Feature Extraction Method of Converter Transformer Vibration Signal Based on Multi-feature Fusion

This paper proposes a method for extracting vibration signal characteristics of converter transformers based on multi-feature fusion, in order to effectively identify operating faults of converter transformers. This method uses the actual measured vibration signal of the converter transformer as the experimental signal, carries out EMD (empirical mode decomposition) decomposition, and extracts the waveform index, peak index, margin index, kurtosis index and wavelet entropy of the IMF component to form the features Parameter matrix. Then PCA is used to reduce the dimensionality of the matrix to form the eigenvectors, and finally the fuzzy C-means clustering is used to classify the eigenvectors under different working conditions of the commutation transformer. Experimental results show that the vibration signal feature extraction method based on multi-feature fusion overcomes the limitation of single feature and has good distinguishing performance under different working conditions.