Power transformer fault diagnosis method based on vibration signal and deep learning
Aiming at the problem of low accuracy of real-time diagnosis of power transformer mechanical faults,this paper proposes a power transformer fault diagnosis method based on vibration signal and deep learning.Firstly,the vibration signal on the surface of the power transformer case is decomposed by the Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(ICEEMDAN)to obtain the reconstructed signal,and the fuzz-y entropy value is introduced to construct the vibration eigenvectors.Then,a Convolutional Neural Network-Bidi-rectional Gated Recurrent Unit(CNN-BiGRU)is used to form a basic classification network to achieve feature clas-sification,and an Efficient Channel Attention Mechanism(ECAM)is introduced to improve the CNN learning per-formance.Finally,a Multi-strategy Co-optimization Bald Eagle Search(MSCOBES)algorithm is designed based on the hybrid improvement of ICMIC chaotic mapping,adaptive dynamic perturbation and elite inverse learning,and the improved algorithm is applied to realize hyper-parameter optimization of CNN-BiGRU to obtain the optimization of power transformer fault diagnosis based on MSCOBES-CNN-BiGRU-ECAM model.In the experiment for the test transformer,the experimental results show that the proposed method can reach an accuracy up to 99.4%for the power transformer with different types of mechanical fault.
power transformerfault diagnosisICEEMDANCNN-BiGRUMSCOBESICMIC chaotic mappingadaptive dynamic perturbationelite inverse learning
万方数据
维普
