Real Time Monitoring Technology of Substation Equipment Status Based on Convolutional Neural Network

Detecting substation equipment failures at an early stage is challenging due to subtle abnormal changes that traditional monitoring methods may overlook. To address this, a convolutional neural network (CNN) algorithm model is proposed for real-time monitoring of transformers, circuit breakers, and switch devices in substation equipment. The study preprocesses sample data by considering factors such as grid harmonics, clutter, magnetic field, ripple, temperature, and humidity, which influence equipment operation. The CNN model's convolutional and subsampling layer parameters are optimized to balance feature extraction and model complexity. Experimental results show that the algorithm's loss value gradually decreases with iterations, stabilizing at 0.1000 after 100 epochs, with test accuracy reaching 1. However, some sample categories are misclassified. Comparative analysis with decision tree, Support Vector Machine (SVM), and k-Nearest Neighbors (KNN) algorithms demonstrates CNN's superior performance in accuracy, F1 score, false positive rate, training time, and parameter count. This study highlights CNN's potential for enhancing substation equipment monitoring systems.

I. Introduction

In recent years, Convolutional Neural Networks (CNNs) have emerged as a robust deep learning framework, demonstrating exceptional performance in image processing, speech recognition, and other domains. CNNs are particularly adept at extracting complex data features, making them an ideal candidate for processing heterogeneous data from substation equipment monitoring. Despite these advantages, existing studies predominantly focus on monitoring single devices or specific types of faults. This narrow scope limits their applicability in scenarios requiring comprehensive monitoring and classification of multiple device statuses. Motivated by these challenges, this study aims to address the limitations of existing technologies by developing a CNN-based real-time monitoring model tailored for substation equipment. This model is designed to achieve high accuracy and efficiency in identifying the operational status of transformers, circuit breakers, and switch devices, thereby advancing multi-class fault classification capabilities.