Physics-Informed SiC Power Module Thermal Model

Accurate beyond 150 °C

Integrating physics-informed logic with transfer learning, this model maintains high thermal prediction accuracy under elevated temperatures and varying switching frequencies, providing a new tool for reliability design and real-time monitoring of SiC power modules.

Yizheng, Tang · Zhan, Cao · Zhu, Lingyu · Sun, Hao · Weicheng, Wang · Zhili, Li · Huo, Wei · Ji, Shengchang

IEEE Transactions on Power Electronics 2025

Advantages

Accurately describes nonlinear temperature-dependent effects above 150 °C

Achieves excellent generalization across operating frequencies via few-shot transfer learning

Offers interpretability and generalization compared to traditional ITM