An online 11 kv distribution system insulator defect detection approach with modified YOLOv11 and mobileNetV3

Area of Science:

• Electrical Engineering
• Computer Vision
• Artificial Intelligence

Background:

• Smart distribution grids require robust insulator defect detection to prevent power loss and cascading outages.
• Challenges include complex backgrounds, limited datasets, and small-scale defect detection.
• Deep learning offers a promising solution for automated insulator inspection.

Purpose of the Study:

• To develop a novel, lightweight deep learning model for detecting insulator defects in 11 kV distribution systems.
• To improve the accuracy and efficiency of insulator defect detection using unmanned aerial vehicles (UAVs).
• To validate the proposed model's performance against existing methods.

Main Methods:

• A modified YOLOv11 architecture incorporating C3K2, SPFF, and C2PSA blocks was developed.
• A MobileNetV3 classifier was integrated for a lightweight framework suitable for distribution automation systems (DAS).
• Data augmentation techniques were employed during the preprocessing phase.

Main Results:

• The proposed model demonstrated effectiveness in detecting insulator defects in real-life scenarios.
• Performance comparisons showed superiority over earlier YOLO versions and other existing methods.
• Case studies confirmed the robustness and reliability of the detection method.

Conclusions:

• The novel YOLOv11 and MobileNetV3 based approach provides an efficient and accurate solution for insulator defect detection.
• The lightweight framework is suitable for deployment on DAS-enabled devices for automated grid monitoring.
• This method contributes to enhancing the reliability and safety of smart distribution grids.