A High Precision On-Line Detection Method for IGBT Junction Temperature Based on Stepwise Regression Algorithm

• Published in: IEEE access 2020, Vol.8, p.186172-186180
• Main Authors: Shao, Lingfeng, Hu, Yi, Xu, Guoqing
• Format: Article
• Language: English
• Subjects: Algorithms; Capacitance; Capacitors; condition monitoring; Converters; Electrodes; Electronic systems; Energy conversion; insulated gate bipolar transistor; Insulated gate bipolar transistors; Junctions; Life prediction; Logic gates; on-line junction temperature extraction; Parameter sensitivity; Regression models; reliability; Reliability analysis; Semiconductor devices; stepwise regression; Temperature measurement; Temperature sensitive electrical parameters; Temperature sensors
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2020.3028904

The insulated gate bipolar transistor (IGBT), one of the most vulnerable component, is one of the most precious central component in the converter interior. High junction temperature will lead to device failure, which is the main reason of failure of power electronic system. Therefore, on-line high precision measurement of IGBT module junction temperature is the basis of life prediction and reliability evaluation of high-power power conversion equipment. In this paper, the principle of IGBT junction temperature extraction and the latest development of related technologies are summarized. In particular, the working principle and shortcomings of temperature sensitive electrical parameter (TSEP) method are summarized. The change of junction temperature will affect the inter-electrode capacitance in the internal structure of IGBT, which will cause the change of temperature sensitive electrical parameters. The single temperature sensitive electrical parameter method is easily affected by IGBT structure and inter-electrode capacitance. This paper presents an algorithm for high precision on-line detection of IGBT junction temperature. The parameter types are optimized by stepwise regression and the model is established accordingly. In this paper, IGBT: FF50R12RT4 is used as the experimental equipment. By comparing the junction temperature model established based on multiple linear stepwise regression algorithm with the junction temperature model based on traditional temperature sensitive electrical parameters, it is proved that the algorithm has better fitting degree and precision, and the algorithm can be used for high precision online extraction of junction temperature.