Saving Birds from Power Grids: A Deep Learning Embedded System for Sustainable Energy Infrastructure
DOI:
https://doi.org/10.5281/zenodo.18621524Keywords:
High-voltage transmission lines, Convolutional Neural Networks, Bird Detection, YOLO-v2 Network, Edge ComputingAbstract
Energy sustainability remains one of the humanity’s most pressing challenges. Electrical power systems, while essential for modern infrastructure, pose significant risks to wildlife, particularly birds. When birds approach to high-voltage power grids, they face fatal electrocution, leading to ecological harm and potential damage to electrical infrastructure. To mitigate this issue, we propose an embedded system that combines low-cost hardware with deep learning for real-time avian detection and deterrence. Our solution employs the YOLO-v2 neural network, implemented on a K210 embedded processor featuring dual 64-bit cores running at 400 MHz, achieving an inference speed of 45 frames per second. The model was trained on a curated dataset of 1,500 diverse bird images, optimized for network input, and achieved an accuracy of 89% during validation. This system demonstrates a practical, efficient, and scalable approach to reducing avian fatalities near power grids while maintaining grid reliability
Downloads
References
[1] H.-T. N. N. T. a. K. C. M. Vo, "Bird detection and species classification: using YOLOv5 and deep transfer learning models.," International Journal of Advanced Computer Science and Applications 14.7 , vol. 14, no. 7, 2023.
[2] K. a. Z. S. Yang, "Deep learning-based object detection improvement for fine-grained birds," Ieee Access, vol. 9, pp. 67901-67915, 2021.
[3] C. a. Y.-q. L. ". d. o. t. l. b. o. D.-Y. m. I. I. P. X. 1. I. T. 1. I. C. I. 2. H. C. J. 3. 2. P. 1. S. I. Zou, "Bird detection on transmission lines based on DC-YOLO model.," in Intelligent Information Processing X: 11th IFIP TC 12 International Conference.Springer Internationa, Hangzhou, China, 2020.
[4] J. a. A. F. Redmon, "YOLO9000: better, faster, stronger," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017.
[5] A. G. Howard, "Mobilenets: Efficient convolutional neural networks for mobile vision applications," arXiv preprint arXiv, no. 1704.04861, 2017.
[6] B.-S. e. a. Lin, ""Fall detection system with artificial intelligence-based edge computing," IEEE Access, no. 10 , pp. 4328-4339, 2022.
[7] Kendryte, Kendryte, [Online]. Available: https://www.canaan-creative.com/zh/proDetail/210.
[8] Gareth, "kaggle," [Online]. Available: https://www.kaggle.com/datasets/nelyg8002000/birds-flying.
[9] "kaggle," [Online]. Available: https://www.kaggle.com/datasets/cubeai/bird525-classification-for-yolov8.
[10] TianzhongSong, "github," [Online]. Available: https://github.com/TianzhongSong/Tensorflow-quantization-test.
[11] D. a. M. E.-S. Sinha, "Thin mobilenet: An enhanced mobilenet architecture.," in 2019 IEEE 10th annual ubiquitous computing, electronics & mobile communication conference (UEMCON). IEEE, 2019.
