Degradation State Assessment of IGBT Module Based on Interpretable LSTM-AE Modeling Under Changing Working Conditions

Researches indicate that the Insulated Gate Bipolar Transistor (IGBT) module is a crucial yet failure-prone constituent of power electronic devices. As a result, assessing the degradation state of IGBT module is significant to guarantee safe device operation. However, existing studies are usually conducted under constant working conditions, which is inconsistent with the strong nonlinearity of the degradation process and strong coupling of parameters of the IGBT module caused by changing working conditions. Additionally, the designed models often lack interpretability, making it less credible. Some of them may even need to be processed with the full life data after the degradation state is obtained, which does not satisfy the requirements of online applications. Therefore, this paper proposes an interpretable LSTM-AE modeling method for degradation state assessment of IGBT modules under changing working conditions. Firstly, the interpretability of the model is improved by designing a contrastive output structure. Then, by analyzing the characteristics of IGBT module degradation process, the properties of the degradation are summarized, and the prior knowledge is integrated into the model as a constraint to enhance the accuracy of the degradation state assessment and further improve the interpretability. Afterwards, a variable working condition power cycle experiment system is designed and constructed to collect degradation data. Finally, the performance of the proposed method is verified. The results show that the degradation state extracted by the proposed method are more consistent with the theoretical analysis results, which can accurately assess the degradation state of IGBT modules.