Online Junction Temperature Estimation System for IGBT Based on BP Neural Network

Abstract

IGBT module is widely used in industry. The life and reliability estimation of IGBT module are the focus of recent research in this field. IGBT module’s life and reliability are closely related to its junction temperature. Therefore, the prediction of IGBT junction temperature becomes essential of research. In this paper, an online IGBT junction temperature extraction system based on BP neural network is presented. The VCE(SAT) and turn-off delay time of thermoelectric parameters related to junction temperature are extracted online by the system, and the corresponding relationship between these two thermoelectric parameters and junction temperature is obtained using BP neural network algorithm, and the system is implemented with FPGA and stm32. Experimental results show that the system can predict junction temperature with the error less than 2 centigrade.