Abstract:

The junction temperature estimation enables online health management and performance optimisation of SiC power modules. A multitude of temperature sensitive electrical parameters (TSEPs) is gaining increased attention due to their non-invasive nature for online measurements, but they are conventionally focused on certain individual TSEPs, which lack accuracy and robustness given the dynamic variations of operational conditions. This work proposed a fusion method by combining theprincipal component analysis (PCA) and multiple linear regression (MLR) to estimate the junction temperature from a TSEP data set. The presented fusion method predicts temperature with better accuracy and higher noise immunisation compared with the single TSEP estimation method.