192 / 2022-03-14 17:00:37

Transformer winding fault diagnosis using the vibration radar image feature mining method

power transformer,winding fault,vibration,radar iamge,feature mining

Power transformer, as the essential equipment of power transmission and voltage transfer, should be ensured to operate under the safe and stable conditions, which is the prerequisite for the power system high stability and more economic benefits. However, the transformer suffers unavoidably from lighting impulse, huge short-circuit current shock. The turn-to-turn short-circuit and winding deformation induced by the impulse will increase sharply the risk of total insulation failure and mechanical un-stability. Real-time online perceiving the winding condition, diagnosing the defect degrees and developing preventive measures in advance become a highlighted technical requirement. The vibration sensor technology, as a qualitative method of winding defect online monitoring, is still lack of enough research on the actual experiment simulation, feature mining and defect diagnosis. Therefore, we propose a method of the transformer winding fault diagnosis based on vibration radar image feature mining. In the paper, the 10kV shrunken transformer winding defect simulation experimental platform is built. And the experimental transformer is designed referring to the 110kV power transformer. The vibration signals on the upper end of the winding and the external hyper-sensitivity position are collected by the accelerator sensor. Depending on the basis of the frequent-domain signal under the normal and same load current operation condition, the differential data is converted into the radar image, which process can transfer low-dimensional sequence data into two-dimensional image data. The diagnostic features can be designed based on the global and local feature mining algorithms. Finally, the winding fault online diagnosis method based on the vibration radar image feature mining can be proposed.