A bird target detection model designed for substation scenarios.

Area of Science:

• Electrical Engineering
• Computer Vision
• Wildlife Management

Background:

• Bird activity poses a significant threat to power grid reliability, causing failures and outages.
• Existing bird detection methods are often expensive and lack sustained efficacy.
• Substation security is crucial for uninterrupted power supply.

Purpose of the Study:

• To develop an efficient and accurate bird detection system specifically for substation environments.
• To enhance the reliability and safety of power grids by mitigating bird-related incidents.
• To introduce a novel deep learning model for bird target detection in critical infrastructure.

Main Methods:

• Proposed YOLO-birds method utilizing Faster-BiFPN for optimized feature extraction and fusion.
• Incorporated SPPBiF attention mechanism to improve detection of varied scales and small objects like birds.
• Employed Focal-EIoU loss function to enhance robustness against low-quality image samples.
• Created a custom dataset of bird images relevant to substation security threats.

Main Results:

• YOLO-birds achieved a mean Average Precision (mAP50) of 90.2%, demonstrating superior performance.
• The developed method effectively detects birds in substation environments.
• Validated effectiveness against existing bird detection techniques.

Conclusions:

• The YOLO-birds method presents a highly effective solution for detecting birds in substations.
• This technology can significantly aid in preventing bird-induced power grid accidents.
• Improved substation security and power grid reliability are achievable through this advanced detection system.