Dynamic identification method of insulation defects in distribution transformers based on continuous data visualization

Abstract:
In order to realize the dynamic identification of insulation defects of distribution transformers with minimal errors, a method based on continuous data image is proposed to improve the accuracy of dynamic identification. The defect data is transformed into plane image structure data by continuous data imagery method, and input into convolution neural network to extract defect feature vectors. The feature vectors are input into the support vector machine classifier, and the training sample set is obtained by normalization. The linear and nonlinear classification results of samples are obtained after the training of vector machine, and the insulation defect recognition of distribution transformer is realized. The results show that when the number of iterations is more than 120 and the kernel function is 20, the method has the best effect and the accuracy is higher than 0.976, and the average accuracy is 0.9855, and the error is less than 0.15.