A Junction Temperature Monitoring Method with Miller Plateau based on Principal Component Analysis and Artificial Neural Networks

Insulated gate bipolar transistors (IGBTs) are widely used in power electronic converters, while the influence of junction temperature on IGBT's reliability cannot be ignored. Accurate and stable junction temperature monitoring methods are crucial, with temperature sensitive electrical parameter (TSEP)-based methods being most commonly used. Many TSEPs are affected by conduction current, requiring recalibration when current changes. A junction temperature monitoring method based on principal component analysis (PCA) and artificial neural networks (ANN) is proposed in this paper, using Miller plateau waveform as the TSEP, which is experimentally verified in a double-pulse platform. PCA is used for dimensionality reduction and feature extraction, while ANN establishes the relationship between junction temperature and Miller plateau waveform, reducing the weight of conduction current-related factors. This allows the method to be independent of conduction current. Finally, the proposed method is compared with a method based on Miller plateau voltage, demonstrating its accuracy and independence.