The brushless doubly-fed machine (BDFM), widely employed in marine shaft power generation systems, faces significant operational risks when stator winding inter-turn short-circuit faults occur. However, the unique dual-winding structure of BDFG (brushless doubly-fed generator) complicates fault identification regarding which winding (control or power) is affected and the severity level. This study establishes a short-circuit model for BDFG and investigates characteristic electrical signals under different fault conditions by analyzing electromagnetic signal variations. The research first examines air-gap magnetic density changes during generator operation under short-circuit faults. Subsequently, real-time online detection of faults is achieved through current signal analysis on the power winding side. Finally, Park’s vector modulus analysis is proposed to distinguish fault locations between dual windings and evaluate short-circuit severity. This integrated approach provides a diagnostic framework for identifying fault positions and assessing damage levels in BDFG systems.