Support Vector Regressor-Based SiC MOSFET Junction Temperature Prediction Based on Temperature Sensitive Electrical Parameters and External Dependencies

Temperature sensitive electrical parameters (TSEPs) have been used as a set of indicators to achieve online and non-intrusive junction temperature (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$T_{j}$</tex>) monitoring for SiC MOSFETs. However, TSEPs are influenced by both the junction temperature and the operating conditions of the device where prediction accuracy can be impacted if not carefully accounted for. This paper develops a Support Vector Regressor (SVR)-based junction temperature prediction method considering the external dependencies of TSEPs. Such external dependencies are statistically quantified through correlation analysis, providing guidelines for feature selection in the SVR model. For instance, results from correlation analysis suggest that while being influenced by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$T_{j}$</tex>, the on-state resistance (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R_{D S, o n}$</tex>) is also impacted by the device drain current and gate driving voltage. Incorporating these external dependencies into the input feature set significantly improves model accuracy, with the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^{2}$</tex> score increasing from 0.918 to 0.995 when <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R_{D S, \text{on}}$</tex> is selected as the input TSEP feature. This highlights the suitability of the prediction method for scenarios where SiC MOSFETs are operated under varying conditions.