An Online Junction Temperature Estimating Method for SiC MOSFETs Based on Steady-State Features and GPR

The thermal sensitivity electrical parameter (TSEP) method has gained popularity for online junction temperature (T-j) estimation to enhance the operational reliability of SiC mosfets. However, the performance of existing TSEP methods is affected by varying operating conditions. Achieving a balance between T-j estimation accuracy and cost remains a challenge. To address these issues, this article proposes an online T-j estimation method. First, the on-state voltage and on-state body diode voltage drop are combined as features for T-j estimation. The selection of these two features with different sensitivities under various load current cases improves the accuracy of T-j estimation and enhances the overall sensitivity. Second, Gaussian Process Regression is employed to eliminate the effect of load current from the T-j estimation model, ensuring robustness to operating conditions. Finally, an online T-j estimation strategy based on these innovations is proposed and its effectiveness is validated through multiple experiments in a dc-dc converter under various operating conditions. Compared to conventional methods, the proposed approach demonstrates higher accuracy and stronger robustness against operating conditions.