Hybrid Fault Diagnosis Method of Transformer Based on Double Antenna RFID Sensor and Deep Learning

The double-antenna RFID sensor and deep learning method are used to realize the online diagnosis of hybrid faults for transformer. The double-antenna RFID sensor is used to realize the acquisition and wireless transmission of the vibration signal generated by the transformer operation. The equipment has the advantages of low cost, low power consumption and easy to realize positioning. Transformer mixed fault vibration signal contains multi-source noise, showing strong nonlinear and non-stationary characteristics. SSDA model has significant advantages in complex nonlinear data processing due to its introduction of sparse penalty term, so this paper uses SSDA model to realize the feature extraction of mixed fault signals. The training data of SSDA also plays a key role in its feature learning performance. The mode function obtained by the VMD method has high accuracy and contains all the original information, and is used to calculate the envelope spectrum signal as the training data of SSDA in this paper. Finally, the parameters in VMD and SSDA were optimized by QPSO algorithm to achieve the best fault diagnosis performance. The experimental results show that the effective communication distance of the RFID sensor proposed in this paper reaches 17.5m, and the hybrid fault diagnosis accuracy reaches 99.8%.