Incipient fault detection based on Just-in-time-learning and wavelet transform for DAB DC-DC converter

In this paper, a novel fault detection method for incipient fault occurring in Dual-active-bridge (DAB) DC-DC converter is proposed. The incipient fault has the characteristics of small amplitude, weak influence on system performances and easily being covered by noise, hence it is difficult to detect incipient fault for DAB DC-DC converter. This work proposes to use wavelet transform to process data, which can amplify the fault feature and simultaneously weak the noise. And then Just-in-time-learning (JITL) algorithm is utilized to model the system dynamic process online, followed by singular-value-decomposition (SVD) and covariance calculation. The threshold is obtained when applying the proposed method to the normal data, and then to the test data in order to get its covariance, which when is larger than the threshold, an alarm will be triggered to imply the fault occurrence. Finally, establish a simulation model in the Matlab environment, and analyze the experimental results to verify the effectiveness of the proposed method.