Ai-Enhanced Fault-Tolerant Control of Bidirectional Dual Active Bridge Converters Using Wavelet based Neural Classification

This thesis presents a simulation-based fault-tolerant control strategy for a bidirectional Dual Active Bridge (DAB) converter, motivated by the growing demand for reliable and efficiënt power conversion in applications such as electric vehicles, battery energy storage systems, and renewable energy integration. Although DAB converters are widely adopted due to their high efficiency and bidirectional capability, their performance can be significantly degraded by open-circuit switch faults, which may compromise system stability and power continuity. To address this issue, a fault detection and diagnosis framework combining wavelet-based feature extraction and Artificial Neural Network (ANN) classification is proposed. Inductor current signals obtained firom a MATLAB/Simulink model of the DAB converter are decomposed using discrete wavelet transform to extract time-frequency features that characterize both normal and faulty operating conditions. These features are used to train an ANN classifier to distinguish between normal operation and open-circuit faults at individual switches. The ANN training converged within six epochs and achieved a Mean Squared Error (MSE) of 1.43 x 10"15, indicating fast convergence and stable learning behaviour under the simulated conditions. Simulation results demonstrate that the proposed method is capable of detecting open-circuit faults at switches Tl, T2, T5, and T6 within approximately 0.21-0.23 s. Following fault identification, a redundancy-based passive fault-tolerant control strategy is activated, allowing the converter to recover stable operation within approximately 0.28-0.29s. The post-fault current waveforms return to within predefined tolerance bands with reduced oscillations, confirming the effectiveness of the proposed approach in maintaining bidirectional power flow during fault conditions. Overall, the results indicate that the integration of wavelet-based signal processing and ANN-based fault diagnosis can enhance the reliability of DAB converters under open-circuit fault scenarios in a simulated environment.