Adaptive Weighted Health Index Construction Based Rolling Health Status Assessment of Power Converter Systems

ABSTRACT Accurate and timely health status assessment of power converter systems is crucial for ensuring the reliability and safety of power equipment. Conventional health assessment methods for power converters often rely on static models or fixed‐weight Health Index (HI), which lack adaptability to evolving degradation patterns and fail to prioritize recent operational data, limiting prediction accuracy and timeliness. In this study, a rolling prediction framework is proposed for health status assessment of key components in power converter systems, which is built upon an adaptively weighted HI and rolling Support Vector Regression (SVR). First, the HI is constructed from multiple degradation‐related features, where an inverse standard deviation weighting scheme is applied to dynamically capture the relative contribution of each feature, yielding an adaptive and interpretable HI. Then, a rolling prediction mechanism is introduced using an SVR model to characterize the nonlinear relationship between raw features and the HI. In this framework, the training set is continuously updated through a sliding time window, while exponentially decaying weights are applied to emphasize more recent data. Finally, two experiments on circuit breakers and Insulated‐Gate Bipolar Transistors (IGBT) are conducted to demonstrate the effectiveness of the proposed method.