Multiscale Park's Vector Approach for Identifying Early Stator Fault Severity Using MCSA in DFIGs

Early fault detection of Doubly-fed Induction Generators (DFIG) is a key problem for reliable operation of wind power generation. Motor current signature analysis (MCSA) is the most reliable and widely used technique, gaining favor because it is non-invasive. In the initial stage of wind generator fault, the fault characteristic information is weak and easy to be covered by noise and interference signals, which makes early fault detection difficult. In this manuscript, a Multi-scale Park's Vector Approach for early fault severity identification of DFIGs stator based on MCSA is proposed, and features of early weak faults are extracted and enhanced by multi-scale method. In this method, the three-phase current of the generator stator is converted into orthogonal two-phase current by Park's Vector Transform, and then the two-phase current is computed in multi-scale layers to obtain park's current vector trajectories of different scales. Curve fitting is performed for the current vector trajectories of different levels and scales. The stator faults are identified and the severity of the faults is quantitatively evaluated according to the spatial trajectory change of the fitted curve. The proposed method is verified by the experimental data of healthy, 2, 7 and 15-turn stator short-circuit faults of a doubly-fed induction generator. The results show that the Multi-scale Park's Vector Method can not only detect the stator inter-turn fault, but also estimate the severity of the fault, and realize the enhancement of the early weak fault characteristics.