This paper investigates anomaly diagnosis for grid-tied three-phase inverters in cyber–physical smart grids, with an emphasis on distinguishing physical IGBT open-circuit physical faults from anomalies induced by denial-of-service (DoS) cyber-attacks. A super-twisting-based second-order interval sliding-mode observer is developed to estimate three-phase currents with bounded errors in the presence of uncertainties and disturbances. Based on analytical residual relationships, fault localization is achieved using the residual sign pattern and magnitude ratios for single-switch and same-leg double-switch open-circuit faults. In contrast, DoS-induced anomalies primarily manifest as effective current attenuation without deterministic residual sign or ratio patterns, enabling fault-type discrimination. Simulation results demonstrate that the proposed method achieves reliable anomaly diagnosis within one fundamental cycle, without requiring additional sensors or training data.
