Transformer fault diagnosis based on LIF-GADF technique and EfficientFormer-TL algorithm

As a key bridge between generator and grid, the stable operation of power transformer is crucial to the safety and reliability of power system. The traditional dissolved gas analysis (DGA) method has problems such as difficulty in fault feature extraction and poor accuracy in practical application, and a more accurate diagnostic technique is urgently needed. To this end, this paper proposes a new transformer oil fault diagnosis method based on laser-induced fluorescence (LIF) technology. First, the fluorescence wavelength data in transformer oil is collected using LIF technology, and then the Gram Angle Field (GADF) algorithm is used to transform the one-dimensional spectral data into a two-dimensional image, which is then input into the improved EfficientFormer algorithm combined with migration learning for classification and identification. Through this method, four types of samples, including thermal fault oil, electrical fault oil, localized moisture oil and original transformer oil, can be effectively identified, and the accuracy rate in fault diagnosis is as high as 99.00%. The experimental results show that compared with other common deep learning methods, the proposed method can significantly improve the accuracy and robustness of fault diagnosis, which fully validates its potential and application value in transformer fault diagnosis. This study provides new ideas for transformer fault detection and has strong practical application prospects, which can further improve the fault warning capability and operational security of the power system.