A Substation Safety Measure Ticket Generation Algorithm based on Neural Network and Inference Engine
Yin Wu, Huicai Zhao, Xingfu Huang, Liqiong Huang, Yuyang Hu · 2025
Intelligent substations use fiber optic as the information transmission medium, transforming the interconnection relationship between secondary devices into fiber optic connections. The corresponding secondary circuits and hardware platforms are replaced with virtual circuits and software templates, resulting in a significant change in maintenance work methods and higher requirements for operators' maintenance work. This article proposes an intelligent substation safety measure ticket generation algorithm based on multi-scale feature convolutional neural network (MSF-CNN) and inference machine. Firstly, a neural network model suitable for safety ticket generation was designed and trained. Combining the structural characteristics of DenseNet, multi-scale feature extraction and logical relationship modeling were achieved through multi-layer convolution, concatenation, and pooling processing. The model uses hyperbolic tangent function as the activation function to improve the accuracy of regression prediction. Secondly, a custom word vector and equipment association modeling method based on SCD file parsing are proposed for the electricity proprietary terms and multi item logical relationships involved in the safety measures ticket text, in order to improve data expression accuracy and model understanding ability. In addition, based on the functional correlation and input-output interaction characteristics of substation equipment, a three-layer neural network-based inference engine was constructed to generate maintenance safety measures for primary and secondary equipment. The experimental results showed that the MSF-CNN model performed the best in accuracy, recall, and F1 score, with an accuracy of 92.5%, recall of 90.8%, and F1 score of 91.6%.