model corresponding to the IGBT to be constructed, using the target gate switch delay model and the target electromagnetic transient model.
8. An electronic device, comprising: The system includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, causes the processor to perform the steps of the electromagnetic transient IGBT real-time model construction method as described in any one of claims 1-6.
9. A computer readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed, it implements the method for constructing a real-time electromagnetic transient IGBT model as described in any one of claims 1-6.
10. A computer program product, characterised in that, The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, wherein when the program instructions are executed by a computer, the computer performs the electromagnetic transient IGBT real-time model construction method as described in any one of claims 1-6.