Abstract: In the power grid system, the fault of oil-immersed transformer can affect the power supply lines. The traditional manual testing cannot meet the needs of transformer fault detection. Therefore, research on the application of intelligent fault diagnosis for oil-immersed transformer is carried out. Taking a substation transformer as an example, two neural network algorithms, BP and QIA-BP,are selected to carry out oil-immersed transformer fault diagnosis research. The efficiency and accuracy of BP neural network detection are compared with those of manual detection, and the tracking effect of QIA-BP neural network is compared with that of BP neural network. The accuracy of the QIA-BP intelligent diagnosis system is tested by white box test, which verifies the effectiveness and superiority of QIA-BP neural network intelligent fault detection for transformer fault diagnosis.