CN-121997121-A - IGBT health state high-robustness diagnosis method, system, medium and equipment

CN121997121ACN 121997121 ACN121997121 ACN 121997121ACN-121997121-A

Abstract

A high-robustness intelligent diagnosis method, system, medium and equipment for IGBT health state based on shell temperature distribution features comprises the steps of arranging temperature sensors on the outer surfaces of substrates under a plurality of chips of an IGBT module, collecting shell temperature data of each measuring point, establishing a three-dimensional finite element thermal-force coupling model of the IGBT module, simulating heat transfer performance changes of a substrate solder layer in different degradation states, generating a shell temperature simulation data set containing health, early degradation, serious degradation and failure states, constructing an inverter experiment platform, collecting shell temperature data under real working conditions, constructing an experiment test data set after Kalman filtering and sliding window mean processing, training a convolutional neural network model based on the simulation data set, inputting the three two-dimensional convolutional layers and the full connecting layer into a normalized diagnosis temperature matrix, and outputting the normalized diagnosis temperature matrix to realize intelligent classification of the type and degree of non-uniform degradation of the solder layer.

Inventors

ZHU LINGYU
CHEN MIN
TANG YIZHENG
Zhao Ku
WANG WEICHENG
XU HONGHAI
JI XUEJUN
CHENG YUELIANG
TANG HAINING
ZHANG TAO

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 intelligent diagnosis method for high robustness of the IGBT health state based on the shell temperature distribution characteristics is characterized by comprising the following steps of: arranging temperature sensors on the outer surfaces of the substrates right below the chips of the IGBT module, collecting shell temperature data of each measuring point, and constructing a shell Wen Juzhen for representing the shell temperature space distribution characteristics of the IGBT module; The shell Wen Juzhen is preprocessed, the temperature of each measuring point is subtracted by the ambient temperature to obtain a temperature rise matrix, then the temperature rise value of the corresponding parallel branch is subjected to cross difference according to the parallel half-bridge structure in the IGBT module, and divided by the sum of the temperature rises of the corresponding branches to construct a normalized diagnosis temperature matrix so as to eliminate the influence of the ambient temperature and the power loss fluctuation; Establishing a three-dimensional finite element thermal-force coupling model of the IGBT module, simulating the heat transfer performance change of the substrate solder layer under different degradation states, and generating a shell temperature simulation data set containing health, early degradation, serious degradation and failure states; building an inverter experiment platform, collecting shell temperature data under a real working condition, and constructing an experiment test data set after Kalman filtering and sliding window mean value processing; based on the simulation data set training convolutional neural network model, the network comprises at least three two-dimensional convolutional layers and a full-connection layer, the input is a normalized diagnosis temperature matrix, and the output is an IGBT health state type to realize intelligent classification of the type and degree of uneven degradation of the solder layer.
2. The intelligent diagnosis method for high robustness of IGBT health status based on shell temperature distribution characteristics according to claim 1, wherein preferably, the dimension of the shell Wen Juzhen corresponds to the number and layout of the power chips in the IGBT module, and the measuring point is located at the substrate position right below the IGBT chip or the FWD chip.
3. The intelligent diagnosis method for high robustness of IGBT health status based on shell temperature distribution characteristics according to claim 1, wherein the construction mode of the normalized diagnosis temperature matrix is as follows: A. B, C, D sets of measuring points respectively correspond to the sum of the temperature rises of the low-side IGBT, the low-side FWD, the high-side FWD and the high-side IGBT, and the relative temperature difference between adjacent branches is calculated: ; Where Δta is equal to the sum of Δt1, Δt3 and Δt5, Δtb is equal to the sum of Δt2, Δt4 and Δt6, Δtc is equal to the sum of Δt7, Δt9 and Δt11, and Δtd is equal to the sum of Δt8, Δt10 and Δt12.
4. The intelligent diagnosis method for high robustness of IGBT health status based on shell temperature distribution characteristics of claim 1, wherein the substrate solder layer degradation status comprises health status, early degradation with area reduced by 20% -35%, serious degradation with area reduced by 35% -50%, and failure status with area reduced by at least 50%, and the degradation is unevenly distributed.
5. The intelligent diagnosis method for high robustness of IGBT health status based on shell temperature distribution characteristics according to claim 1, wherein temperature data is processed by Kalman filtering, and noise is suppressed by adopting a sliding window averaging mode.
6. The intelligent diagnosis method for high robustness of IGBT health status based on shell temperature distribution characteristics according to claim 1, wherein the convolutional neural network does not comprise a pooling layer, an input layer receives a normalized temperature matrix with a size of 2 x 2 or 3 x 4, spatial characteristics are extracted through a plurality of two-dimensional convolutional layers, and finally a health status classification result is output through a full connection layer.
