Power Grid Fault Diagnosis Method Based on Improved Inception-ResNet Model Graph Semantic Extraction

Wei Xing, Ling Zheng, Jiayin Bai

Abstract: Traditional power grid fault diagnosis methods have problems such as large parameters, poor real-time performance, susceptibility to malformed data interference, and low accuracy. In order to solve the above problems, a power grid fault diagnosis method based on improved Inception-ResNet graph semantic extraction is proposed. Real-time monitoring of each node of the power grid based on PMU measurement data. The semantic reconstruction of PMU data is realized by converting data features into image features, thus completing the conversion from quantitative analysis of data form to qualitative analysis of image form, greatly reducing the influence of malformed data on fault diagnosis. The improved Inception-ResNet algorithm is used to extract the semantic features of the fault image, and the corresponding fault type is obtained according to the fault feature and the fault diagnosis result is output. Experiments and simulations show that this method can effectively reduce training parameters and training time, shorten diagnosis time, and improve diagnosis accuracy.

Keywords: PMU Graph Data, Power Grid Fault Diagnosis, Deep Learning, SE-Inception-ResNet