Diagnosis of Transformer Winding Looseness Based on VMD Morphological Gradient Spectrum and BAS-RF

To effectively extract the state information of winding in the transformer vibration signal，a new method based on variational modal decomposition（VMD）and morphological gradient spectrum is proposed for extracting feature vectors. The beetle antennae search-random forest（BAS-RF）is utilized to recognize the dis⁃ charge types. First，the measured vibration signals of the transformer windings under three different loose states are decomposed by VMD to obtain several modal components. Then，the multi-scale morphological gradient spectrum is calculated to form the initial characteristic sample. In order to prevent the disaster of dimensionality，the dimension reduction of the feature vectors is carried out by the principal component analysis. Finally，the number of decision trees in the random forest and the depth of the trees are optimized to construct a classifier model using the beetle antennae search to realize the recognition of the loose state of the transformer winding. Experimental results show that this method can effectively extract the characteristic information of transformer winding looseness and has excellent anti-noise performance. The constructed BAS-RF model has a high recogni⁃ tion accuracy and recognition speed. © 2023 Nanjing University of Aeronautics an Astronautics. All rights reserved.