Adaptive Verification Method for Safety Measures of Smart Substation Secondary System Based on Cosine Similarity and DNN

Abstract

The secondary safety measures of energy substations have the disadvantages of complexity, implied and unintuitive, and the correctness of safety measures is difficult to guarantee. Aiming at the characteristics of secondary system information digitization, an adaptive verification method based on cosine similarity and deep convolutional network (DNN) for Smart Substation secondary system security measures is proposed. Intelligent data processing (IED) monitoring data in the secondary system is preprocessed by means of a cosine similarity matrix to reduce the number of training samples. After deep neural network training and fine-tuning, the classification of online IED data and security measures is completed, thus completing the verification of secondary security measures. The proposed method provides a new method for the verification of secondary system security measures of a new generation of Smart Substation.