Transformer Winding Looseness Fault Diagnosis Based on CEEMDAN-MPE and ISSA-KELM

Transformer winding vibration can easily cause winding looseness and deformation, thereby affecting the safe and stable operation of equipment and even the power system. To accurately diagnose the mechanical state changes of transformer windings, this study proposes a transformer winding looseness fault diagnosis method based on complete ensemble empirical mode decomposition with adaptive noise multi-scale permutation entropy (CEEMDAN-MPE) and improved sparrow search algorithm optimized kernel extreme learning machine (ISSA-KELM). Firstly, the transformer vibration signal is adaptively decomposed into multiple intrinsic mode components (IMF) using CEEMDAN. Secondly, the correlation coefficient is used as an evaluation indicator to select IMF components with rich fault feature information. Then, calculate the multi-scale permutation entropy (MPE) of the selected IMFs and extract the nonlinear dynamic features of the signal. Finally, based on the MPE value, a winding mechanical fault diagnosis model with SSA optimized KELM is established. The experimental results show that CEEMDAN-MPE can successfully characterize the nonlinear dynamic characteristics of the operating state of transformer windings, and the SSA-KLEM model established can effectively identify the types of transformer winding looseness and deformation faults, with a diagnostic accuracy of 98.75%.