Abstract: Junction temperature is a key determinant of power semiconductor reliability and lifetime. Five approaches to online junction-temperature monitoring of insulated-gate bipolar transistor (IGBT) power modules are reviewed: temperature-sensitive electrical parameter (TSEP) methods, monolithically integrated temperature sensing, thermal modeling, fiber-optic sensing, and data-driven methods. Their principles, applications, accuracy factors, and engineering constraints are compared. TSEP methods are easy to integrate but require calibration and operating-condition compensation. Monolithic sensors measure temperature near local junction regions and support degradation monitoring, although chip area and packaging interfaces remain constraints. Thermal models suit embedded control but are sensitive to parameter drift and boundary conditions. Fiber-optic sensors provide temperature information near chip hot spots but face packaging and cost challenges. Data-driven methods support multisource fusion and health assessment, but require broader datasets, better interpretability, and long-term maintenance. Physics-informed modeling, online calibration, distributed multipoint sensing, and integration with lifetime prediction are key directions for improving engineering reliability.