Research on Digital Technology of Mechanical Structure Fault Diagnosis of Power Transformer Based on Comprehensive Feature Extraction and SABO-PNN

Most of the current fault diagnosis based on vibration analysis method has fixed feature vectors and is prone to diagnostic errors. To identify the transformer structural faults more effectively, a mechanical structure fault diagnosis method for power transformers based on integrated feature extraction and SABO-PNN is proposed. Firstly, double feature extraction is performed on the original vibration signals based on wavelet packet transform and optimized variational modal decomposition based on the Subtraction-Average-Based Optimizer (SABO). Then, weight coefficients are assigned to the vibration feature vectors based on the fuzzy analytical hierarchy process, and the combined eigenvalues are calculated from the feature vectors and fuzzy weights. The integrated eigenvalues are used as input vectors to improve the probabilistic neural network using the SABO algorithm and train the vibration signals for diagnosis. Based on digital twin technology, a vibration signal monitoring platform is constructed, which can display vibration signals and diagnostic analysis results in real time. Based on the experimental prototype of 400 kVA transformer, the vibration signals from multiple measurement points of the tank under different fault conditions were collected and diagnostic analyses were performed. The test results show that the method can accurately diagnose the mechanical structure faults of the transformer.