CN-121997643-A - IGBT temperature field calculation method, system, medium and equipment

CN121997643ACN 121997643 ACN121997643 ACN 121997643ACN-121997643-A

Abstract

A method for calculating IGBT temperature field based on data driving includes setting up finite element thermal model based on geometry structure and material thermal parameters of IGBT module, utilizing finite element thermal model to simulate steady state temperature field distribution under each working condition, setting up multi-working condition training data set containing input variable and corresponding temperature field output, setting up main network-branch network nerve network model, using multi-working condition training data set to train main network-branch network model, inputting power loss value and heat dissipation parameter measured in real time into IGBT temperature field prediction model to obtain initial temperature field prediction result, updating power loss value according to initial temperature field prediction result, inputting updated power loss into IGBT temperature field prediction model again to make new round of temperature field prediction, repeating execution until power loss difference between two adjacent iterations is smaller than preset convergence threshold, stopping iteration and outputting final steady state temperature field.

Inventors

ZHU LINGYU
TANG YIZHENG
WANG WEICHENG
XU HONGHAI
JI XUEJUN
CHENG YUELIANG
TANG HAINING
ZHANG TAO
HUANG YUANPING

Assignees

西安交通大学
国电南瑞科技股份有限公司
国网浙江省电力有限公司电力科学研究院

Dates

Publication Date: 20260508
Application Date: 20251231

Claims (10)

1. The IGBT temperature field calculation method based on data driving is characterized by comprising the following steps of: step 1, establishing a finite element thermal model based on the geometry structure and material thermal parameters of an IGBT module; step 2, taking the power loss distribution of the IGBT chip and the FWD chip as a first input variable, taking the flow speed and the temperature of a cooling medium as a second input variable, and generating a plurality of working condition combinations in a preset parameter range through Latin hypercube sampling; Step 3, simulating steady-state temperature field distribution under each working condition by using the finite element thermal model, and constructing a multi-working condition training data set comprising input variables and corresponding temperature field outputs; Step 4, constructing a main network-branch network neural network model, wherein the main network adopts a Fourier neural operator network to process the spatial distribution characteristics of power loss, and the branch network adopts a fully-connected neural network to process the heat dissipation boundary conditions; Step 5, training the main network-branch network neural network model by using the multi-working condition training data set, and optimizing network weights by adopting Sobolev H1 loss functions to obtain a trained IGBT temperature field prediction model; Step 6, obtaining load current, switching frequency and bus voltage of an IGBT module under actual running conditions, and calculating power loss values of an initial IGBT and an FWD chip by combining a pre-calibrated power loss model; And 7, updating a power loss value according to the initial temperature field prediction result, re-inputting the updated power loss into the IGBT temperature field prediction model to perform new temperature field prediction, and repeatedly executing until the power loss difference value between two adjacent iterations is smaller than a preset convergence threshold value, stopping the iteration and outputting a final steady-state temperature field.
2. The method for calculating the IGBT temperature field based on the data driving according to claim 1 is characterized in that preferably, the main network maps the positions of the IGBT and the FWD chip and the corresponding power loss thereof into a two-dimensional power loss diagram, scalar power of each grid point is converted into a high-dimensional feature vector through a dimension lifting layer, convolution operation is carried out in a frequency domain through multi-layer Fourier transform and inverse transform operation, space-frequency features are extracted by combining residual connection with a nonlinear activation function, and finally a feature matrix with fixed dimension is output through a projection layer.
3. The method for calculating the IGBT temperature field based on data driving according to claim 1, wherein the branch network is composed of three layers of fully-connected neural networks and is used for encoding the flow speed and the temperature of the cooling medium into a group of high-dimensional feature vectors as the modulation weights of subsequent feature fusion.
4. The method for calculating the IGBT temperature field based on the data driving according to claim 1, wherein in the step 4, the feature fusion is to multiply a feature matrix output by a main network with a feature vector output by a branch network element by element, and then the feature matrix and the feature vector are processed by a convolution layer and a nonlinear activation function, and finally a single-channel temperature field image is output.
5. The method for calculating the IGBT temperature field based on the data driving according to claim 1, wherein in the step 1, the temperature distribution of the surface of the IGBT module is measured by a thermal infrared imager and is compared with a finite element simulation result, and if the maximum temperature difference is not more than 1 ℃, the finite element model is judged to be effective.
6. The method for calculating the IGBT temperature field based on data driving according to claim 1, wherein the power loss model is an analytical or table look-up model obtained by fitting actual measurement loss data under various working conditions based on IGBT device manual parameters, and calculates conduction loss and switching loss according to load current, bus voltage and switching frequency.
7. The data-driven IGBT temperature field calculation method according to claim 1, wherein the single reasoning time of the IGBT temperature field prediction model is less than 0.2 seconds, and the maximum error of temperature field prediction is less than 2 ℃.
8. A system for carrying out the method of any one of claims 1-7, comprising: the data acquisition unit is used for acquiring the running parameters and the heat dissipation conditions of the IGBT module; the power loss calculation unit is used for calling a power loss model to calculate the chip loss based on the operation parameters; the neural network reasoning unit integrates the trained main network-branch network neural network model, and is used for receiving the power loss and the heat dissipation parameters and outputting a temperature field; the iteration control unit is used for judging whether the electrothermal iteration converges or not and outputting a final temperature field result; And the display and storage unit is used for visualizing the temperature field distribution and storing historical data.
9. A computer storage medium comprising computer instructions which, when run on a computer, cause the computer to perform the method of any of claims 1-7.
10. An electronic device, the electronic device comprising: A memory, a processor, and a computer program stored on the memory and executable on the processor, wherein, The processor, when executing the program, implements the method of any one of claims 1-7.

