A Confidence-Scheduled Hybrid Method for DC-Bias Estimation and Suppression in Bidirectional Full-Bridge LLC Converters During Reverse Power Transfer

DC-bias may accumulate in bidirectional full-bridge LLC converters during reverse power transfer because the magnetizing branch lacks an inherent DC-blocking mechanism. This bias may cause asymmetric flux excitation in the transformer core, thereby increasing magnetic stress and even leading to core saturation. To address this issue, a confidence-scheduled hybrid DC-bias estimation and suppression method is proposed. An integration-based indicator is constructed for sensitive weak-bias detection, while a reduced-order extended Kalman filter (EKF) is introduced to improve noise immunity and dynamic tracking under strong-bias conditions. Moreover, a confidence-scheduling mechanism is developed to adaptively fuse the two estimates according to bias severity. Based on the fused estimate, a two-level suppression strategy is implemented for severe- and weak-bias conditions. Simulations and experiments on a 2 kW prototype verify that the proposed strategy achieves fast detection, highly accurate robust estimation with a steady-state error of less than 2%, and effective suppression over a wide operating range without additional bulky DC-blocking hardware.