Analysis of Frozen Data Anomaly and Update Method of Electromechanical Energy Meter Terminal based on Deep Learning

Abstract

In view of the lack of advanced and mature substation fault detection and facility detection technology, combined with the characteristics of the actual application environment of substation, a substation operating equipment autonomous monitoring and fault diagnosis detection system based on deep learning intelligent detection robot is proposed. That is, the deep learning algorithm, Big data analysis technology and patrol robot with HD camera are organically combined. The image information collected by the high-definition camera is fused with the data information collected by a variety of sensors, and then the fault tree and Big data analysis algorithm are used to carry out real-time intelligent detection and analysis of all equipment in the substation, and the early warning can be sent to the relevant equipment maintenance personnel in a timely manner. The experimental results indicate that, the number of input nodes in the fault tree is 7, the number of output nodes is 2, the number of center vectors is 14, the number of nodes in the basis function layer is 7, and the threshold of the basis function is set to 0.8257. In actual training, after 31 iterations, the training results can quickly converge to the target value, the training error meets the requirements, and the fault diagnosis accuracy reaches over 90%. It has been proven that the diagnostic performance of the system is good, achieving the expected design effect.