AI and computer vision driven fault detection framework for enhancing response accuracy in transmission and distribution production command systems

Reliable transmission and distribution production command increasingly relies on intelligent visual systems capable of accurately interpreting inspection imagery under complex environmental conditions. Conventional vision-based grid monitoring approaches are sensitive to illumination variations, limited in feature discrimination, and dependent on manually tuned parameters, resulting in false alarms, missed detections, and inconsistent response accuracy. To address these limitations, this study proposes the Intelli Bald Eagle Enriched Attention Block YOLOv8 Network (IBE-ABYv8-Net) to enhance production command reliability through optimized visual perception and adaptive learning. A real-world dataset consisting of 1600 transmission and distribution inspection images, collected under diverse environmental and operational conditions, is used for model development and evaluation. Contrast enhancement and histogram equalization are applied to improve visual clarity and normalize illumination inconsistencies. ORB-based local feature descriptors are employed to capture rotation-invariant structural characteristics of grid components, while deep convolutional features extract high-level semantic representations of faults and equipment conditions. YOLOv8 serves as the real-time detection backbone, enabling multi-scale feature extraction for accurate fault localization and classification. An enriched multi-scale attention block is integrated to emphasize fault-relevant regions and suppress background interference, improving the detection of small and visually subtle defects. Bald Eagle Optimization (BEO) automatically tunes critical hyperparameters, including learning rate, confidence threshold, and IoU threshold, ensuring stable and adaptive performance without manual intervention. The model achieves 98.94% accuracy outperforming IF-YOLO and CNN-RNN models. BEO-driven optimization reduces false alarms under uneven illumination and cluttered backgrounds, ensuring reliable real-time response accuracy in AI-driven transmission and distribution production command applications.