Transformer is an important device in power system. In order to evaluate the aging state of transformer timely and effectively

this paper proposes an ultrasonic feature extraction and diagnosis method for oil-paper insulation aging. Through accelerated thermal aging tests

oil-paper insulation samples covering the full life cycle were obtained. The ultrasonic signals of all samples were obtained by using the self-built insulating oil ultrasonic testing platform

and the corresponding polymerization degree of insulating paper was measured and the aging state was divided. The GASF image of the ultrasound signal was acquired using the Gramian angular summation field (GASF)

and the HOG features of the image were extracted based on the histogram of oriented gradient (HOG). An aging diagnostic model based on ultrasonic features of oil-paper insulation was established based on learning vector quantization (LVQ) neural network

and the optimal number of neurons in the competing layers was determined to be 9 by the trial-and-error method. The results show that the diagnostic accuracy of the diagnostic model is more than 90%

and the diagnostic method has the potential to monitor the oil-paper insulation

which has a certain academic value and engineering application significance.