An Intelligent Decision-Support Framework for Substation Fault Management Using BOA-Optimized Deep Learning and IoT-Based Image Processing

Abstract

Achieving operational efficiency and reliability in power substations becomes increasingly difficult as system complexity rises, necessitating the deployment of advanced monitoring and control technologies. The proposed Intelligent Power Auxiliary Control and Monitoring System integrates Internet of Things (IoT) technology with two-dimensional image processing to facilitate real-time monitoring and fault detection within substation environments. The system employs IoT-based sensors to capture critical auxiliary control parameters, including voltage, current, temperature, and equipment status. For image-based analysis, the YOLOv8 algorithm is utilised as an object detection mechanism, enabling precise identification of substation components and anomalies. Deep learning analysis is conducted using a Residual Neural Network (ResNet), which supports high-accuracy fault recognition through comprehensive monitoring of system parameters. The ResNet’s performance is further refined through weight parameter optimisation via the Butterfly Optimisation Algorithm (BOA), which improves convergence speed and classification accuracy. The system's effectiveness is validated through empirical analysis using actual substation data, demonstrating improvements in both fault detection accuracy and operational responsiveness. Evaluation findings confirm that the BOA-optimised ResNet model outperforms conventional deep learning approaches in terms of diagnostic accuracy and computational efficiency. The research contributes to the development of autonomous, intelligent auxiliary control systems capable of enhancing the safety and stability of substation operations. To aid in maintenance scheduling and operator decision-making, the system incorporates a fuzzy rule-based decision layer that interprets predictive outputs and initiates context-aware operational responses