7. The intelligent diagnosis method for high robustness of IGBT health status based on shell temperature distribution characteristics according to claim 1, wherein random noise within a range of + -0.01 ℃ to + -0.1 ℃ is superimposed in simulation data to simulate actual measurement errors.
8. A system for carrying out the method of any one of claims 1-7, comprising: A temperature sensing unit for arranging a plurality of temperature sensors at the substrate right below the IGBT chip; The data acquisition and preprocessing unit is used for acquiring shell temperature data and executing environment temperature compensation and normalization differential processing; The finite element simulation module is used for constructing an IGBT model and generating a multi-state shell temperature data set with noise; the convolutional neural network diagnosis model is used for receiving the normalized temperature matrix and outputting a health state classification result; and the display and alarm unit is used for visualizing the diagnosis result and giving an early warning when the degradation is detected.
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 health state high-robustness diagnosis method, system, medium and equipment Technical Field The invention relates to the technical field of IGBT module detection, in particular to an IGBT health state high-robustness intelligent diagnosis method, system, medium and equipment based on shell temperature distribution characteristics. Background Wind power generation has been attracting attention as one of the most dominant forms of new energy. In a wind power generation system, a wind power converter plays a core role of connecting a wind power generator with a power grid, and is also an integral part of the whole wind power generation system. The core component of the wind power converter is an IGBT power module. However, IGBTs are prone to degradation, failure under unstable operating conditions, thereby causing failure of the entire system. However, the IGBT generally operates in a full package, and the internal degradation thereof is difficult to directly detect, so that the critical state parameter is proposed to be the key for diagnosing the health state of the IGBT. 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 an IGBT health state high-robustness intelligent diagnosis method, system, medium and equipment based on shell temperature distribution characteristics, wherein the shell temperature distribution characteristics are used as state parameters, relative change values are obtained through normalization processing, errors caused by working condition changes are reduced, and finally the neural network is utilized to realize the health state high-robustness intelligent diagnosis. The intelligent diagnosis method for high robustness of the IGBT health state based on the shell temperature distribution characteristics comprises the following steps: arranging temperature sensors on the outer surfaces of the substrates right below the chips of the IGBT module, collecting shell temperature data of each measuring point, and constructing a shell Wen Juzhen for representing the shell temperature space distribution characteristics of the IGBT module; The shell Wen Juzhen is preprocessed, the temperature of each measuring point is subtracted by the ambient temperature to obtain a temperature rise matrix, then the temperature rise value of the corresponding parallel branch is subjected to cross difference according to the parallel half-bridge structure in the IGBT module, and divided by the sum of the temperature rises of the corresponding branches to construct a normalized diagnosis temperature matrix so as to eliminate the influence of the ambient temperature and the power loss fluctuation; Establishing a three-dimensional finite element thermal-force coupling model of the IGBT module, simulating the heat transfer performance change of the substrate solder layer under different degradation states, and generating a shell temperature simulation data set containing health, early degradation, serious degradation and failure states; building an inverter experiment platform, collecting shell temperature data under a real working condition, and constructing an experiment test data set after Kalman filtering and sliding window mean value processing; based on the simulation data set training convolutional neural network model, the network comprises at least three two-dimensional convolutional layers and a full-connection layer, the input is a normalized diagnosis temperature matrix, and the output is an IGBT health state type to realize intelligent classification of the type and degree of uneven degradation of the solder layer. In the intelligent diagnosis method for high robustness of the IGBT health state based on the shell temperature distribution characteristics, the dimension of the shell Wen Juzhen corresponds to the number and layout of the power chips in the IGBT module, and the measuring points are positioned at the positions of the substrate right below the IGBT chips or the FWD chips. In the intelligent diagnosis method for high robustness of the IGBT health state based on the shell temperature distribution characteristics, the construction mode of the normalized diagnosis temperature matrix is as follows, A, B, C, D four groups of measuring points respectively correspond to the sum of the temperature rises of the low-side IGBT, the low-side FWD, the high-side FWD and the high-side IGBT, and the relative temperature difference between adjacent branches is calculated: Where Δta is equal to the sum of Δt1, Δt3 and Δt5, Δtb is equal to the sum of Δt2, Δt4 and Δt6, Δtc is equal to the sum of Δt7, Δt9 and Δt11, and Δtd is equal to the sum of Δt8, Δt10 and Δt12. In the high-robustness intelligent diagnosis method for the IGBT health state based on