Fault Diagnosis Algorithm for Dry-Type Transformer Based on Deep Learning of Small-Sample Acoustic Array Signals

The normal operation of electrical equipment is related to the stability of the power system. The dry-type transformer, as an important part of the distribution network, directly guarantees that users can use high-quality electricity. At present, most of the fault diagnosis of dry-type transformers is limited to the detection and maintenance of power outages, and there are few studies on nondestructive testing of power outages. In this letter, the operation state of the dry-type transformer is judged by the small-sample acoustic array signal, and the highly correlated intrinsic mode components are extracted by empirical mode decomposition (EMD); the highly correlated intrinsic mode components are further denoised by combining the adaptive wavelet basis transform. Then, the Hilbert transform is used to fuse the multichannel signals to form the original eigentensor. The principal component analysis is used to reduce the dimensionality of the original eigentensor to reduce the feature information surplus. The improved residual network is used to classify different features of dry-type transformers. It is verified that the proposed method has a high accuracy of 97.8% under the premise of small-sample datasets, which is better than that of the same type of detection method and has good robustness.