In this paper

image data obtained from the video monitoring system of high voltage disconnector are taken as the research subject and the opening and closing condition identification of the disconnector is taken as the research target

the an image identification algorithm for the intelligent monitoring system of high voltage substation is proposed. For keeping the sample data to possesses diversified features

a variety of data enhancement techniques is adopted to process the sample set in the pre-processing module

which not only enhances the feature quantity

but also expands the dimension of sample expression. In the training module

network structures of different depths are constructed based on ResNet deep learning neural network

models with layers of 18

34

50

101 are obtained after training and the performance of each data model is assessed and compared on the validation sets. In addition

several other common neural network models are constructed and performance comparison with the method proposed method in this paper is made. The test results show that the proposed method can reach the identification accuracy of 99.7% on the validation sets and show advantages in performance. This method can provide algorithm support for disconnector monitoring system and meet the performance requirements of engineering applications.