Online Observation for IGBT Module Loss and Spatial Temperature with Aging Tracking: A Data-driven Method

The loss and heat flux of IGBT module change with package degradation, resulting in the variation of case temperature and its distribution. Inspired by the aging mechanism of module, this study performs an online observation for IGBT's time-varying profile by collecting temperature of multiple points in the bottom case, which achieves a spatial temperature reconstruction beyond junction temperature with aging track capability. Moreover, the loss can also be obtained by the thermal performance at steady state. A deep neural network (DNN) based data-driven model is established by collecting data in varied degradation level and operating condition, which favors the online observation by its low computation cost. A dedicated optical fiber positioning is proposed to facilitate temperature acquisition, and the process of data-driven modeling is illustrated in detail. Results show that the spatial temperature inside the module and the power loss can be accurately reconstructed within 5% error, and the module degradation status are well tracked.