Efficient Junction Temperature Estimation of SiC Power Modules Based on Temperature-Dependent Lumped Thermal Model

Abstract

Silicon carbide (SiC) power modules exhibit superior performance at high temperatures compared to silicon counterparts, and their thermal performance at such high temperature is significantly influenced by the properties of temperature-dependent materials. A junction temperature estimation based on the electrothermal coupling effect becomes significantly inefficient due to step-by-step updates of the temperature-dependent thermal parameters in iteration calculation. Thus, this article proposes an efficient estimation approach to estimate the junction temperature of multichip SiC power modules. A 3-D lumped thermal model (LTM) is developed, incorporating temperature-dependent thermal parameters in its nonlinear state-space equations. Dynamic thermal curves from finite element (FE) simulation are utilized to accurately identify these nonlinear thermal parameters via an adaptive particle swarm optimization (APSO) algorithm. In particular, the nonlinear state-space equations are effectively solved by the trapezoidal rule-backward differentiation formula 2 (TR-BDF2) method, which implements calculations in two stages between the trapezoidal rule (TR) and backward differentiation formula (BDF2), leading to enhanced stability and a significant reduction in computation time. The proposed method achieves a computational speed of 1948 times faster than the conventional Runge-Kutta (R-K) method. The computational errors are within approximately $1~^{\circ }$ C, experimentally confirming that the proposed approach is superior in the efficient and accurate estimation of junction temperature at high temperatures.