Description

IGBT temperature field calculation method, system, medium and equipment Technical Field The invention relates to the technical field of IGBT module detection, in particular to a data-driven IGBT temperature field calculation method, a data-driven IGBT temperature field calculation system, a data-driven IGBT temperature field calculation medium and data-driven IGBT temperature field calculation equipment. Background IGBTs have been widely used in the fields of electric automobiles, rail transit, renewable energy systems, and high-voltage direct-current transmission. However, the reliability of IGBTs is often limited by thermal failure, which accounts for more than half of the device failures. Efficient and accurate measurement of the reconstruction of critical temperature profiles with limited sensing points is an urgent issue to be addressed, which is critical to health management and life prediction. The above information disclosed in the background section is only for enhancement of understanding of the background of the invention and therefore may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Disclosure of Invention The invention provides a data-driven IGBT temperature field calculation method, a data-driven IGBT temperature field calculation system, a data-driven IGBT temperature field calculation medium and data-driven IGBT temperature field calculation equipment, which are used for efficiently and accurately calculating temperature field distribution under various working conditions, and solving the contradiction between efficiency and accuracy in the temperature field calculation method based on finite elements. The IGBT temperature field calculation method based on data driving comprises the following steps: step 1, establishing a finite element thermal model based on the geometry structure and material thermal parameters of an IGBT module; step 2, taking the power loss distribution of the IGBT chip and the FWD chip as a first input variable, taking the flow speed and the temperature of a cooling medium as a second input variable, and generating a plurality of working condition combinations in a preset parameter range through Latin hypercube sampling; Step 3, simulating steady-state temperature field distribution under each working condition by using the finite element thermal model, and constructing a multi-working condition training data set comprising input variables and corresponding temperature field outputs; Step 4, constructing a main network-branch network neural network model, wherein the main network adopts a Fourier neural operator network to process the spatial distribution characteristics of power loss, and the branch network adopts a fully-connected neural network to process the heat dissipation boundary conditions; Step 5, training the main network-branch network neural network model by using the multi-working condition training data set, and optimizing network weights by adopting Sobolev H1 loss functions to obtain a trained IGBT temperature field prediction model; Step 6, obtaining load current, switching frequency and bus voltage of an IGBT module under actual running conditions, and calculating power loss values of an initial IGBT and an FWD chip by combining a pre-calibrated power loss model; And 7, updating a power loss value according to the initial temperature field prediction result, re-inputting the updated power loss into the IGBT temperature field prediction model to perform new temperature field prediction, and repeatedly executing until the power loss difference value between two adjacent iterations is smaller than a preset convergence threshold value, stopping the iteration and outputting a final steady-state temperature field. In the data-drive-based IGBT temperature field calculation method, the main network maps the positions of IGBT and FWD chips and the corresponding power loss thereof into a two-dimensional power loss diagram, scalar power of each grid point is converted into a high-dimensional feature vector through a dimension increasing layer, convolution operation is carried out in a frequency domain through multi-layer Fourier transform and inverse transform operation, residual connection and nonlinear activation function are combined to extract space-frequency features, and finally a feature matrix with fixed dimension is output through a projection layer. In the data-driven IGBT temperature field calculation method, the branch network is formed by three layers of fully-connected neural networks and is used for encoding the flow speed and the temperature of a cooling medium into a group of high-dimensional feature vectors serving as the modulation weight for the subsequent feature fusion. In the data-driven IGBT temperature field calculation method, in the step 4, the feature fusion is to multiply a feature matrix output by a main network and a feature vector output by a branch network el