Utilization of Least Squares Algorithm for Online Identification of Foster Thermal Network Parameters

Predicting failure and estimating remaining lifetime of power seminconductor (PS) can be carried out using multiple indicators. One of many indicators of PS degradation is an increase of its thermal impedance. This paper proposes a method enabling tracking those changes and updating PS thermal model parameters during real time operation. The presented algorithm is based on ordinary least squares (OLS). Both accuracy, and resiliance to correlation and systematic errors are improved by utilizing initial parameters guess for regularization.