Fault diagnosis of grid-connected inverters using a fractional-order predefined-time sliding mode observer

• A fractional-order predefined-time observer ensures current estimation within a fixed time. • The proposed observer achieves faster convergence and reduces chattering effectively. • A novel fault index improves sensitivity and minimizes false alarms under disturbances. • A lightweight fault localization method identifies switching faults via current analysis. • The scheme is validated on a three-phase inverter under diverse fault and disturbance cases. Fault detection in grid-connected inverters is often hindered by the slow convergence rate of conventional observers and their limited robustness against disturbances. To overcome these limitations, this study proposes a fractional-order predefined-time sliding mode observer (FOPTSMO) that enables rapid convergence and improved disturbance rejection. A dynamic model of the inverter system is first established to support observer design. The proposed FOPTSMO ensures that the estimated current converges to the actual current within a predefined time, with theoretical proof of convergence and stability. Compared to existing predefined-time observers, it achieves faster convergence while effectively mitigating chattering effects. To enhance fault detection accuracy, a novel index is introduced based on the continuous integration of current residuals, improving sensitivity to fault occurrences. In addition, a lightweight fault localization strategy is developed by analyzing output current characteristics under various switching fault scenarios. The proposed approach is validated through case studies conducted on a Simulink-modeled three-phase inverter under four representative conditions: single-switch and double-switch open-circuit faults, grid voltage fluctuations, and inductance parameter imbalances. Simulation results confirm that the proposed method significantly improves detection speed, fault localization accuracy, and robustness under various disturbance conditions.