Online identification of thermal parameters in automotive inverter power modules under dynamic operation

As automotive power electronic systems advance toward higher power density and reliability, accurate thermal monitoring of power modules has become increasingly critical. To address the limitations of existing monitoring methods, including parameter drift, strong data dependence, and limited applicability to complex topologies, this paper proposes an online thermal-parameter identification method for automotive inverter power modules. A thermal model of an automotive SiC power module is established, with thermal resistance used as a key monitoring parameter. An online junction temperature measurement method based on the body-diode forward voltage (V_F) is employed to ensure accurate temperature acquisition. Within a unified modeling framework, the Extended Kalman Filter (EKF), Particle Swarm Optimization (PSO), and Bayesian Optimization (BO) are applied and comparatively analyzed to identify thermal model parameters under dynamic and repetitive thermal loading conditions. Experimental results obtained from a three-phase, six-switch automotive inverter platform demonstrate that the proposed method can effectively identify thermal-parameter variations and maintain stable convergence performance.