Online Fault Diagnosis of Dual Active Bridge Based on R-RVFL Neural Network

Dual active bridge (DAB) converters are key components in modern power electronic systems. However, they are vulnerable to open-circuit faults (OCFs) in their insulated gate bipolar transistors (IGBTs) under high-frequency switching conditions, which can significantly impair the operational reliability. Although data-driven intelligent diagnostic methods have demonstrated improved diagnostic capabilities in recent years, they still encounter critical challenges, such as high computational complexity and low training efficiency. To achieve rapid and accurate detection of OCFs in DAB converters, this article proposes an online fault diagnosis method based on regularized random vector functional link (R-RVFL) neural networks. Specifically, multiple electrical signals are first processed through Gaussian filtering and then fused using a multifeature strategy to construct informative feature sets for fault classification. Subsequently, the fast-learning algorithm of the R-RVFL neural network is utilized to extract a nonlinear mapping between selected features and fault patterns. To balance the testing accuracy and computational time of the diagnostic algorithm, R-RVFL parameters are adjusted to reduce the computational burden while regularization coefficients are fine-tuned to enhance generalization and mitigate overfitting. Comprehensive comparisons with state-of-the-art techniques demonstrate the superiority of the proposed approach in terms of diagnostic accuracy, computational efficiency, and response time.