Abstract：In order to solve the limitations of current routine test items and test cycles in traditional transformer testing methods, a new method of transformer oil state monitoring based on multi-frequency ultrasonic is proposed. First, carry out multi-indicator comprehensive analysis of the results of oiling experiment and evaluate the status of the selected transformer oil samples, and then a multi-frequency ultrasonic device is used to transmit the ultrasonic waves of different frequencies to the selected transformer oil, and the ultrasonic receiver module obtained the ultrasonic parameters (wave speed, amplitude-frequency data, and phase-frequency data) in real time. The BP neural network is used to establish the mapping relationship between the ultrasonic spectrum parameters and the state of transformer oil and verify whether the multi-frequency ultrasonic equipment can monitor the status of transformer insulation oil in real time, accurately and comprehensively.The results show that the BP neural network can establish a good correspondence relationship between ultrasonic parameters and transformer oil status levels. The use of multi-frequency ultrasonic equipment for transformer oil online monitoring is an effective method.