Information-Physical Fusion Identification Model for Vibration Patterns of Unbalanced Transformer Winding

Aiming at identifying the vibration patterns of transformer winding under unbalanced operation, an online identification method based on information-physical fusion is presented in this paper. The feature information in electromagnetic and mechanical subsystem of virtual/physical entities is obtained through digital simulation and physical experiment, respectively. Meanwhile, a global mode-state information database is built. Key parameters such as winding current and vibration acceleration are extracted, then the mathematical mapping relationship of unbalance factor, current, and vibration acceleration is analyzed. Then the fused feature indicator is defined to evaluate the corresponding patterns of sample spaces. On this basis, the Autoformer generalization model is constructed to cluster the mapped pattern subspaces from various sample spaces, which formulates the information-physical fusion identification scheme for winding vibration patterns. Finally, the effectiveness and feasibility of the information-physical fusion identification scheme are validated. Results indicate that the proposed method can effectively reflect the unstable winding vibration patterns in unbalanced modes, which provides a feasible solution for transformer status tracking and monitoring.