The multi-frequency ultrasound technique for transformer oil parameter detection is given. The corresponding hardware detector and upper system software are designed and developed. Based on the formulated multi-frequency ultrasound detection technique, a large number of ultrasound signals with different frequencies could be gathered in a scanned monitoring frequency in a second by the hardware part, then the state diagnosis of transformer is finished via hundreds of scannings in a short time. The design principle of system and main functional units are given, and the latter mainly includes the transmitting part and receiving part. Next, the multivariate statistical analysis algorithm is elaborated, and based on which, the ultrasonic parameters that contain the operating information of transformers are made statistics and analyzed by the upper software, to get the ultrasound characteristic values, between which and ten situations of transformer, further, relationship is built, so as to judge the operating situation of transformers. The experimental detection shows that the developed system can detect the transformer oil parameters such as the water content, breakdown voltage (BDV) and active water online and in real time. Thus the proposed transformer oil detection method based on multi-frequency technique solves the problems of cumbersome detection steps and failed online testing and has advantages of sensitive detection and lower cost, when it is compared with the traditional detection systems.