Detection Method for Loose Points in Transformer Iron Cores Based on HO and FMD

Transformers are indispensable components of modern power systems, and timely identification of core loosening faults is of great significance for ensuring reliable operation. In this study, a novel diagnostic approach is introduced that combines Hippopotamus Optimization (HO) with Feature Modal Decomposition (FMD). The HO algorithm is employed to optimize the parameter selection of FMD, enabling the extraction of representative fault-related features from vibration signals. These features are subsequently classified through a Temporal Convolutional Network (TCN). Experimental investigations indicate that the proposed method can accurately distinguish four loosening levels of the transformer core (0%, 20%, 50%, and 100%), achieving a classification accuracy of 95%. The results confirm the efficiency and robustness of this technique for transformer fault detection.